Datasets:

WilliamWen

/

summarization_cata

like 0

The Heck coupling reactions has become one of the most useful C(sp2)-C(sp2) bond-forming reactions in organic synthesis [1–5]. The reaction has been applied to many areas, including bioactive compounds, natural products, drug intermediates, fine chemical syntheses, UV absorbers, antioxidants and industrial applications [6–9]. However, existed protocol gained increasing importance both in large-scale industrial processes and the development of new materials and biologically active compounds [10–12]. But in the other hand, existed protocol suffers from the inherent drawback of the required pre-synthesizing the organic halides and accompanied formation of a stoichiometric amount of hazardous halide salt, which can cause significant environmental concerns.Furthermore, the examination of ionic liquid functionalized novel catalytic systems involving other transition metals such as Co [13–15], Ni [16,17] and Cu [18,19] have also been receiving much attention other than palladium (Heck reaction) catalysts [20–23]. In general, copper based catalytic systems are easily available because of their cheaper price, functional group tolerance and large-scale procedures. Nevertheless, only limited reports are available in the copper catalysed Heck type reaction and are suffering the problems concerning extraction from the reaction mixture, waste production, high toxicity & price, air-sensitivity and leaching [24]. Henceforth, we felt it would be of keen interest to eliminate these negative aspects of the copper complex. In this regard, we have developed a stable, selective, suitable ligand that leads to efficient heterogeneous copper complex with high turnover and reprocessability.In the last two years, Nemanja Vucetic et al., have described bis-layer supported ionic liquid catalyst with an unprecedented activity in the Heck reaction [25] and Daniel Rauber et al., have been reported fluorinated phosphonium ionic liquid in Heck reactions [26]. Furthermore, Saithalavi Anas et al., have proposed polymer supported copper catalyst for the Heck reaction [27]. However, Issa Yavari et al., has illustrated copper-catalysed Mizoroki-Heck coupling reaction using an efficient and magnetically reusable Fe3O4@SiO2@PrNCu catalyst [28]. All the above techniques provide excellent yield, but some have disadvantages such as lengthy work-up procedure, harsh reaction conditions (organic co-solvents) and requires absolutely dry and inert media. To the best of our knowledge, no one has reported copper functionalized 1-glycyl-3-methyl imidazolium chloride (Fig. 1 ), which proved to be highly efficient at ambient temperature for C–C bond formation under solvent free condition.Though, recovery and leaching can occur in the extractive work up leading to a loss of the catalyst in the reaction mix on the one hand and requests supplementary effort to purify the extracted product. To overcome such problems, novel complex was developed [29] by covalent linking of organo catalytic unit with an ionic liquid moiety (often chloroglycine). This imparts a low solubility of catalyst in the solvents used for extraction of the product on the one hand and high solubility in the reaction medium on the other hand [29]. This strategy was applied to Heck reaction providing high yield and good recyclability of the organo catalyst.The objectives of the present study are to: (i) develop an efficient synthetic process for the facile conversion of Heck reaction. The present method developed for the Heck reaction offer many advantages including high conversion, solvent free and the involvement of non-toxic reagents.At the initial of this project, the influence of various reaction parameters has been investigated for the 1-iodo-4-methoxybenzene and styrene as model reaction like best catalytic system, temperature, solvent and base. To begin, the impact of different bases to the model reaction was studied and these results are summarized in Table 1 . Initially, no product was found when the absence of base (Table 1, entry 1). However, reaction was carried using several bases with 14%–55% yields (Table 1, entries 2–9). Furthermore, Et3N is superior in comparison to various bases (Table 1, entry 10). After inventing the best catalyst systems, we further optimized the reaction conditions in presence of Et3N. The experiments showed the time decreasing from 24 ​h to 10 ​h gave the same results (Table 1, entries 11 and 12). But the yields were dropped appreciably the time decreasing from 10 to 8 ​h (Table 1, entry 13). Further, the amount of the base was increased from 2 mmol to 3 ​mmol, the yield increased 96%–97% (Table 1, entry 14), in the other hand dissimilar yield was achieved when decreasing the base amount from 2 mmol to 1 ​mmol (Table 1, entry 15). By all these investigation results, it was concluded that 2 ​mmol Et3N is sufficient to brought out the complete Heck reaction in 10 ​h at 25 ​°C.Next, in order to find an appropriate solvent for the model reactions, the coupling of 1-iodo-4-methoxybenzene and styrene was carried out with different solvents and Et3N. Among the reports, in the absence of solvent was the most productive, as compared with the polar and non-polar solvents (Table 2 , entries 1–11). This may be due to the easy coordination of complex with organic co-solvents. It has also been reported that H2O molecule sometimes is required to activate the Cu(II) catalyst. In our case, carrying out the reaction in water gave a negative effect on the product yield in comparison with solvent free condition (Table 2, entry 14). This lower yield could be due to complex delocalization under aqueous condition. However, temperature also plays an important role in the model reaction. When we conducted the Heck reaction of 1-iodo-4-methoxybenzene and styrene as the model substrate at 30 ​°C, there was no change in the yield (Table 2, entry 12).Subsequently, in order to optimize the reaction conditions for a particular catalyst, the coupling reaction of 1-iodo-4-methoxybenzene and styrene was performed in presence of Et3N by using different catalysts and the results are given in Table 3 . Unfortunately, no product was detected in the absence of catalyst (Table 3, entry 1). To our delight, when using a various copper salts, the coupling reaction gave trace to 14% (Table 3, entries 2–6). Following, we studied the activity of synthesized 1-carboxy ethyl-3-methyl imidazolium bromide [Cemim]Br, 1-(2-aminoethyl)-3-methylimidazolium bromide [Aemim]Br, 1-glycyl-3-methyl imidazolium chloride [Gmim]Cl catalyst and it was initiate that [Gmim]Cl catalysed furnished less yield of desired product (Table 3, entries 7–9). An additionally we tested the combination of CuCl2 and the different ionic liquids led to significant activity to the model reaction (Table 3, entries 10–12). However, Among them heterogeneous [Gmim]Cl–Cu(II) catalyst was found to be the best catalyst providing 96% yield in 10 ​h (Table 3, entry 13). After, similar yield was reached when escalating the catalyst amount from 0.1 to 0.2 ​mol% (Table 3, entry 14) and it was found that conversion decreases with decrease in catalyst loading (Table 3, entry 15). Later, the reaction conditions for the active catalyst system were further optimized the reaction duration. The testing showed that on increasing the time from 10 to 11 ​h gave the same (Table 3, entry 16). Though, the yields were dropped significantly on decreasing the time from 10 to 9 ​h (Table 3, entry 17). In contrast to other protocols, this coupling reaction can be performed even at 30 ​°C, while at 25 ​°C only 10 ​h reaction time is necessary to complete the Heck coupling reaction without any loss (Table 3, entry 18).With suitable conditions in hand, we examined the generality and limitation of the catalyst system. As shown in Table 4 , more than 22 different stilbenes derivatives were synthesized in moderate to excellent yields. More specifically, 9 different aryl chlorides were tested: para- and ortho-methyl substituted chlorobenzenes gave the corresponding products (Table 4, entries 2 and 3). Similarly, 4-nitro and 4-methoxy substituted aryl chlorides were successfully transformed in 82%–84% yields (Table 4, entries 4 and 5). In addition, aryl bromides with electron-withdrawing substituent (CN) also work well under our standard conditions with 730 TON (Table 4, entry 6). Next, we used methyl acrylate olefins. When 4-methyl and 4-nitro substituted aryl chloride are treated with methyl acrylate under the optimized reaction condition, substituted stilbene product was isolated only in 78% and 83% yields (Table 4, entries 8 and 9). However, 7 different aryl bromides were tested: aryl bromide showed good reactivity with other activated substituted groups such as para-(methyl/methoxy/chloro) and afforded the corresponding coupling products in moderate to good yields (Table 4, entries 10–13). Moreover, 4-methoxy and 4-methyl substituted aryl bromide are treated with methyl acrylate gave substituted products in reasonable yields (Table 4, entries 15 and 16). Likewise, functionalized aryl iodides containing both electron withdrawing and electron donating groups felt efficient coupling reaction with styrene. Nevertheless, para-methoxy and methyl substituted aryl iodides reacted well with styrene gave the corresponding products in excellent yields (Table 4, entries 18 and 19). Additionally, excellent reactivity was observed, when para-CN-aryl iodine were subjected to coupling with styrene (Table 4, entry 20). While, reaction of 4-methoxy iodo benzene with methyl acrylate resulted in 87% yield (Table 4, entry 22). Notably, in all the cases, complete trans form was detected and the formations of cis or homo coupling products were not observed.Isolation of the heterogeneous catalyst was easily performed by extraction or centrifugation. The isolated catalyst was washed with diethyl ether and dried in air. The regenerated catalyst was used for the reaction of 1-iodo-4-methoxybenzene with styrene for nine runs to afford trans-stilbene with 96%–91% isolated yields (Table 5 ). The precise mechanism of the catalytic reaction needs to be elucidated, but it is noticeable that the mechanism is strongly modified depending of the halobenzene employed, obtaining trans-stilbene as the main product (Scheme 1 ).In conclusion we have introduced [Gmim] Cl–Cu(II) as a catalyst for Heck reaction with Et3N in the absence of an organic co-solvent. Aryl iodides were reacted efficiently with styrene and methyl acrylate at 25 ​°C in the presence of the catalyst. Noteworthy features of this catalyst system are (1) its catalytic activity was tested in Heck reaction; (2) 0.1 ​mol% of catalyst was sufficient to furnish the trans-stilbenes with excellent yields (up to 96%). (3) The catalyst can be readily recovered and reused without significant loss of its activity.All solvents and chemicals were commercially available and used without further purification unless otherwise stated. The 1H NMR spectra were recorded on a Bruker 500 ​MHz using CDCl3/DMSO‑d 6 as the solvent and mass spectra were recorded on JEOL GC MATE II HRMS (EI) spectrometer. FT-IR was recorded on AVATRA 330 spectrometer with DTGS detector. Column chromatography was performed on silica gel (200–300 mesh). Analytical thin-layer chromatography (TLC) was carried out on precoated silica gel GF-254 plates.[Gmim]Cl–Cu (II) complex was synthesized following the literature [30].In a conical flask (100 ​mL), a mixture of 1-iodo-4-methoxybenzene (1 ​mmol), styrene (1.2 ​mmol), triethylamine (2 ​mmol) and [Gmim]Cl–Cu (II) (0.1 ​mol%) was added and stirred at 25 ​°C for a period as indicated in Table 4 (The reaction was monitored by HPLC and TLC). The heterogeneous mixture was extracted with ethyl acetate or diethyl ether (3 ​× ​5 ​mL). The organic phase was separated and dried over anhydrous Na2SO4 and evaporated. The resulting crude was purified by flash chromatography to give the desired pure product with excellent yield.1-Iodo-4-methoxybenzene (1 ​mmol) was reacted with styrene (1.2 ​mmol) in the presence of [Gmim]Cl–Cu(II) (0.1 ​mol%) and triethylamine (2 ​mmol) at 25 ​°C. After completion of the reaction (TLC/HPLC), the product was extracted as stated in the preceding general method. The white solid [Gmim]Cl–Cu(II) was isolated by centrifugation. Furthermore, the recovered complex was washed with diethyl ether and dried in air. The resulting catalyst was charged to another batch of the similar reaction. This was repeated for 9 runs to complete the reaction in 10 ​h to give the desired product with 96%–91% yield (Table 5).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 the Management of PC-Campus for providing required facilities and SIF (IITM) for providing the spectral data.

A novel 1-glycyl-3-methyl imidazolium chloride-copper(II) complex [[Gmim]Cl–Cu(II)] was found to be a heterogeneous catalyst for an efficient Heck reaction with good to excellent yield under solvent free condition. This protocol provides a simple strategy for the generation of a variety of new C–C bonds under environmentally benign condition. The catalyst was reused up to nine consecutive cycles without any significance loss in its activity.

Today, lower quality fuels (high viscosity) containing high amounts of sulfur are extracted due to decreasing oil reserves [1]. As it is known, when fuels are burned, organic sulfur compounds in them are oxidized and emit SO2 gas which is harmful to the atmosphere and the environment, and these gases cause acid rain and corrosion [2]. Therefore, it is of crucial importance to desulfurize these low quality fuels.Hydrodesulfurization (HDS) as conventional desulfurization is widely used in the world. In HDS, organic sulfur compounds react with H2 gas and H2S is released as a result of the carbon–sulfur bond cleavage in organic compound [3]. However, HDS has some disadvantages [4–6]: the use of high temperature, high pressure, expensive H2 gas and expensive catalysts with high chemical stability and high thermal resistance that must not be affected by severe operating conditions. Also, aliphatic sulfur compounds are easy to remove in HDS, while refractory aromatic sulfur compounds are difficult to remove [7].To eliminate these disadvantages, alternative desulfurization processes such as adsorptive desulfurization [8], extractive desulfurization [9], oxidative desulfurization [10], biodesulfurization [11] are used. Among them, the most advantageous and promising method is oxidative desulfurization (ODS). In ODS, at relatively low temperatures such as 20–60 °C, at atmospheric or near atmospheric pressures, organic sulfur compounds are oxidized by using H2O2 and a catalyst to convert first to their sulfoxides and then to their sulfones, which are more polar compounds, and finally these oxidized sulfur compounds are removed from the fuel by extraction with a polar extractant such as methanol, acetonitrile, dimethyl formamide etc. or by adsorption [12].Desulfurization is also carried out with simultaneous oxidation and extraction [13]. In HDS, it is difficult to remove aromatic sulfur compounds, especially alkyl-substituted aromatic sulfur compounds which are prevented from accessing into the catalyst pores due to steric hindrance [14]. On the contrary, in ODS, using a liquid homogeneous catalytic system such as formic acid or acetic acid- H2O2 (HP) oxidant [15], alkyl-substituted aromatic sulfur derivatives are easier to remove due to an increase in electron density [1,16–19] on the sulfur atom as shown in Fig. 1 . In particular, bonding the naphthenic ring to the thiophenic ring significantly increases the ODS yield of the compounds such as THBNT, THDBT and OHDNT [20]. When the phosphotungstic acid-HP system is used, the molecular size of the catalyst becomes important. Since phosphotungstic acid is a bulky molecule [21], the ODS reactivity of aromatic sulfur compounds having alkyl groups adjacent to the sulfur atom decreases due to spatial obstacle [22]. In a study [23] in which ODS of model sulfur compounds was performed by phosphotungstic acid-HP, it was reported that sulfur removal decreased in the order DBT > 4-MDBT > 4,6-DMDBT. When the solid heterogeneous catalyst is used, the sulfur atom is prevented from entering the catalyst pore and its interaction with the sulfur atom due to the steric hindrance of alkyl groups adjacent to sulfur becomes weak, consequently causing a decline in the ODS reactivity [24]. Desulfurization using t-butylhydroperoxide in the presence of Mo/Al2O3 catalyst is in the order DBT > 4-MDBT > 4,6-DMDBT≫ BT [25]. With the use of TiO2 anatase-supported V2O5 catalyst and HP, the ODS yield is in the order DBT > BT > 4-MDBT > 2-MT > 2,5-DMT > 4,6-DMDBT [26]. In the H3PW12O40/TiO2-HP system, the desulfurization at 30 °C increases in the order 4,6-DMDBT < BT < DBT [27].In ODS reactions, the mixture consists of two immiscible liquid phases as organic phase (real fuel or model fuel solution containing sulfur compounds such as DBT, 4,6-DMDBT dissolved in a non-polar solvent such as n-hexane, n-heptane or iso-octane) and aqueous phase (H2O2 solution). Therefore, quaternary ammonium salts as phase transfer catalysts (PTCs), one end of which is hydrophilic and the other end hydrophobic, are generally used, reducing the liquid–liquid interface tension [28] and enabling the transfer of oxidizing species to organic phase, so that the ODS increases significantly [29]. Sometimes using ionic liquid (IL) instead of aqueous phase, ODS is further increased such that the IL acts as extractant during oxidation [30]. For the last 20–30 years, ultrasound wave has been used to accelerate oxidation reactions and increase ODS more. Sonication has two simultaneous effects in accelerating ODS reactions. The ultrasound wave creates cavitation bubbles in liquid and the implosion of these bubbles produces very high temperatures and pressures locally in the liquid. At the extremely high temperatures, chemical bonds of organic compounds are broken and reactive radicals are generated (Sonochemical effect). Microjet, microturbulence and shock waves created by imploding cavitation bubbles significantly accelerate the mass transfer by increasing the emulsification of the organic and aqueous phase (Sonophysical effect). Thus higher desulfurization efficiencies are achieved in a shorter time [31].ODS reactions are generally heterogeneous reactions, i.e., there are two or more phases in the mixture that are immiscible with each other. The solution of the organic phase, which is formed by dissolving model sulfur compounds in a non-polar solvent such as hexane, heptane or toluene, has been referred as to denotations such as model fuel, model diesel, model liquid fuel, model sulfur solution. The aqueous phase consists of an oxidant and a catalyst. In many studies, the reactivity of the model sulfur compounds has been determined and the optimum conditions (temperature, oxidant volume, catalyst amount, organic phase/aqueous phase volume ratio, time etc.) for maximum desulfurization have been found. These conditions have then been applied to real fuels to achieve desulfurization.Many solid, liquid and gas oxidizers have been evaluated. Inorganic chemicals such as oxone [32], sodium persulfate [33], potassium superoxide [34], potassium dichromate [35], sodium percarbonate [36], sodium perchlorate [37], hydrogen peroxide [38], sodium hypochlorite [39], solid oxidizers such as cyclohexanone peroxide [40] and organic chemicals such as t-butylhydroperoxide [41] and cumene hydroperoxide [42] as liquid oxidizers are used. The most distinctive feature of cyclohexanone peroxide as solid organic oxidizers and cumene hydroperoxide and t-butylhydroperoxide as liquid organic oxidizers is that they can all dissolve in the organic phase or fuel, thereby directly oxidizing sulfur compounds [43,44]. The structural formulas of oxidizing substances are shown in Fig. 2 . Gaseous oxidants are generally oxygen [45], nitrogen dioxide [46] and ozone [47], and the solubility [48–51] of these gases in non-polar solvents is generally higher than that in water.Catalysts used in ODS are divided into two types; homogeneous catalysts soluble in liquid phase and heterogeneous catalysts insoluble in liquid phase.Catalysts used in heterogeneous catalysis are solid and insoluble in liquid mixture [52]. Nanoparticles improve the adsorption of sulfur compounds due to their large surface area [53]. Photocatalyst under UV [54] or visible light [55], nano-sized silica particles including mesoporous silica [56], aluminum oxide particles [57], transition metal oxides [58], activated carbons [59], modified metal–organic frameworks [60], Ni catalyst also called sponge metal [61], nanocomposite [62], graphene oxide [63], activated carbon (AC)-supported phosphotungstic acid [64] and fly ash-modified fenton catalysts [65] are used. In the case of using heterogeneous catalysts, the catalytic ODS mechanism [66–70] is illustrated in Scheme 1 . DBT, which is transferred from the organic phase to the aqueous bulk phase by ultrasound, diffuses to the outer surface of the solid catalyst by passing across the liquid film (boundary layer) around the supported catalyst particle. DBT is adsorbed on active sites on the external surface of the catalyst or on active sites on the internal surface of the inner pores by diffusing through the pore. HP interacts with active sites on the inner and outer surface and forms oxidizing active complexes. After DBT adsorbed on these active centers is converted into its sulfones by undergoing an oxidation reaction, DBT sulfone is desorbed and transferred successively to the boundary layer, aqueous phase and organic phase. In addition to enhancement of adsorption and desorption, ultrasound significantly increases not only the external and internal diffusion but also the collision frequency of reactants with active sites, thus causing increased UAODS performance.Matsuzawa et al. [71] carried out the photocatalytic oxidation of DBT using a Hg-Xe lamp of 200 W at wavelength > 290 nm in the presence of anatase-type TiO2 (P25) as a heterogeneous photocatalyst and air (in which oxygen acts as an electron scavenger [72], thus causing oxidation only by electron vacancy (h+) [73] of TiO2 [74–76]) in a polar acetonitrile solution. They found the photooxidation rate in combination with H2O2, photocatalyst and indirect ultrasound (45 kHz and 50 W) was higher than the oxidation rate with H2O2 and photocatalyst, and this effect was due to the reactivation of the TiO2 surface and increased mass transfer. However, they stated that the direct oxidation rate of 4,6-DMDBT using only H2O2 under photoirradiation was higher than the photooxidation rates in the cases of HP-photocatalyst-US and HP-photocatalyst. In addition, it is reported that the oxidation reaction rate of the methyl group in 4,6-DMDBT increased by using aliphatic and cyclic alkanes as a non-polar solvent instead of the polar solvent acetonitrile, since oxygen was more soluble in non-polar solvents [77].In another study [78], using photocatalytic anatase TiO2, 30 wt% HP and ultrasound with duty cycle, the catalytic oxidative desulfurization of gum turpentine, which is similar to crude sulphated turpentine and a by-product of Kraft process [79] to obtain wood pulp, spiked with dimethyl disulfide was investigated and a desulfurization efficiency of 100% was found at 28 °C, 120 W power dissipation and 20 kHz US frequency, 70% duty cycle, 15 g L–1 HP concentration, 4 g L–1 TiO2 loading for 50 ppmw DMDS initial concentration. Also, it was reported that total treatment cost (0.31 $ L–1) with (US + H2O2 + TiO2) system is less as compared to US, only 30 wt% H2O2, only Fenton, only TiO2, US + 30 wt% H2O2, US + Fenton and US + TiO2. In addition, the authors investigated the effects of individually US/Fenton and US/TiO2 processes on desulfurization, but it was found that the desulfurization efficiencies of those processes were lower than the desulfurization efficiency of the US/HP/TiO2 process. It has been explained that the reason for the very high desulfurization of the US/HP/TiO2 process is the production of more hydroxyl radicals from HP along with the support of the TiO2 catalyst and the generation of additional hydroxyl radicals as a result of the increase of active sites by deformation of the catalyst under US. It was also stated that homogeneous distribution of the catalyst particles and better mixing due to the high turbulence caused by the collapsed bubbles enhance the sulfur removal. Cavitational yields (4.65 × 10–9, 4.71 × 10–9 and 6.61 × 10–9 g J−1 for US/Fenton reagent, US/TiO2 and US/HP/TiO2, respectively) were calculated by the authors to confirm the differences in desulfurization in the three processes. In this study, it was determined that the total cost of the other treatment methods was 2.22, 43.12, 14.69, 17.50, 1.255, 0.70 and 0.595 $ L−1 for US, only HP(30%(v/v), only Fenton, only TiO2, US + HP(30%(v/v), US + Fenton and US + TiO2, respectively. Although a high sulfur removal is obtained from gum turpentine in the presence HP and TiO2 under US, oxidative desulfurization of DMDS as an aliphatic sulfur compound is quite easy, the initial sulfur quantity (50 ppm DMDS) is very low, and the reaction time is 120 min. Therefore, it is not a favorable method.In the studies performed by Yu et al. [80] and Zhao et al. [81], sonophotocatalytic oxidative desulfurization of hydrotreated diesel oil and model diesel oil using CdO as semiconductor and H2O2 as oxidant was investigated and desulfurization efficiencies were found to be 72.7 and 99.47%, respectively. The high desulfurization in the latter under 20 kHz and 150 W US can be attributed to primarily the use of the model sulfur solution prepared by dissolving the organosulfur compound in a solvent instead of hydrogenated diesel fuel, which consists of a complex mixture of aliphatic hydrocarbons and aromatic hydrocarbons [82], and acetic acid to increase the oxidizing power of H2O2 and secondarily to the catalyst with a smaller grain size (i.e. larger surface area) and more homogenized structure, which is synthesized under ultrasound [83], hence causing a higher catalytic activity.Behin and Farhadian [84] performed the ODS (followed by extraction with a binary solvent of methanol and water in ratio of 1:1 in volume) of nonhydrotreated kerosene with a total S content of 1553 ppmw at 0.05 cm s−1 superficial gas velocity for 15 min. by passing ozone as a homogeneous photocatalyst through an airlift reactor and using H2O2 under both US of 20 kHz frequency (60 W power) and UV in a wavelength range of 280–400 nm. Despite a 48% loss of aromaticity due to ozone, and to a lesser extent polar solvent, a desulfurization efficiency of 91.7% was reached. It is revealed that the high desulfurization yield at optimum conditions was due to HO· (oxidation potential [85] of hydroxyl radical, 2.80 V) and HO2· (oxidation potential [85] of hydroperoxyl radical, 1.7 V) radicals formed in the mixture during the reaction rather than the increased mass transfer and the physical properties of raw kerosene are almost unchanged.In addition, sonolysis of sulfur compounds in water was carried out at high ultrasonic frequencies without using catalysts and oxidants. The dilute solution containing 21.46 ppm S BT in water was subjected to sonodegradation at 21 °C under 352 kHz and 80 W US, and it was explained that the dominant mechanism was the oxidation of BT as a result of the formation of hydroxyl radicals from water [86]. However, in the sonolysis of a dilute T solution containing 32 ppm S in water at 22 °C under 850 kHz and 40 W US power, it was revealed that the dominant mechanism was pyrolysis as a result of high temperature caused by collapsed cavitation bubbles rather than hydroxyl radical formed in the medium since T can diffuse readily into the cavitation bubble due to T's lower boiling point (i.e., more volatile) than BT [87].AOPs were utilized in combination with sonolysis. Despite high desulfurization under both US and UV or visible light in AOPs [71,80,81], where photocatalysts are used, these high desulfurization yields were reached in 6,5 and 3 h, respectively, for the respective studies. In photocatalysis, a light energy such as UV or visible light is absorbed by photocatalyst (e.g., TiO2), and the electron is excited by passing from the valence band to the conduction band, and thus an electron-hole pair is formed on photocatalyst. The positive electron holes (h+) react with the water adsorbed on the catalyst to produce hydroxyl radicals. In addition, oxygen on catalyst surface reacts in series with the excited electron (e–) to produce hydroxyl radicals and also US generates hydroxyl radical from HP. Consequently, enhanced hydroxyl radical production renders sonophotocatalytic ODS yield high. The reactions are as follows [88,89]: Photocatalyst → + h ν e - + h + H 2 O + h + → H + + HO• 2 H 2 O + 2 h + → H 2 O 2 + 2 H + H 2 O 2 → ) ) ) 2 H O · O 2 + e - → O 2 · - → 2 H + + e - H 2 O 2 → H + + e - H O · + H 2 O In the Sono-Fenton process, FeSO4 is used along with HP under US irradiation. In the Fenton reaction, Fe2+ is first oxidized by HP to produce the HO· radical and then the reaction of Fe3+ with HP produces the complex intermediate Fe-OOH2+ which decomposes rapidly to form HO2· radical and Fe2+ under US [90]. Fenton reaction is substantially accelerated by US [91]. As a result, sulfur removal further increases due to enhancement of hydroxyl radicals in organic-aqueous phase interfacial area. The medium must be acidic to maximize production of free radicals [92]. The reactions in the Sonofenton process are as follows [93]: F e 2 + + H 2 O 2 → F e 3 + + H O · + O H - F e 3 + + H 2 O 2 → F e - O O H 2 + + H + Fe - O O H 2 + → ) ) ) F e 2 + + H O 2 · As noted above, reaction times are very high in studies [71,80,81], where photocatalyst was used. Therefore, this will lead to higher electrical energy consumption for US, UV and heating, if any, increasing the operating cost in AOP.In the study [84], in which ozone and HP were used as oxidant under US-UV, it was explained that the reason for high sulfur removal in a short time was indirect hydroxyl radical production from O3 and direct hydroxyl radical from HP by UV and US. In addition, it is stated that ultrasound greatly accelerates the gas–liquid mass transfer through micro-streaming produced by the violent collapse of bubbles and allows ozone to react with sulfur compounds by increasing the gas–liquid interfacial area. Moreover, dissolved ozone gas acts as nucleation sites to form cavitation bubbles, causing the formation of more cavitation bubbles [94]. Thus, this synergistic effect accelerates significantly the ultrasound-assisted photo oxidative desulfurization reaction rate.In a sonophoto-fenton process [95] in which oxalic acid was used, a sulfur removal of>93% was achieved from 100 ppm DBT in toluene at 0.05 mol L−1 Fe2+ concentration, 0.15 mol L−1 oxalate concentration, pH = 2, a volume ratio (organic phase/HP) of 10:1, 25 °C and 15 min under both 37 kHz, 95 W indirect US and UV in the presence of air. It was revealed that FeII(C2O4), which is formed by the reaction of Fe2+ with oxalate anion (C2O4 2−) in the reaction medium, as well as Fe II ( C 2 O 4 ) 2 2 - complex which is formed by the reaction of FeII(C2O4) with C2O4 2−, is responsible for this high desulfurization. The authors reported that FeII(C2O4) and Fe II ( C 2 O 4 ) 2 2 - caused the formation of HO·, HO2· and O· radicals in the aqueous phase to oxidize DBT under US and UV irradiation. It was stated that Fe(II)-oxalate complex as catalyst can be reused three times (a decrease of 1.33 and 1.56% for the first and second run, respectively) without significantly losing its activity by regenerating it after each reaction.The effect of solid catalysts to increase ODS has also been studied [96], and it was found that the use of US for total desulfurization of 2,3-DMBT and 2,3,7-TMBT, which are the two most abundant components in JP-8 fuel, in the presence of H2O2, formic acid and phosphoric acid-activated carbon increases the total desulfurization in the absence of US (mechanical stirring) by around 2.4-fold. It is also reported that desulfurization by chemically activated carbon (MW-99) with phosphoric acid is superior to desulfurization by thermally activated carbon (Norit SX-1) due to the larger surface area of MW-99 and the greater number of its surface acid centers. Sulfur removal of 98 and 94% (followed by adsorption with activated alumina) from JP-8 and diesel, respectively, was performed with MW-99 under optimum conditions (65 °C, 2 h, 60% amplitude, 20 kHz sonication, pH = 1.4).Khlaif and Bded [97] carried out the ODS (followed by extraction) of crude oil containing 1.95% total S by weight in the presence of US and AC using different volumes of acetic acid and 50 wt% H2O2. As a result of the increase of the amount of AC used from 3 to 9 g, the number of active sites in AC increased, thus improving ODS and an optimum desulfurization of 81.325% was obtained by using 9 g AC, 40 mL H2O2, 30 mL acetic acid at 50 °C.Using phosphotungstic acid (H3PW12O40@ TMU-17-NH2) incorporated in robust zinc-based MOF with enhanced efficiency as a solid catalyst, simultaneous extraction and oxidation of model oil containing BT, DBT and 4,6-DMDBT, each of which has concentration of 500 mg L–1, were performed in the presence of acetonitrile under indirect sonication of 37 kHz [98]. Although the pore volume and surface area (137 cm3 g−1 and 814 m2 g−1) of the composite MOF catalyst formed by encapsulating H3PW12O40 in TMU-17-NH2 were lower as compared to those of the neat MOF (239 cm3 g−1 and 1050 m2 g−1), a sulfur removal of 98, 87 and 71% was reached with 20 mg of the MOF composite containing 20 wt% phosphotungstic acid at model oil/MeCN 1:1 vol ratio, O/S ratio of 2:1 and room temperature for DBT, 4,6-DMDBT and BT, respectively, at the end of 15 min. The reason for the lower reactivity of 4,6-DMDBT compared to DBT is that the alkyl substituted aromatic compound is sterically prevented from entering the 3D framework. Also lower desulfurization was achieved with DMF solvent instead of MeCN depending on the fact that adsorption of solvent on the heterogeneous catalyst increases with increasing boiling point [99,100] and polarity [101–103]. The low desulfurization with DMF can be attributed to the fact that not only the boiling point of DMF (153 °C) is significantly higher than that of MeCN (82 °C) [104] but also higher polarity [105] of the former compared to the latter causes stronger interaction with Zn2+ in the modified MOF composite [106], thus reducing adsorption of DBT. The former is bound to Zn2+ cations in the MOF composite [106]. The three possible adsorption mechanisms [107,108] are π-π interaction between sulfur compounds and aromatic rings of modified MOF, hydrogen bonding between NH2 groups and S, and strong Zn2+-S interaction between phosphotungstic acid-TMU-17-NH2 and aromatic sulfur compounds. TMU-17-NH2 is probably structurally similar to TMU-16-NH2 with positive zeta potential [109]. H2O2 and aromatic sulfur compounds are adsorbed on the catalyst, the phosphotungstic acid anion is oxidized with hydrogen peroxide and as a consequence, the polyoxoperoxo complex anion formed oxidizes aromatic sulfur compounds [110]. In addition, water in the reaction medium can result in the radical decomposition of H2O2 by forming an aqueous complex with Zn2+ in Zn(II)-based MOF, hence generating a strong oxidant radical HO·[111] and electrophilic activation of hydrogen peroxide to convert sulfur compounds to their sulfoxides as oxidized sulfur compounds is caused by Zn-based MOF [112].Metalloporphyrin [113] and metallophthalocyanine [114] catalysts, which are metal complexes, are also used in ODS reactions. Metal removal from the latter is not easy compared to the former [115]. The degree of ODS can be changed by adding different electron-withdrawing or electron-donating substituents to these complexes [113,116]. In addition, the stability of these complexes can be increased by forming nanocomposite catalysts, thus ensuring that they can be reused in oxidation reactions [117].Wang et al. fulfilled two separate studies [118,119] concerning sonocatalytic ODS (followed by extraction with methanol) of benzothiophene in the presence of H2O2 at 60 °C using core–shell nanosphere modified with metallophthalocyanine (tetra-substituted carboxyl iron phthalocyanine, FeC4Pc) encapsulated into magnetic mesopore silica nanoparticles and silica nanotube catalyst with magnetite nanoparticles-coated interior surface and FeC4Pc-modified inner and outer surface. Higher desulfurization of the former (at the same conditions, desulfurization near 94.5%) compared to the latter (76% desulfurization yield at 30 min and molar ratio of H2O2/S = 15) can be considerably clarified by the fact that the particle size (60 nm) and the average pore size (2.6 nm) of the nanosphere composite catalyst are smaller than the outer diameter of the nanotube catalyst (200 nm), hence providing larger surface area for adsorption, though the catalyst loading is not specified in the latter. In these two studies, it was reported that high desulfurization is due to the radical decomposition of H2O2 to HO· on metallophthalocyanines. HO· radical from H2O2 by ultrasound wave can also be formed [120]. It is also stated that both catalysts can be easily isolated from the mixture by applying an external magnetic field after the reactions due to their superparamagnetic properties and reused in the next reactions.Uniform Ni skeletal catalyst was synthesized at a size of 2.5–10 µm under 90 kHz ultrasound and crude oil containing 2.645% S by weight is subjected to oxidation with two treatment cycles using a mixture of ozone-air and 0.2% by weight catalyst based on the oil volume for 5 min in a US bath with frequency of 22 kHz [61]. Sulfur removals from gasoline and diesel fractions in crude oil were found to be 52 and 27.4%, respectively, as well as improvement of gasoline and diesel fractions.By using 0.5 g of the modified GO/COOH solid catalyst with increased surface acidity formed by the addition of –CH2COOH group to the epoxy or hydroxyl groups of GO as a result of the reaction of graphene oxide (GO) with chloroacetic acid, a desulfurization of 95%, which is higher than desulfurization in the case of using non-acidified GO, was performed from the DBT solution containing 1000 ppm S with 30 wt% H2O2 within 300 min on sonication [121]. It was put forward that the adsorption-oxidation mechanism is the conversion of DBT to DBT sulfone by the peroxyacid group formed on the GO/COOH surface via activation of H2O2 by the carboxyl group in GO, and then π-π interaction of DBT sulfone with GO/COOH and adsorption of DBT sulfone through hydrogen bonding. In addition, it was stated that ultrasound contributes to high desulfurization due to the increase in the surface area caused by the exfoliation of GO/COOH as well as the increased collision frequency of the reactants due to the significantly increased mass transfer.As phosphotungstic acid hydrate as oxidizing agent is dissolved in the aqueous phase, thus making it difficult to be reused by recovery [122], activated carbon-supported phosphotungstic acid (PTA) catalysts were synthesized and two separate studies [123,124] were carried out on UAODS of 2000 ppmw DBT. In the first study [123], a DBT conversion of 93.4% was reached using 40 mL of model oil, at PTA/AC-10 catalyst/model oil 1.25: 100 mass ratio and H2O2/model oil 0.1 vol ratio under 70 W US power at 60 °C and 10 min, while in the second study [124] under the same conditions except the use of US at 100 W power, DBT conversion well below the conversion reached in the first study was obtained. The reason for the low conversion can be attributed to the weakening of the ultrasound wave (bubble shielding effect) as a consequence of absorption and scattering of US waves by these bubbles by resulting in the formation of dense cavitation bubble cloud around the transducer under high power [125]. Therefore, an optimum power intensity is needed as an important factor for high conversion in liquid phase reactions. In both studies, it was reported that desulfurization improved due to the increase in the number of surface acid sites by the increase in the amount of phosphotungstic acid in AC, and beyond a certain phosphotungstic acid amount, the sulfur removal is unchanged due to the reduction in surface area as a result of the destruction of microchannels in AC and the occupation of pores in AC by phosphotungstic acid.In a similar study [126] where the same catalyst (HPW/AC-10) was synthesized, the optimum conditions were determined using RSM for reasonable desulfurization of the model oil containing 2800 ppm S consisting of a mixture of DBT, BT and T in the presence of individually, 30, 20 and 10 wt% H2O2 at different catalyst quantities, different AP/OP volume ratios and different times under 37 kHz US. By applying these optimum parameters to kerosene with 1370 ppmw S, a 99% desulfurization was successfully achieved, followed by four-cycle extraction.In a study [127] where O2 in air was used as oxidant instead of thermally unstable H2O2, modified heteropolyacid catalysts (H5PV2Mo10O40/SiO2 and H5PV2W10O40/SiO2) supported on silica were synthesized. At optimum conditions (catalyst weight/model oil volume 11.09 g L–1, POM weight /SiO2 (wt. %) 39.879, sonication time 199.209 min.) found using the response surface method at 65 °C and 1.3 L min−1 air flow rate, a higher desulfurization (90 vs. 70%) of DBT was achieved in a shorter time (199 vs. 360 min.) under 20 kHz and 360 W direct US compared to the desulfurization in the case in which ultrasound is not used. It was demonstrated that the reason for low desulfurization is the polymerization of DBT due to the low concentration of oxygen dissolved in the organic phase (limited aerobic medium) under magnetic stirring, thus causing the polymer formed to accumulate on the modified heteropolyacids. While this polymerization is thought to be probably initiated by the DBT cation radical formed as a result of electron transfer from DBT to vanadium incorporated heteropolyacid [128], it was found that US increases the dissolved oxygen concentration and prevents polymer deposition on the catalyst surface. DBT conversion 10% more with H5PV2W10O40/SiO2 than the conversion percentage with H5PV2Mo10O40/SiO2 was obtained since the standard reduction potential of V5+ and W6+ (1 and −0.090 eV, respectively) is higher than that of Mo6+ (-0.913 eV), thus having stronger oxidizing power [129,130]. The oxidation mechanism [131,132] in the UAODS system can be elucidated by the electron transfer-oxygen transfer (ET-OT) reaction, in which oxygen is involved, between the modified heteropolyacid and DBT as follows: C 12 H 8 S + P V 2 5+ W 10 O 40 5 - → ET C 12 H 8 S + ∙ + P V 5+ V 4+ W 10 O 40 6 - C 12 H 8 S + ∙ + P V 5+ V 4+ W 10 O 40 6 - → OT P V 5+ V 4+ W 10 O 39 6 – – O – C 12 H 8 S P V 5+ V 4+ W 10 O 39 6 – – O – C 12 H 8 S → OT C 12 H 8 SO+ P V 2 4+ W 10 O 39 5 – P V 2 4+ W 10 O 39 5 − → Oxidation by oxygen + O 2 + 2 H + P V 2 5+ W 10 O 40 5 - + H 2 O Model oil with 1000 ppm total S content containing BT, DBT and 4,6-DMDBT was sonicated at 300 W, 45% amplitude and 20 kHz fixed frequency using 30 wt% H2O2 in the presence of MoO3 supported on γ-Al2O3 catalyst for 30 min [133] and at the optimum conditions (H2O2/S = 3 molar ratio, 45 °C, 30 g L–1 catalyst/model oil ratio) found by RSM with central composite design, a DBT → DBT sulphone conversion above 98% was found. Moreover, a desulfurization improvement of over 95% was achieved for DBT even after 6 cycles without losing the catalyst effect, due to US, which prevents the agglomeration of catalyst particles and H2O2 and causes desorption of adsorbed polar sulfones and water impurities from the catalyst surface. For BT, DBT and 4,6-DMDBT, the highest desulfurization was achieved when the MoO3 content on the catalyst was 10 wt% and at this loading, it was proved by XRD analysis that MoO3 is homogeneously dispersed on the support and MoO3 crystals are not seen. It was suggested that the sulfur compounds are oxidized by highly reactive molybdenum peroxide and molybdenum diperoxides formed in situ.In a similar study [134] where the same reagents and the same ultrasonic parameters were used, complete oxidation of DBT in the model oil containing 600 ppmw total S was achieved in the presence of MoO3 loading of 10 wt.%/Al2O3 at H2O2/S = 3.8 molar ratio, 30 g L–1 catalyst/model oil ratio, 45 °C and 30 min. Besides, the addition of aromatic compounds (tetralin, naphthalene and 2-methyl naphthalene) individually to the model oil formed by dissolving DBT in hexane to mimic diesel fuel appreciably reduced the UAODS yield although the resulting DBT selectivity is high due to the competitive adsorption of the aromatic compounds on the catalyst surface. Further, in both studies [133,134] it was shown that the active sites responsible for the adsorption of sulfur compounds are tetrahedrally coordinated Mo6+ oxides, above a Mo-saturated monolayer coverage (which is at 10 wt% Mo loading), agglomeration of amorphous MoO species results in the formation of MoO3 crystals and cause a reduction in the number of active sites, as well as the reduction of surface area, by blocking micropores of the catalyst [135], thus reducing the UAODS.Using persulfate agent in toluene and hexane as solvent, 98 wt% H2O2 and 1% Si-Al/Al2O3 as solid catalyst, 99.72% of sulfur (followed first by extraction with acetone, then by adsorption with activated charcoal and ultimately by sonication under 30 kHz US of the diesel sample treated with acetic acid) in hydrotreated diesel fuel containing 766.73 ppmw total S was removed at around 65 °C and atmospheric pressure [136]. It can be thought that the oxidation mechanism [137] is based on sulfate ion radical caused by thermal activation of persulfate, hydroxyl radical formed as a result of the reaction of sulphate ion radical with H2O2 and activation of S2O8 2− by hydrogen peroxide, which causes the formation of hydroxyl radical. Moreover, US can cause homolytic cleavage of the persulfate agent [138] and hydrogen peroxide [90]. The surface hydroxyl groups [139] on Al2O3 (Fig. 3 ) in the solid catalyst in the reaction medium can induce the formation[137] of S O 4 ∙ - radical from persulfate by interacting with H+ formed by the reaction (4) and, hence accelerating the UAODS reaction. (1) S 2 O 8 2 - → 2 S O 4 · - (2) S 2 O 8 2 - + H 2 O 2 → 2 S O 4 · - + 2 O H · (3) 2 S O 4 · - + H 2 O 2 → 2 S O 4 2 - + 2 O H · (4) S 2 O 8 2 - + O H · → S O 4 2 - + S O 4 · - + 1 / 2 O 2 + H + Since homogeneous Fenton catalysts (FeSO4) dissolve in the aqueous phase and consequently, making their recovery difficult [140] after ODS reactions, water-insoluble Fenton-like catalysts supported on coal fly ash (which is a very cheap waste from coal-fired power plants) were synthesized [141]. Approximately 30% desulfurization was carried out as a result of simultaneous oxidation and extraction of sulfur compounds from commercial diesel fuel containing 595 ppm S using 10 wt% H2O2 and ethanol solvent in the presence of the Fenton-like catalyst in an ultrasonic bath at 47 kHz frequency and 147 W power [65]. It has been suggested that the oxidation stems from the hydroxyl radicals formed from the reaction between Fe2+ and H2O2. Hydroxyl radicals [90] formed from the decomposition of H2O2 by US may also contribute to this desulfurization. Furthermore, since coal fly ash contains metal oxides [142], H2O2 helps desulfurization by being adsorbed on the supported catalyst as well as forming surface-bound hydroxyl radicals on the support [143].US has also been applied to oil sands [144] as an oil deposit consisted of a mixture of clay, sand, bitumen and water. A total sulfur removal efficiency of 82% has been reported by simultaneous oxidative and extractive desulfurization of semi-solid Alberta bitumen containing 5.2 wt% S using 3 wt% H2O2, saturated NaOH and tetrahydrofuran under a 28 kHz frequency and 200 W powerful indirect ultrasound at 20 °C and 20 min [145]. Then, an 88% bitumen recovery from oil sand and a 42% sulfur reduction from bitumen was fulfilled using the same reagents, the same reaction conditions and ultrasonic parameters simultaneously. In addition, possible metalloporphyrins [146–148] in bitumen can accelerate the UAODS reaction of bitumen. Moreover, it was stated that since ionic NaOH cannot dissolve oil sand sufficiently and effectively, mid-polar THF is used owing to its high dissolving power.The UAODS process was not limited to liquid fuels, but also applied to mesophase materials [149]. It was demonstrated a sulfur removal (followed by extraction with equal volumes of methanol and sodium hydroxide (0.5 wt%)) of 91.1% from coal tar pitch with 0.9 wt% S containing predominantly polycyclic aromatic hydrocarbons (also called polynuclear aromatic hydrocarbons) was carried out using xylene as dispersant and solvent, trichloroacetic acid as catalyst, 30 wt% H2O2 in the absence of surfactant under 20 kHz and 300 W direct US at 60 min. and 70 °C [150]. On the other hand, the use of surfactant did not increase UAODS.Apart from hydrogen peroxides, organic peroxide has also been used as oxidant. In this type of study [151], approximately 35% desulfurization (followed by extraction three times with acetonitrile) was performed from a high-viscosity bunker-C oil MFO 380 (max kinematic viscosity 380 cSt) with 3.17 wt% S using viscosity-reducing heptane and 3 mL of t-butyl hydroperoxide as oxidant in the presence of 0.2 g MoO3 as solid catalyst under direct US at a frequency of 20 kHz and 70% amplitude at atmospheric pressure, 90 min and 80 °C. Unlike HP, TBHP has the advantage of being soluble in both aqueous and organic phases, therefore, in desulfurization reactions where the aqueous phase is not used, it is in direct contact with sulfur compounds without the need for mass transfer. It was reported that the much higher-reactivity peroxo molybdenum complex formed as a result of the reaction of t-BHP with MoO3 is responsible for the oxidation of sulfur compounds to their sulfones. When ultrasonic cavitation bubbles in sonochemistry implode, very high temperatures and pressures occur locally in the liquid (hot spot theory) [152]. Therefore, it can be deduced that reactive oxygen species, which are generated by thermal decomposition of t-BHP in this reaction, such as t-butoxyl (H3C)3 - O·, hydroxyl HO· and t-butyl peroxyl (H3C)3 - O - O· radicals [153], further contributes to the oxidation of bunker-C oil.In the presence of heterogeneous catalysts with which sulfur compounds interact electronically on the solid surface, adsorption, where mass transfer is an important factor, takes place through catalyst pores [154], whereas homogeneous catalysts dissolve in liquid (ie, aqueous phase). After UAODS reactions, isolation, recovery and reuse of homogeneous liquid catalysts, as well as the homogeneous solid catalysts dissolved in the aqueous phase, from the reaction mixture are quite problematic since they are in the same phase as reactants, which increases the process cost [155].Reactions, in which homogeneous catalysts are involved, can be divided into two classes; 1) Reactions in the absence of PTC 2) Reactions in the presence of PTC. Among the homogeneous solid catalysts, catalysts such as phosphotungstic acid [156] as polyoxometallate class, Fe(II)SO4 [157] and CuSO4 [158] were employed, while organic acids such as acetic acid [159] and formic acid [160] were utilized as homogeneous liquid catalysts.In the absence of PTC, the ODS mechanism [161] is shown in Scheme 2 . Peroxyformic acid formed in situ by the reaction of HP and formic acid in aqueous phase is transferred to the organic phase where DBT is oxidized, by the effect of ultrasound.In a study [162] where the sonoreactor was optimized to increase the UAODS yield, a sulfur removal of 98.25% was achieved from model fuel containing 1000 ppmw DBT in n-decane using 16 mL of 34.5 wt% H2O2 and 40 mm-diameter sonotrode with an immersion depth of 3 cm at acetic acid/H2O2 64: 300 molar ratio in 7.4 cm-diameter glass reactor under 20 kHz, 500 W and 80% amplitude direct US at 48 °C within 30 min.UAODS of a model fuel containing 100 ppmw DBT (10.8 mM/l) in toluene was performed using FeSO4, acetic acid and 30 vol% hydrogen peroxide (HP) [163]. It was stated that the hydroperoxyl radicals formed were responsible for the oxidation of the sulfur compound rather than the hydroxyl radicals formed, hence by explaining that lower scavenging of HO2· radicals is important. An DBT removal of approximately 33.34 wt% from model oil has been reached at acetic acid/HP = 2 vol ratio, toluene/HP = 10 vol ratio, at 1.5 M Fe2+ concentration, 90 min and atmospheric pressure under 70 W and 35 kHz indirect US at 25 °C.In a similar study [164] in which desulfurization of benzothiophene (BT), 3-methyl thiophene (3-MT) and thiophene (T) was performed using 25 mL of 30 vol% HP + CH3COOH and Fe2+, sulfur removals of 79.4, 77.9, 77% − 76.3, 76.9, 77.6% and 77.5, 76.5, 76.1% were obtained from concentrations of 100, 300 and 500 ppm for BT, 3-MT and T, respectively, under 2.5 bar, 35 W and 35 kHz indirect US at 90 min and 25 °C, such that these conversions were higher than those obtained at atmospheric pressure due to the elimination of transient cavitations at high pressure. In addition, according to the cavitation bubble dynamics model, it was revealed that the high desulfurization is caused by the sonophysical effect (microconvection) of US.In a study [165] in which a sample of raw coal containing 2.16 wt% total S as solid fuel was treated with peroxyacetic acid, oxidative desulfurization of raw coal improved due to the increased reactivity of the coal depending on the increased specific surface area, the total pore volume and the mean pore size of ​​the treated coal compared to those of the untreated coal since abrasion of coal particles upon sonication occurs; 17.59% of the total sulfur present in the coal was removed using 10 mL 98 wt% acetic acid and 50 mL 30 wt% HP under 20 kHz and 720 W direct US at 30 °C within 5 min. It was shown that the greatest contribution to desulfurization is that US increases the production of hydroxyl radical in the presence of HP and acetic acid in the mixture, whereas the hydroxyl radical production rate is significantly low when there is only HP.In a similar study [166] in which the same reactants were used, the raw coal was subjected to ultrasonic treatment followed by microwave. The US applied reduced the particle size of the coal, increased its total porosity (i.e., specific surface area, total pore volume and average pore diameter of the raw coal are 0.88 m2.g−1, 0.00213 cm3.g−1 and 9.68 nm, respectively, whereas specific surface area, total pore volume and average pore diameter of the coal sample after US treatment are 1.66 m2.g−1, 0.00771 cm3.g−1 and 18.56 nm, respectively) and increased hydroxyl radicals. But at the same time, microwave increased the reaction rate dramatically as the reactants in the mixture absorbed the electromagnetic radiation generated [167]. At the end of the ultrasonic treatment at acetic acid (98 wt%)/HP (30 wt%) 1:5 vol ratio under 20 kHz and 720 W direct US for 50 min at 40 °C, followed by microwave treatment under 600 W power at a frequency of 2.45 GHz at 100 °C for 6 min, a desulfurization of nearly 22% was obtained from raw coal containing 1.93 wt% organic S, which results from the resonance nature of the thiophenic compound according to mercaptan and sulfoether, whereas the percentage of pyritic sulfur (in the form of FeS2) removed as inorganic sulfur was reported to be about 85%.In another study [168] using the same reactants, two coal samples (XS with 0.85 wt% organic S and YN with 2.69 wt% organic S) completely free of inorganic sulfur as a result of pretreatment with dilute nitric acid were subjected simultaneously to ultrasonic and microwave treatment with a power of 560 W each for 50 min. Sulfur removals of 23.53 and 76.58% were achieved for XS and YN, respectively. Consequently, it turns out that from these three studies concerning coal, simultaneous operation (US-MW) is more efficient.In desulfurization of model fuels prepared by dissolving model sulfur compounds in a non-polar solvent such as octane, heptane or hexane, an extraction step is not required since the sulfones as oxidized sulfur compounds are easily determined by instrumental devices such as GC-FID, HPLC, hence easily finding the conversion to sulfones. However, as there are also aliphatic and aromatic hydrocarbons in addition to sulfur compounds in real fuels, it is not possible to determine the sulfur compounds with these devices. After separating sulfones by an extractant, the total sulfur percentage in the fuel can be determined by using devices such as microcoulometric analyzer, sulfur analyzer with UV fluorescence, XRF and GC-SCD.Alkaline solutions have also been used in UAODS. In simultaneous oxidative and extractive desulfurization [169] of ultra low-sulfur diesel spiked with 500 ppmw 4,6-DMDBT, it has been shown that desulfurization in single step can be improved without an extraction step mainly due to the hydroxyl radicals formed as well as secondarily the formation of carbonate radical C O 3 - ∙ by resulting in radical decomposition of HP under US in the range of pH 6 ∼ 8 with basic sodium carbonate. Approximately 94% desulfurization was reported at diesel/acetonitrile 1: 2 vol ratio, 0.8 M HP 30 wt%, 30 mM Na2CO3 under 23 kHz frequency direct ultrasonic pulse at 60 °C in 2 h.As shown in Table 1 , the other studies [170–175] using acetic acid as organic acid in addition to HP are common in the literature. In addition, acetic acid is relatively low-cost [176]. In studies [97,165,177] in which desulfurization of crude oil, coal and model diesel fuel with the help of US by using acetic acid-HP oxidant system was performed, it was indicated that high desulfurization efficiency is reached in a short time at relatively low temperatures. The oxidation of sulfur compounds is caused by peroxyacetic acid and hydroxyl radicals formed in situ in the aqueous phase. It has also been shown that nitrogen compounds have an inhibitory effect on oxidative desulfurization as the oxidation reactivity of the nitrogen compounds present in the fuel (e.g. quinoline) is higher than that of the sulfur compounds [172]. Moreover, the effects of different US loop reactor types on UAODS were also examined [173]. It is stated that the aqueous phase separated after the UAODS reaction and the extractant separated after the extraction step can be reused for the fresh feedstocks containing 208 ppmw S DBT and the same feedstocks subjected to oxidation treatment, respectively, though the desulfurization efficiencies in reuses are lower than those in their first uses [171]. UAODS efficiencies of diesel fuel feeds containing different sulfur amounts in the presence of acetic acid under the relevant reaction conditions are shown in Table 1.One of the most important reasons why HDS is still widely used today is that fuel loss after HDS process is very low [178]. In laboratory-scale studies, after the ODS process, the properties of the fuel are almost unchanged [172,179–186], but the loss of fuel in the extraction step (i.e., the reduction of fuel recovery) after the ODS process on large scales can pose a major problem. Moreover, whether the properties such as density, viscosity, cetane number, boiling range distribution of the desulfurized fuel produced in large quantities (factory scale) have changed is a matter of investigation separately and must be checked one by one. In most research papers [171,187–189], when H2O2/S mole ratio initially increases, desulfurization generally increases, then reaches a certain value and decreases slowly after this optimum value. It was reported that this decline is due to dilution of the aqueous phase.In a study [172] in which nitrogen was removed by US from a synthetic fuel solution with 252 ppmw N prepared by dissolving quinoline in a hydrotreated petroleum product feed containing 3.6 ppm S, 92% nitrogen removal (followed by extraction with methanol) was achieved in the case where only acetic acid is used in the absence of HP as oxidant. It is stated that this value is higher than the value (79% nitrogen removal) obtained without oxidation treatment by only liquid–liquid extraction with methanol, hence underlining that acetic acid has the capacity to extract nitrogen compounds.The effects of different sonoreactor types on desulfurization and denitrogenation (followed by silica gel adsorption) of hydrotreated diesel fuel containing 241 ppmw S and 161 ppmw N were also evaluated [173]. It was shown that the most effective reactor in terms of cost and performance optimization was sonitube.In an oxidation study [190] accomplished under 20 kHz and 70 W direct US followed by extraction with DMF, it was stated that while the initial sulfur content in the model fuel containing DBT increased from 1220.80 ppmw to 3976.86 ppmw, desulfurization also increased to 98.35%. In the UAODS [175] followed by extraction, as acetic acid/oil ratio increased to 1.50 wt%, the desulfurization of diesel containing 849 ppmw S improved. This was attributed to the strong oxidant peracetic acid formed in situ.Heterogeneous reactions with solid–liquid systems using solid oxidants were also carried out. HP-acetic acid at S/oxidant 1:10 molar ratio, KO2-Acetic acid, Na2S2O8 alone, Na2S2O8-acetic acid and oxone alone at S/oxidant 1:10 and 1:30 molar ratios at different times at 80 °C were used [186] for UAODS of model oils and diesel fuel. Sulfur and nitrogen removal were individually performed by ultrasonic horn device under 21.1 kHz and 80 W direct US and ultrasonic cup horn device under 19.9 kHz and 80 W US from mild hydrotreated diesel feedstock containing 226 ± 2.17 ppmw total S and 158 ± 2.81 ppmw total N as well as three model solutions containing 1.2 mg mL−1 DBT or DMDBT and 1.2 mg mL−1 quinoline individually. In UAODS reactions of model solutions in both reactor types, when oxone alone is used at a molar ratio of S/oxidant = 1:30 without acetic acid, very high desulfurization efficiencies compared to other oxidant systems (100% sulfur removal for DBT and DMDBT in 90 min, a nitrogen removal of 40% for quinoline in the same time) were achieved. For scale-up purposes, the US cup horn was chosen as it closely resembles the geometry of continuous flow reactors and sulfur was removed (followed by SiO2 adsorption) from hydrotreated diesel fuel at molar ratios of (S + N)/oxidant 1:10, 1:20 and 1:30 by oxone at different times. In addition to obtaining a diesel fuel containing 0.91 ± 0.48 ppmw N (a nitrogen removal of 99.4%) at a molar ratio of 1:30 in 90 min, a sulfur removal of 99% was achieved. In the case of extraction with MeOH instead of adsorption, significantly low desulfurization (65%) was obtained for the same molar ratio and the same time, but diesel fuel recovery with SiO2 adsorption was lower than that with methanol extraction by 11%. It was stated that excess oxone can be reused for the same diesel fuel without losing its activity in four treatment cycles followed by adsorption with SiO2 each (from 84% sulfur removal at the end of the 1st cycle up to 95% at the end of the 4th cycle). Although oxone is a relatively inexpensive oxidant and provides high desulfurization, a 15% diesel loss after adsorption with SiO2 makes it very difficult to use in large scales, on the contrary, low desulfurization efficiencies were obtained by extraction with methanol due to low extractive performance of the extractant selected for oxidized sulfur compounds. This major difference between extraction and adsorption performance could possibly be due to SiO2 adsorbing not only oxidized sulfur compounds but also sulfur compounds [191].After biphasic UAODS reactions in the presence of HP and acetic acid, how to valorize the aqueous phase or eliminate the sulfur compounds and their oxidized counterparts in the aqueous phase is a crucial environmental issue.A 96.45% sulfur removal [192] (followed by extraction with acetonitrile at 1000 rpm mechanical stirring speed for 25 min at room temperature) was achieved from model diesel fuel containing 3976.861 mg S L–1, which is prepared by dissolving DBT in homogeneous solution (n-dodecane + n-heptane + n-hexadecane), using 10 mL HP and 10 mL acetic acid under 20 kHz frequency, 70 W power and 80% amplitude direct US at 70 °C in 30 min. The aqueous phase (total organic carbon TOC content 1200 mg L–1) containing DBT, DBTO2 and acetic acid, that is separated after the heterogeneous UAODS reaction and called diesel wastewater, was diluted individually 10- and 20-fold with distilled water and subsequently subjected to homogeneous ODS reaction at C(Fe2+) = 2 mmol L–1 and C(HP) = 20 mmol L–1 Fenton's reagent concentration (with acetic acid by adjusting pH to 3.1) under 200 W and 20 kHz direct US for 120 min. At the end of the homogeneous ODS reactions of the two aqueous phase samples diluted 10- and 20-fold with pure water, a removal of 75 and 76% for DBTO2, 28 and 66% for TOC, respectively, were obtained. HPLC analysis of the treated diesel wastewater confirmed the formation of benzoic acid followed by aliphatic carboxylic acids (e.g., oxalic acid) after 30 min as a result of oxidative degradation of small amounts of remaining DBT. It was stated that this sono-Fenton process has the potential to remove organic pollutants from diesel waste water and the treated water can be reused.In order to further remove the sulfur in the fuel (i.e., to obtain ultra-low or low-sulfur fuel), advanced oxidation processes, which are used in the removal of organic pollutants from wastewater, have also been utilized in UAODS reactions. For this purpose, FeSO4 was added to the aqueous phase containing HP-acetic acid and a 98.32% desulfurization degree [193] (followed by extraction two times at DMF/oil 1:1 vol ratio for 2 min each at room temperature) of hydrotreated Middle Eastern diesel fuel containing 568.75 ppmw total S was obtained at optimum conditions (40 °C, Fe2+/HP 0.05 mol/mol, pH = 2.10 and reaction time of 15 min) under 200 W and 28 kHz direct US. Explaining that the high desulfurization is due to the Fe2+ ion which generates more hydroxyl radicals from HP, it has been determined that the US-Fenton’s reagent system follows the second order reaction kinetics.In a similar study [184] where Fenton’s reagent as oxidizer and acetic acid were used, 97.5% sulfur removal from original diesel fuel containing 1936.48 ppmw total S (followed by extraction at DMF/oil 1:1 vol ratio under vigorous mixing at room temperature) has been achieved at optimum operating conditions (70 °C, 10 min, 8 W cm−2 ultrasonic intensity, O/S molar ratio 6: 1, FeSO4/HP mass ratio 2:10 and acetic acid/HP volume ratio 1:2) under direct US at 28 kHz frequency. It was reported that the diesel loss after oxidation-extraction is less than 8 wt% and although the density and cetane index decreased a little, the other properties of diesel fuel did not change much.By virtue of very severe process conditions (Hydrotreated diesel fuel with 421.45 ppmw total S obtained as the feeding material by hydrotreatment of diesel fuel containing 9997 ppmw total S for two-stage HDS, 7 MPa, 628 K, LHSV 1.8 h−1) necessary to reduce very high-sulfur diesel fuels by HDS to less than 10 ppmw S (9.5 ppmw S), diesel fuel containing 9997 ppmw total S was first processed by HDS in milder conditions (with 99.8% diesel fuel recovery) to obtain a fuel containing 421.45 ppmw S and then subjected to oxidation reaction (followed by extractions two times at DMF/oil 1: 1 vol ratio for 2 min each at room temperature) at 70 °C, HP/Diesel Oil 3/100 vol ratio, pH = 2.1 and Fe2+/HP 0.05 mol g−1 in the presence of Fenton’s reagent and acetic acid under 28 kHz and 200 W direct US in 15 min [178]. Along with the 92.2% diesel fuel recovery, diesel fuel containing 9 ppmw total S (97.86% sulfur removal) was obtained. Therefore, it was stated that integrating the ODS unit as a complement to the HDS unit is potentially advantageous in terms of overall process cost and efficiency.It was reported that by using individually Fenton’s Reagent and Fenton-type reagent (Cu2+-HP), which is used to enrich hydroxyl radicals, in the presence of acetic acid (pH = 1.9 ∼ 2.1), a desulfurization degree (followed by extraction twice at DMF/fuel 1:1 vol ratio at room temperature for 10 min each) of 95.2 and 89.2%, respectively, was achieved for FCC diesel fuel [185] with 1936.48 ppmw total S at 60 °C, HP/S 6:1 molar ratio and M2+ (Fe2+ or Cu2+)/HP 0.05 mol mol−1 under 28 kHz and 200 W direct US in 15 min, which is an indication that metal ions catalyze the UAODS reaction creating a synergistic effect.In a study [179] conducted to remove sulfur from a straight run diesel oil sample containing 960 ppm S (followed by extraction one time with DMF at extractant/oil volume ratio of 1:2), a desulfurization yield of 94.7% was obtained at the optimum conditions (HP/formic acid (FA) 1: 1 vol ratio, (HP + FA)/oil 1:10 vol ratio, 50 °C and 10 min) under 28 kHz–40 kHz and max 200 W direct US. It was observed that the degree of desulfurization almost does not increase due to the decomposition of HP after the optimum reaction time, the sulfur removal is slightly reduced due to side reactions after the optimum oxidant/oil volume ratio, and the desulfurization removal does not change beyond the optimum temperature. Moreover, it was stated that beyond optimum conditions, oil recovery decreases and also production costs will increase.In a similar study [180] under the same optimum conditions as the previous study [179] (except that extractant DMF/oil volume ratio is 1: 1 and extraction time is twice), the effect of HP/FA volume ratio under direct US was investigated and a sulfur removal of 92.8% has been obtained from FCC diesel oil containing 1948 ppmw total S at the end of the UAODS process. Beyond the optimum oxidant/catalyst volume ratio (1:1), it was reported that desulfurization decreases due to nonproductive decomposition of excess HP to oxygen and water as there is not enough formic acid in the medium to form high-concentration peroxyformic acid in-situ by reaction of HP with FA.The effect of extraction on desulfurization after the oxidation reaction of sulfur compounds in FCC diesel containing 1985 ppm total S with HP-FA oxidant system under indirect US was investigated [194]. Taking into account oil recovery and the consumption of extraction solvent, a desulfurization of 94.2% was achieved as a result of extraction two times at DMF/oil volume ratio of 1:1 at 30 °C for 20 min each.Recently, RSM-Box-Behnken Design has been used to find the optimum desulfurization, to examine the effect of reaction parameters and interactions between the parameters on UAODS yield and also to find which parameter or parametric relationships are more important on desulfurization such that fewer experiments are performed with this program, thus resulting in less time-consuming study.Using RSM [181], a sulfur removal of 95.46% from kerosene containing 2490 ppmw total S was achieved at the ratio of nO/nS = 15.02, nacid/nS = 107.8 and US power/fuel volume = 7.6 W mL−1 (followed by extraction with acetonitrile, extractant/kerosene volume ratio = 1, extraction stage = 1, ambient temperature, 700 rpm, 30 min.) at 20 kHz frequency and 400 W direct US at 50 °C within 10.5 min. It was observed that above the optimum nacid/nS and nO/nS ratios, the desulfurization was almost unchanged as performic acid formation and decomposition reactions occur together in an acidic medium and the equilibrium concentration of peroxyformic acid was reached due to the decomposition of HP. When the two ratios in the relation of power/volume and nO/nS to sulfur removal are above a certain value, no increase in desulfurization was observed due to dilution in the aqueous phase and the weakening of the ultrasonic wave emitted to the mixture by enlargement of the bubble cloud at the probe tip at high power. The fact that there is no significant increase in desulfurization above a certain value of the two ratios in the relation between power/volume and nacid/nS is due to the reason mentioned above. While a sulfur removal of 29.92% from kerosene is achieved by extraction alone employing acetonitrile without oxidation reaction, the desulfurization is 74.9% by oxidation and water washing without extraction process, which shows that formic acid extracts oxidized sulfur compounds sulfoxides and sulfones during the oxidation reaction.In a similar study [195] with the same oxidant system by applying RSM, a sulfur removal higher than 98% was achieved at HP/S molar ratio of 10.82, FA/S molar ratio of 379.75 and 52 °C (which are the three independent reaction parameters selected) under 70 W and 20 kHz direct US and at 15 min for model fuel containing 500 ppm total S prepared by dissolving BT in toluene. With the same values of these 3 optimum parameters found, a sulfur reduction of approximately 95.6% (followed by extraction at acetonitrile/kerosene volume ratio of 1 for 30 min at room temperature) was achieved from kerosene containing 2720 ppmw total S under 250 W direct US in 20 min. The results revealed that the decrease in desulfurization at low acid/S and high O/S values is due to the dilution of the formic acid by increased surplus HP, thus lowering peroxyformic acid concentration and also the formation of vapor-filled bubbles rather than gas-filled bubbles with increasing HP. It was found that the importance degree of the independent reaction parameters was in descending order: Acid/S molar ratio > HP/S molar ratio> (Acid/S molar ratio)2 according to the ANOVA results of the quadratic correlation equation (where the smaller than 0.05 the P value and the larger the F value, the more important the parameter).In a study [196] with the same oxidant system, using the RSM-Box-Behnken Design (BBD), where temperature and US power/gas fuel volume (W mL−1) were selected as constant parameters and O/S, Acid/O molar ratios and sonication time as process variables, 87% sulfur removal from gas oil containing 2210 ppmw total S (followed by one-time extraction at acetonitrile/gas oil volume ratio of 1: 1 under vigorous stirring for 30 min at room temperature) was achieved at O/S 46.36 molar ratio, acid/O 3.22 molar ratio in 19.81 min for 50 °C and 7.78 W mL−1 under a direct US of 20 kHz. However, in the case of 4-step extraction, 96.2% of the sulfur present in the gas oil was removed, but it was reported that the recovery of gas oil decreased to 81.25%. After the oxidation reaction under the same conditions, the extraction performances under mechanical mixing and under direct US were compared. It was observed that the desulfurization yields were approximately the same, thus showing that US does not have a positive effect on extraction. In addition to these, as a result of the preliminary cost analysis of this batch process, it was determined that a total operating cost of $ 0.43 was incurred for the treatment of 1 L gas fuel and also 31.7 and 56.3% of this total cost were liquid–liquid extraction and US Power/gas oil volume, respectively. It was stated that this calculated cost will be less in continuous-flow UAODS systems as there are stagnant zones in the mixture in batch UAODS systems, thus leading to a higher consumption of US power density per unit volume of fuel in the batch systems. According to ANOVA analysis, it was determined that the importance of variables is in the order: sonication time > acid/O molar ratio > O/S molar ratio> (acid/O × sonication time)> (sonication time)2> (acid/O molar ratio)2> (O/S molar ratio) × (acid/O molar ratio)> (O/S molar ratio)2. It was explained that sulfur removal decreased due to the scavenge of hydroxyl radicals at high acid/O molar ratio and enhancement of side reactions in case there is excess HP in the medium towards high O/S molar ratio. In high acid/O and high O/S molar ratios, it was explained that peroxyformic acid stabilizes at low pH of the aqueous phase as a result of very high concentration of formic acid after a certain value, thus resulting in a lower desulfurization by limiting the production of active oxidizing radicals, which are generated by the decomposition of performic acid.The RSM-BBD was applied to a batch reactor in a continuous study [188] in which the aqueous phase consisting of HP and FA is injected by nozzles of different diameter to just below the bottom end of the probe (which is the active site where radicals are produced). O/S molar ratio, acid/S molar ratio and sonication time were selected as independent variables at 50 °C under 20 kHz and 360 W direct US and the optimum parameters (nO/nS = 38.88, nacid/nS = 116.47 and sonication time 29.2 min.) were determined under batch conditions. According to ANOVA, it is stated that the most important terms are in the order: acid/S molar ratio> (O/S molar ratio × acid/S molar ratio) > sonication time. These optimum parameters have been applied to two continuous reactors in series (where in the first reactor, the aqueous phase was injected to the lower end of the probe) at different feed rates (thus causing different retention times) and different fuel phase/aqueous phase volume ratios (herein (Vacid/VO) = 1.117). For non-hydrogenated diesel fuel containing 1550 ppmw total S, a desulfurization of 83.39% (followed by a single extraction with acetonitrile/organic phase volume ratio of 1:1 at 1000 rpm mixing speed for 30 min at room temperature) was reached at Vf (volume of the fuel phase)/Vaq (volume of the aqueous phase) 5: 1 vol ratio, 40 mL min−1 total outlet flow rate (33.33 mL min−1 diesel fuel + 6.67 mL min−1 aqueous phase), a residence time of 3 min in the first reactor and 2.5 min in the second reactor using 1.5-mm-diameter nozzle from the point of the lowest retention time and lowest aqueous phase volume to minimize the process cost. It was explained that when the nozzle diameter decreases from 1.5 mm to 0.43 mm, the desulfurization decreased to 68.74% due to a decline in the ratio of the hydrodynamic momentum flow rate generated by the US probe to the hydrodynamic momentum flow rate of the dispersed aqueous phase (in which case, aqueous phase will stay in the active zone for much less time as the increasing flow rate by use of the smaller nozzle diameter leads to the increased momentum). In addition, it was shown that the increase of the aqueous phase flow rate from 10 to 40 mL min−1 for all the nozzle diameters leads to a decrease in desulfurization due to the reason mentioned above. Batch sonoreactor and sonoreactors in series operating at different times at a constant volume ratio of Vf/Vaq = 2.96 mL mL−1 and at different Vf/Vaq ratios at constant sonication times of 5.5 min were compared and it was reported that in all cases, the sulfur removal per power density consumed in continuous sonoreactors in series is higher than that in the batch sonoreactor.The effect of pressure on UAODS in a sonoreactor was investigated [197] and the optimum conditions (390 W US power at 20 kHz frequency, gauge pressure 0.03 barg and 22 min) were found by applying RSM-BBD in which pressure, US Power and sonication time were selected as independent variables at T = 50 °C, nO/nS = 15.02 and nacid/nS = 107.8. A sulfur removal of 96.7% (followed by one-time extraction at acetonitrile/kerosene 1:1 vol ratio under 500 rpm stirring speed for 30 min at room temperature) was obtained from kerosene with 2490 ppmw total S. Also, it was disclosed that according to computational fluid dynamics (CFD), desulfurization decreased at pressures above atmospheric pressure (1 barg and 2 barg) due to the progressively decreasing vapor volume fraction, the decreasing bubble collapse pressure, the low dispersion of the aqueous phase into the organic phase and a significant increase in the aqueous phase volume fraction. The authors suggested that the marked rise in the aqueous phase volume fraction did not result in finer emulsion droplets, thus causing the interfacial area between the aqueous and organic phase to diminish. In addition, it was stated that when the US Power increased from 100 to 400 W, the max acoustic pressure and micro-streaming speed increased according to the calorimetric analysis, thus desulfurization was improved due to the increase in mass transfer rate. It was determined that the most important terms affecting desulfurization are in the order: time > Pressure > Pressure × Power > Power according to ANOVA.In a continuous cylindrical sonoreactor with multiple probes (3 probes) and two nozzles [182], through which the aqueous phase is injected just below the first and the second probe tips from the left side of the inside of the reactor, the optimum conditions (Vacid/VO (mL mL−1) 1.12, Vaq = (Vacid + VO) 733.33 mL, Vf = 3666.67 mL, Vf/Vaq (mL mL−1) 5 and temperature 50 °C) were determined under direct ultrasound, each of which has a power of 400 W and a frequency of 20 kHz (all ultrasonic processors ON). >97% of sulfur (followed by extraction with DMF) from diesel fuel containing 1550 ppmw total S was removed using two 1.5-mm-diameter nozzles at 15 min residence time, 277.2 W electrical power, 48.90 mL min−1 total aqueous phase volumetric flow rate (flow rate of each nozzle 24.45 mL min−1) and fuel phase volumetric flow rate of 244.44 mL min−1. According to the CFD simulation results, it was explained that this high desulfurization is due to the higher hydrodynamic momentum ratio (momentum of ultrasonic jet-like streaming/momentum of the aqueous phase injected by the nozzle) as well as secondarily, further oxidation reactions of DBT derivatives with oxidizing radicals (HO2·, O· and HO·) in the active zone just below the probe tips not only when larger-diameter (1.5 mm) nozzles are used instead of 0.4- and 0.9-mm-diameter nozzles but also when each of the aqueous phase flow rates is lower (using two nozzles with an aqueous phase flow rate of 24.45 mL min−1 each instead of using a single nozzle with the aqueous phase flow rate of 48.89 mL min−1). In this case, it was suggested that the aqueous phase is dispersed more homogeneously into fuel when compared to smaller diameter nozzles at higher flow rates.The operating cost of the UAOD system was investigated [183] in a continuous flow jacketed glass reactor where the glass nozzle through which the aqueous phase (85 wt% FA + 35 wt% HP) flows is placed 3 cm below the US probe tip. Residence time (min), FA/S molar ratio and oxidant/S molar ratio were selected as independent variables at a reaction temperature of 50 °C as constant value and RSM based on BBD was applied. A sulfur removal of 86.90% (followed by one-time extraction at DMF:oil 1:1 vol ratio at room temperature and 875 rpm stirring speed for 30 min) was obtained from the partially hydrotreated diesel fuel containing 2760 ppmw total S at optimum conditions (retention time of 16 min, molar ratio of na/nS 54.47 and molar ratio of nO/nS 8.24) under 360 W and 20 kHz direct US. Under these optimum conditions, it was reported that the organic phase/aqueous phase volume ratio is 4.34 and the operating cost (chemical consumption + electricity due to ultrasound irradiation) is 7.73 cents per liter of oxidized diesel fuel. As the largest part of the operating cost was HP consumption, the organic phase/aqueous phase volume ratio was increased to 10 in order to significantly reduce the aqueous phase consumption at residence time 16 min and FA/HP volume ratio 3.16. Eventually, a sulfur removal of 84.38% was achieved with an operating cost of 4.66 cents per liter of oxidized diesel fuel at na/nS 23.64 molar ratio, nO/nS 3.58 molar ratio, 7.07 mL min−1 diesel flow rate and 0.71 mL min−1 aqueous phase flow rate (0.54 mL min−1 85 wt% FA + 0.17 mL min−1 35 wt% HP). According to ANOVA results, it was determined that the most important terms affecting desulfurization in this process are in the order: residence time ≈ na/nS > (residence time)2 > (na/nS)2 > (na/nS × nO/nS) > (nO/nS)2.Sono-desulfurization of gasoline and crude oil was performed at optimum conditions found by applying RSM-BBD in which ultrasonic power, irradiation time and oxidant amount are selected as independent variables [198]. A desulfurization of 80.87% (followed by extraction three times at DMSO/gasoline 1:1 vol ratio and water washing four times) was obtained for gasoline containing 1207 ppmw S at optimum conditions (464.7 W direct ultrasonic power (pulsed ultrasound 2 s on, 2 s off), 5.5 min irradiation time and 8.1 mL HP (HP: FA volume ratio 1:1)), whereas a sulfur removal of 73.37% (followed by first magnetic stirring of oil sample for one h and then extraction with 60 mL of a mixture at acetonitrile:methanol:water 1:1:1 vol ratio) was achieved from the crude oil containing 28,620 ppmw S at optimum conditions (785.1 W direct ultrasonic power, 6.2 min irradiation time, 11.4 mL HP (HP: FA, the same volume ratio) with the same pulsed ultrasound. It was stated that after the oxidation of the gasoline sample, adding distilled water up to 1% of the DMSO volume to DMSO for the extraction of oxidized sulfur compounds decreases desulfurization by 20% compared to extraction alone with DMSO. It was explained that this low desulfurization is due to the fact that water reduces the extraction ability of DMSO as the DMSO and water dipole moments [199] are 3.96 and 1.85 D, respectively, (hence DMSO has greater polarity). The differences between mechanical stirring-heating and desulfurization under US were compared and these differences were reported to be approximately 10 and 30% for gasoline and heavy crude oil, respectively, which demonstrates that UAODS is more effective for high-sulfur fuels. This threefold higher difference can be attributed to the emergence of the higher cavitation intensity [200] as heavy crude oil has higher density, higher viscosity and higher surface tension than gasoline. In addition, the high vapor pressure of extremely volatile gasoline compared to heavy crude oil can limit violent implosion of cavitation bubbles in the liquid mixture [125].RSM-Box-Behnken Design (BBD) was used to evaluate the effects of nformic acid/nS, nO/nS, ultrasound power (UP)/simulated oil volume and temperature on UAODS and to optimize these reaction parameters on the purpose of max attainable desulfurization efficiency [187]. A sulfur removal of approximately 97% from DBT containing 500 ppmw S in toluene is reported at nO/nS = 26.7, nformic acid/nS = 74.6, UP/model oil volume = 7 W cm−3 and at 50 °C under 20 kHz and 400 W direct US in 630 s. Besides, it was stated that the FA (formic acid)/HP molar ratio should be at a certain value (1.4–2.8) in order to maximize the concentration of peroxyformic acid (HCOOOH), which is formed in the equilibrium reaction between HP and HCOOH in the aqueous phase in desulfurization reactions and oxidizes the sulfur compounds.In a study [161] where a computational fluid dynamic (CFD) model was used to examine the hydrodynamic and mass transfer characteristics of model fuel in the ultrasonic horn reactor, it was explained that high desulfurization is caused by physical effects such as jet stream, high turbulence intensity rather than the chemical effect of ultrasound, and the reaction is controlled by chemical kinetic due to the very high mass transfer rate. In the mentioned study, a sulfur removal of 96.35% from the model fuel containing 500 ppmw DBT in toluene was achieved at nO/nS = 26.7, nformic acid/nS = 74.6, UP/Model Oil Volume = 26.7 W mL−1 under 20 kHz direct US at 50 °C in 210 s.It was observed in the studies [178,181–185] that the properties of diesel fuels (density at 15 °C, kinematic viscosity at 40 °C, flash point, water content, cetane index) almost did not change after UAODS process followed by extraction.Three organic acid catalysts (FA, acetic acid and trifluoroacetic acid) were compared and a 76.5% sulfur reduction [201] (followed by extraction at a DMF/oil volume ratio of 1:1) was achieved for the catalytic cracking diesel containing 1452 ppmw total S by using trifluoroacetic acid at oxidant/oil 1:10 vol ratio, 70 °C and 60 min as the optimum operating conditions under indirect 20 kHz US, which is higher than the sulfur removals obtained in the case of using acetic acid and FA catalysts as the acidity [202] of trifluoroacetic acid (pKa = 0.18) is higher than that of formic acid (pKa = 3.75) and acetic acid (pKa = 4.75), thus causing the oxidizing power of the peroxycarboxylic acid formed to increase further.In a study [189] where 1-butyl-3-methyl imidazolium hydrogen sulfate [Bmim][HSO4] and 1-octyl-3-methyl imidazolium hydrogen sulphate [Omim][HSO4] with two different alkyl lengths were synthesized and used instead of aqueous phase, approximately 100% desulfurization yield of the model fuel containing 500 ppmw DBT in n-decane was obtained using [Omim][HSO4] at O/S = 5 molar ratio and mass ratio IL/model fuel = 2 under 30 W power and 25 kHz direct US at 25 °C in 3 min. In the experiments in the absence of ultrasound, it was explained that the desulfurization with [Omim][HSO4] is higher than the desulfurization with [Bmim][HSO4] by applying the same optimum operating conditions as the case of using ultrasound under stirring at 900 rpm. It was noted this high desulfurization is due to the longer alkyl chain of the cation of [Omim][HSO4]. In addition, the reactivity of different sulfur compounds under the same operating conditions was compared and it was reported that the UAODS was in descending order DBT > BT > T > 4,6-DMDBT. It was stated that the lowest desulfurization for 4,6-DMDBT is due to the steric hindrance of two alkyl groups adjacent to the sulfur atom, hence weakening the π-π interaction between the aromatic sulfur compound and the ionic liquid. Under the same optimum conditions, a UAODS efficiency of 76.3% was obtained for the real diesel fuel containing 746 ppmw total S. Moreover, it was reported that [Omim][HSO4] can be used six times without losing its activity in UAODS reactions of the model fuel by regenerating it after each reaction and the solubility of the model fuel in this ionic liquid is very low (1.45 wt%), thus suggesting that the synthesized ionic liquid has the potential to be used both as an extractant and as a catalyst.However, the high viscosity of ionic liquids, their costly synthesis [203], and the change in the solubility [204] of the fuel in the ionic liquid according to the anions and cations formed depending on the starting raw materials, and more importantly, the presence of aromatic groups [177,189] such as imidazolium in IL significantly that reduces the desulfurization reactivity of thiophenes, especially abundant in petroleum products, due to steric hindrance make the UAODS process very difficult to be feasible using ionic liquid.One of the two identical hydrotreated diesel feeds containing 231 ppmw S and 115.5 ppmw N to use expensive oxidants in lower quantities was subjected to pre-extractive desulfurization and the other to pre-adsorptive desulfurization (diesel/methanol volume ratio 1:1 for EDS/N and diesel/fuller's earth (V/W) = 1:0.2 for ADS/N) and then, the UAODS/UAODN reaction (followed by EDS/N and ADS/N individually at the same ratios as those in the pre-treatments) of the two partially desulfurized and denitrogenized fuel samples (S = 196 ppmw and N = 85 ppmw after pre-EDS/N and S = 184 ppmw and N = 52 ppmw after pre-ADS/N) was performed using oxone or HP in US Cup Horn at 80 °C under 80 W and 19.9 kHz direct US for 90 min [205]. As a result of all these processes, diesel fuel with 11 ppmw S and 6 ppmw N is obtained by the pre- and post-ADS/N process, while diesel fuel with 78 ppmw S and 25 ppmw N is obtained by the pre- and post-EDS/N process, thus suggesting that it would be economically feasible to use cheap and efficient adsorbent fuller’s earth instead of expensive extractant methanol. It was stated that this process can be proposed to be complementary to HDS.According to the ODS mechanism [28,29,206] (Scheme 3 ) using phosphotungstic acid in the presence of PTC, the phosphotungstate anion in aqueous phase is oxidized to the peroxophosphotungstate anion (1) by HP, then this active oxidizing complex anion is transferred (3) to organic phase by forming an ion pair (2) with the lipophilic cation of PTC. This complex anion is reduced to phosphotungstate anion by oxidizing the sulfur compounds in organic phase (4). The phosphotungstate anion is transferred to the aqueous phase by the lipophilic cation (5) and the cycle is completed.A DBT removal of 100% from model fuel [206] containing 4000 ppmw S DBT in toluene was performed using HP 30 vol% (phosphotungstic acid concentration of 0.6 mM in aqueous phase and tetraoctylammonium bromide (TOAB) concentration of 7.32 mM in organic phase) under 600 W and 20 kHz direct US at 75 °C in 7 min. The same conditions were applied to diesel fuels with different sulfur content at certain times (18 min for diesel A with 7744 ppmw S, 10 min for diesel B with 3011 ppmw S and 10 min for diesel C with 1867 ppmw S) at 75 °C and a desulfurization yield (followed by extraction with acetonitrile three times at solvent/oil mass ratio of 1:2 at room temperature for 2 min each) of 98.2, 98.7 and 99.4%, respectively, was achieved along with a fuel recovery of 82.8, 87.2 and 85.5 wt%. It was reported n-paraffins, n-alkyl cyclohexanes, n-alkyl benzenes and alkyl naphthalenes as component classes in the diesel C sample selected as representative were not adversely affected during oxidation, but alkyl naphthalenes among the four main components have relatively high polarity and thus they were extracted by acetonitrile.In a similar study [207] (where the temperature, tetraoctyl ammonium fluoride (TOAF) concentration, sonication time, phosphotungstic acid concentration and HP purity were 70 °C, 7.5 mM, 10 min, 0.7 mM and 30 vol%, respectively) with the research [206], under the same direct US power and frequency in a continuous flow sonoreactor, marine fuel with less than 23 ppmw S and jet fuel with 1 ppmw S (each followed by adsorption with activated, acidic Al2O3), respectively, were obtained from marine gas oil containing 1710 ppmw S and Jet Fuel (JP-8) containing 863 ppmw S. 33-fold lower consumption of Al2O3 compared to acetonitrile, loss of alkyl naphthalene less than 1 wt%, regeneration with 94% alumina recovery by washing with DMF solvent and maintaining 99% of its adsorption capacity by calcination at 550 °C have revealed that alumina has the potential of being used in large-scale continuous systems.In another study [208] where 30 wt% HP and phosphotungstic acid were used, the UAODS performances of DBT in the presence of different phase transfer catalyst types at 70 °C under 20 kHz and 600 W direct US were evaluated. It was stated that desulfurization reactions of DBT took place in the presence of TOAB (49.57% conversion), tetrabutylammonium bromide (TBAB) (38.34% conversion), methyltributylammonium chloride (MBAC) (11.4%), methyltributylammonium hydroxide (MBAH) (11.10%) and tetramethylammonium fluoride (8.20%) as cationic-type PTCs, whereas desulfurization reactions did not occur in the presence of 1-octanesulfonic acid as anionic-type PTC, Tween 80 as non-ionic PTC and in the absence of PTC. In addition, in the presence of TOAF and tetraoctadecylammonium bromide (TODAB), 90.30% (97.53% in 20 min for TOAF) and 56.89% conversions were performed in 10 min, respectively. From these results, it was emphasized that the biggest positive effect on UAODS is the long alkyl chain (hence more lipophilic cation) bound to the quaternary cation, and the less positive effect is the hydrophilic anion of quaternary salt. It was stated that the smaller (i.e., the more hydrophilic) the size of the monoatomic anion of quaternary salt for the same alkyl chain length, the more effective the PTC. It was determined by GC-PFPD analysis that 3-bromobenzothiophene and 2-bromobenzothiophene sulfone were formed as intermediates when TOAB was used in UAODS reactions of BT, while in the case of TOAF, intermediate products were not formed. The formation of the byproducts can be shown representatively in Scheme 4 : either by the radical mechanism [209,210] where aromatic sulfur compounds react with bromine radical which is formed by homolytic cleavage [211] of molecular bromine on sonication or by direct reaction [212] with Br2 formed. Bromine radical can also be formed by the reaction of hydrogen peroxide with bromide anion [213].The reason for the absence of intermediates can be explained as follows: the standard reduction potential [214] of fluorine and HP is E° (V) = +2.87 and E °(V) = +1.77, respectively. In case of quaternary ammonium salt containing fluoride anion, H2O2 cannot oxidize the fluoride anion to fluorine as the standard reduction potential of F2/F− is +2.87 V. Therefore, fluoride-containing organosulfur compounds are not found in the reaction products. But as a result of the dissociation of the quaternary ammonium salts containing the other halide anions except fluoride in aqueous acidic media, the halide ions reduce hydrogen peroxide to water, causing the decomposition of hydrogen peroxide [215]. For example, when TOAB is used, HP in the aqueous acidic phase is reduced by oxidizing the bromide anion released by dissociation [216] of the quaternary ammonium salt in water according to the following reaction as E° (V) of Br2 is + 1.07 [214]. Catalytic decomposition [217,218] of HP in acidic medium in the presence of bromide ion is as follows: (1) H 2 O 2 ( a q ) + 2 B r aq - + 2 H aq + → B r 2 + 2 H 2 O (2) H 2 O 2 ( a q ) + B r 2 ( a q ) → O 2 + 2 B r - + 2 H + Br2 formed in reaction 1 reacts with H2O2 in reaction 2 forming bromide ion again. The sum of reaction 1 and 2 is written as 2 H 2 O 2 ( a q ) → 2 H 2 O ( l ) + O 2 ( g ) Br2, which is formed according to reaction (1), participates in bromination reaction with sulfur compounds in organic phase and forms bromo intermediates. Besides, as mentioned before, bromination reaction can be carried out by bromine radical Br· formed by homolytic decomposition of Br2 molecule under US. The reason of the decreased desulfurization in this case can be explained as follows: as HP is decomposed in an acidic environment, the amount of peroxo-phosphotungstate formed in situ may decrease significantly. Additionally, a small amount of peroxo-phosphotungstate, which has a higher ability to oxidize organic compounds than hydrogen peroxide [219], reacts very quickly with Br2 in the medium, causing the amount of peroxo-phosphotungstate to decrease much more. Therefore, the desulfurization under US can be significantly lower. In the case of the quaternary ammonium salt containing fluoride for the same alkyl chain length, HP is not reduced by fluoride, thus high desulfurization efficiencies can be achieved by the high amount of peroxotungstate formed and no intermediates are formed. A similar phenomenon can occur when carboxylic acids such as formic acid are used instead of phosphotungstic acid. The reaction mechanism under ultrasound irradiation in the presence of a quaternary ammonium salt with bromide anion can be explained as follows: (3) Br - B r → B r · + B r · (4) H 2 O 2 → 2 O H · (5) HCOOOH ⇌ H C O O · + H O · (6) H 2 O 2 + B r · → H O 2 · + H B r (7) H O 2 · + H 2 O 2 → H O · + O 2 + H 2 O (8) HO · + H 2 O 2 → H O 2 · + H 2 O (9) Br · + B r · → B r 2 (10) HO · + H O · → H 2 O 2 (11) HO · + B r · → H O B r (12) H O 2 · + B r · → H O O B r Accordingly, Br2 formed through the reaction 1, the PFA formed in-situ and HP in the reaction solution bring about a series of reaction 3–12 generating hydroxyl [158,161,220] and bromine radicals [211] by the decomposition of PFA and HP and the homolytic bond cleavage of Br2. Consequently, the hydroxyl and hydroperoxyl radicals play a dominant role in oxidation of the organosulfur compounds. In a study [221] in which HP reacts with FA at 30 °C in the presence of TBAB, it was confirmed by titrimetric analysis that the HP concentration decreased significantly by the decomposition of HP and the peroxyformic acid concentration was too low. The change of the transparent color of the aqueous solution containing HP, FA and PFA in the absence of TBAB to the yellow color of the bromine water formed by the dissolution of Br2 in water in the presence of TBAB is an additional indicative of the decomposition. The resulting performic acid (or peracetic acid formed in the case of using acetic acid) can also react as follows: (13) RCOOOH + 2 B r - + 2 H + → R C O O H + H 2 O + B r 2 where R is H or CH3 and its concentration may decrease depending on the concentration of Br ion in the medium.Moreover, formic acid can react with the resulting Br2 according to the following reaction [222,223] (14), thus causing formic acid concentration to decrease. (14) HCOOH + B r 2 → C O 2 + 2 H + + 2 B r - In the reaction mechanism in the case of using HP, FA and TOAF, peroxyformic acid generates formyloxyl radical and hydroxyl radical by homolytic cleavage under US [220]. As a result of the reaction of peroxyformic acid with hydroxyl radicals, formyl radical and peroxyformyl radical are formed, which is similar to the reactions [224] of peracetic acid with the hydroxyl radicals. Therefore, in addition to the hydroxyl radicals formed and the high concentration of performic acid, highly reactive formyloxyl and peroxyformyl radicals may also be responsible for the high desulfurization.Diesel fuel [208] containing 0.1 g TOAF was undergone ODS reaction (followed by extraction four times at acetonitrile/oil 1:1 mass ratio at room temperature for 1.5 min each) with an equal volume of 30 vol% HP solution containing 0.2 g of phosphotungstic acid under the same US frequency and power at the same temperature as the previous study [207]. After UAODS reactions of 10 min followed by extraction four times, a sulfur removal of 95, 98.8, 87.5, 99.9 and 96.1% was achieved from F-76 containing 4222 ppmw S, MGO containing 1710 ppmw S, JP-5 containing 113.7 ppmw S, JP-8 containing 863 ppmw S and transportation fuel containing 259 ppmw S, respectively. In addition, after a 98.8% UAODS yield from MGO containing 1710 ppm S in the presence of TOAF, the aqueous phase was reused for two fresh MGO samples with 1710 ppmw S each in the presence and absence of TOAF and a UAODS of 98.15 and 96.01%, respectively, was obtained. Again, under the same conditions, this time using dilute HP (3 vol%), a UAODS of 97.90 and 94.8% was obtained for MGO and F-76, respectively. It was stated that after the UAODS reaction of organic sulfur compounds, 99.49% of the tungsten remained in the aqueous phase according to ICP analysis, hence it could be completely recovered.In a study [225] investigating the effect of quaternary ammonium salts with four different alkyl lengths as PTC on UAODS, using PTC (optimum concentration 0.0116 mol L–1) in the range of 0.03–0.25 g, 12 mL 30% HP and 12 mL formic acid, 28.37, 42.37, 70.02, 86.57 and 94.67% sulfur removal, respectively, were obtained without PTC and in the presence of TMAB, TEAB, TPAB and TBAB at 50 °C in 1.5 h under direct US for 0.028 mL of thiophene dissolved in 24 mL of n-heptane. The highest desulfurization with TBAB was attributed to the bigger radius (thus more stable complex formation [HCOOO–--Q---Br] by higher electron delocalization) of the phase transfer cation TBA+ compared to the radii of the other phase transfer cations for the transfer of [HCOOO–] to the organic phase in the presence of the same anion (Br–) and the higher extraction constant of TBAB. It was revealed that the reaction follows pseudo first order kinetics.In a study [226] in which the effect of two different types of continuous flow reactors on UAODS (followed by extraction at acetonitrile/oil 1: 1 mass ratio at room temperature for 1.5 min with vigorous shaking) of MGO was investigated, a 92.74% sulfur removal was performed using 25 g 30 vol% HP, 0.1 g TOAF, 0.2 g phosphotungstic acid under 600 W US in power at 70 °C in 20 min for treating 20 g MGO containing 1710 ppmw total S in a probe-type reactor operating at 20 kHz, while using 625 g 30 vol% HP, 2.5 g TOAF, 5 g of phosphotungstic acid to treat 500 g of MGO per h in a portable tubular sonoreactor operating at 40 kHz, a desulfurization degree of 92.36 and 89.78% was achieved at 25 °C for 100 W US power-60 min and 200 W US power-30 min, respectively. Then, this tubular sonoreactor was scaled up to a treatment rate of 12.5 lb MGO h−1 and a 92.42% desulfurization performance was accomplished using 7.09 kg 30 vol% HP, 56.75 g TOAF and 28.13 g phosphotungstic acid under 100 W US power at 25 °C in 60 min. It has also been stated that sonoreactors can be connected in parallel to treat more fuel (25 lb h−1) with the same removal percentage. In addition, by using diluted HP (3 vol%), a sulfur removal of 91% was reached in this sonoreactor in 120 min. Moreover, it is predicted that chemical costs can be reduced by recycling the processed phosphotungstic acid, TOAF and HP by connecting sonoreactors in parallel to treat larger quantities of fuel (four times the recycle rate) and electricity consumption can be reduced by using low power US. Thus, in terms of total cost, it was reported that this parallel sonoreactor type has the potential to be applied in large-scale processes and has a greater advantage over batch-operated probe-type reactors for industrial and commercial applications.In a study [227] where ionic liquid was used instead of the aqueous phase, 97.6, 99.4 and 98.9% sulfur removal (followed by stirring for 170 min), respectively, was obtained from 511 ppmw thiophene, 524 ppmw benzothiophene and 530 ppmw dibenzothiophene using 5 g 30 vol% HP, 1.5 g 20% trifluoroacetic acid and 0.3 g TOAF at 50 °C in 10 min in the presence of 5 g of 1-n-butyl-3-methyl imidazolium methylsulfate ionic liquid under 600 W and 20 kHz direct US. A 100% desulfurization was achieved by applying the same conditions for Navy diesel (F-76) containing 4220 ppmw total S instead of model compounds. It is reported that the limitation of this method is that the ionic liquid used can extract sulfur-free aromatic compounds present in the fuel.Nowadays, due to the increase in oil consumption, urban and industrial wastes have been used as an energy source. In the presence of 0.1 g TOAB, 30 vol% HP and 0.2 g phosphotungstic acid, a sulfur removal [228] of 27.5 and 61.8% (followed by extraction three times at acetonitrile/oil 1: 1 mass ratio with vigorous agitation at room temperature for two min each) under 20 kHz direct US at 88 °C in 20 min, respectively, was achieved from pyrolysis oil containing 8800 ppmw total S obtained by pyrolysis of the waste tire at 650 °C for use as clean fuel and also diesel fuel containing 960 ppmw total S. As high carbon black and different hydrocarbon compounds in pyrolysis oil led to low desulfurization efficiency, after UAODS reaction, oxidized compounds were adsorbed in a 6-cm-length column filled with 30 g Al2O3 and a sulfur removal of 68.2 and 99.7% was performed for pyrolysis oil and diesel, respectively. Nevertheless, this sulfur removal value was not considered sufficient as pyrolysis oil contains more benzothiophene and thiophene groups with the lowest ODS reactivity compared to diesel fuel according to GC-SCD analysis. Therefore, two continuous UAODS reactors were connected in series and a desulfurization efficiency of 89% was obtained for the pyrolysis oil containing 0.88 wt% total S under the same UAODS reaction conditions, followed by adsorption in a 6-cm-length column filled with 30 g Al2O3.A series of UAODS experiments [229] were conducted by selecting sonication time, thiophene solution/phosphotungstic acid mass ratio, thiophene solution/HP mass ratio and thiophene solution/TOAB mass ratio as independent variables. At the optimum conditions found by principal component analysis (T:HP:TOAB:Phosphotungstic acid = 1:1.5:0.005:0.01 mass ratio), an approximately 73.5% conversion of thiophene at 500 ppmw concentration to its sulfones has been carried out in the range from 75 to 85 °C in 20 min under 20 kHz direct US. The same conditions were applied to solutions of other model sulfur compounds and the ODS reactivity following the pseudo first-order reaction kinetics was in the order: 4,6-DMDBT > 4-MDBT > DBT > 2-MBT > BT > T. It was explained that the low reactivity of thiophene is due to the low electron density on S atom and the relatively high reaction temperature near the boiling point (84 °C) of thiophene.The conversion of 99% (55.5% at 0.02 M HP) and 99.9% (99.1% at 0.02 M HP), respectively, was achieved from model fuel 1 (500 µg BT mL−1) and model fuel 2 (500 µg DBT mL−1) using 0.2 g phosphotungstic acid, 0.1 g TOAB, 0.65 M HP at 80 °C in 15 min under 20 kHz direct US [230]. The activation energies for the oxidation reactions of DBT and BT following the pseudo first-order reaction kinetics were found to be 45.01 and 60.52 kJ mol−1, respectively.An economic analysis of the study [228] was also evaluated. A sulfur removal [231] (each followed by adsorption in a 6-cm-length column filled with Al2O3) of 68 and 90.91%, respectively, was obtained from pyrolysis oil with high-sulfur content (8800 ppmw total S) obtained by pyrolysis of waste tires in one continuous sonoreactor and two continuous sonoreactors connected in series at pyrolysis fuel/phosphotungstic acid 100:1 mass ratio, 30 vol% HP sol./TOAB 250:1 mass ratio and the convenient feed rates of aqueous and organic phase in such a way that fuel/water volume ratio is 1:1 in the reactors at room temperature and atmospheric pressure in 20 min under 20 kHz direct US. As a result of the benefit-cost analysis, it is explained that a single UAODS unit can be feasible at industrial scales as the benefit/cost ratio is 1.16 and 0.86 for a single reactor and reactors in series, respectively. A recycle rate of 95, 92, 99 (which is obtained by regeneration at 500–600 °C) and 95% was reached for phosphotungstic acid, HP, Al2O3 and PTC, respectively, in a single sonoreactor.A 47% yield (which is higher than the desulfurization efficiency at atmospheric pressure under the same conditions) of UAODS was obtained for the model fuel [232] with 100 ppmw DBT concentration prepared by dissolving DBT in toluene using 0.05 g of TOAB, 2 mL of 30 vol% HP and 4 mL of formic acid at 25 °C in 90 min under high pressure of 1.8 bar and 35 kHz and 70 W indirect US. It was explained that this relatively high desulfurization is caused by the stable complex formation of TOAB with HP and the elimination of transient cavitation by high pressure, thus preventing the production of reducing species such as H2 and CO, which consume oxidizing species formed by the collapse of transient cavitation bubbles in the organic phase. It was stated that as US emulsifies the aqueous and organic phase highly (hence creating a higher interface area) and the mass transfer resistance is relatively large in the absence of PTC under mechanical mixing, the effect of PTC under US on sulfur removal is lower than that under stirring.In a similar study [233] where the effects of PTC on UAODS were elucidated by cavitation bubble dynamics and thermodynamic analysis, at HP/HCOOH 0.6 molar ratio, HP/TBAB 16.11 molar ratio (0.5 g of TBAB) and solvent/oxidant 3.33 vol ratio, a sulfur reduction of approximately 96.65 and 77.63% was achieved from 20 mL of model fuel containing 100 ppmw DBT in toluene with 35 kHz and 70 W indirect US at 40 °C in 90 min under atmospheric pressure and nitrogen atmosphere of 1.8 bar, respectively. On the contrary to the study [232], it was reported that this low desulfurization at high pressure occurs due to lower emulsification and lower interfacial area compared to the situation at atmospheric pressure although transient cavitation is eliminated. It was declared that DBT undergoes almost complete oxidation due to the intensive microconvection with the help of US and the enhanced UAODS by transferring fast the oxidant anion of PTC to the organic phase by a large amount of PTC and oxidant in the medium compared to DBT although UAODS in the presence of PTC is based on an ionic mechanism (with higher activation energy than the activation energy of the UAODS reaction in the absence of PTC) rather than radical mechanism. In addition, it was reported that the effect of PTC under mechanical mixing is less pronounced than the effect under US due to the higher activation energy, the higher ΔG and the lower - ΔS value of the stirring system compared to the ultrasonic system.UAODS reactions [234] of two model fuels containing 500 ppmw model sulfur compound each prepared by dissolving BT and DBT in toluene were carried out using 50 wt% HP and TOAB with different polyoxomethalate catalysts at 30, 50 and 70 °C in the range of 2 to 30 min and it was found that the highest reactivity was obtained with a DBT conversion of 94.8% after 30 min of reaction by using NaPW under 500 W power 20 kHz and 40% amplitude (200 W power output) direct US at 70 °C. According to the BT and DBT conversion results, it was found that the UAODS catalytic activity was in the order Na3PW12O40 > H3PW12O40 > H3PMo12O40 > H4SiW12O40 as well as an increase in sulfur removal with increasing temperature for each catalyst. It was stated that the reason for the activity order H3PW12O40 > H3PMo12O40 is that the peroxotungsten complex formed is more catalytically active than the peroxomolybdenum complex even though the standard reduction potential of Mo(VI) is higher than W(VI). However, it was noted that the acidity of the aqueous phase in the case of phosphotungstic acid does not affect UAODS yield much when compared to the desulfurization results obtained in the case of the most active catalyst, sodium phosphotungstate.At optimum conditions (21.96 mL oxidant volume, 1 g catalyst, 0.1 g PTC and 100% amplitude) found by RSM, in which volume of oxidant (40 vol% HP), catalyst (phosphotungstic acid) mass, TOAB mass and ultrasonic wave amplitude are selected as independent variables, using Minitab 15 software, a desulfurization (followed by extraction at acetonitrile/oil 1:1 mass ratio) of 94.5% for gas oil [235] containing 250 ppmw total S was performed at 65 °C in 20 min under 20 kHz and 750 W direct US. In this study, it was reported that the importance of process independent variables and their interactions according to UAODS results was in the order oxidant volume > ultrasonic wave amplitude > oxidant volume × ultrasonic wave amplitude > catalyst mass > PTC mass > oxidant volume × PTC mass > catalyst mass × PTC mass > PTC mass × ultrasonic wave amplitude and after a certain HP volume, excess HP causes a reduction in sulfur removal by creating a radical scavenging effect.In a study [236] aimed at reducing the kinematic viscosity and sulfur of diesel oil, using the Box-Behnken design as RSM by Design Expert v.7.0.0 software, HP volume (X1), acetic acid volume (X2), PTC (TOAB) mass (X3), the amount of transition metal catalyst (phosphotungstic acid) (X4) and time (X5) were chosen as independent variables. As a result of the screening of the variables, time was found to be insignificant with respect to the desulfurization performances. After applying RSM by screening out the time variable, the importance of the relevant four variables and their interactions with each other for UAODS according to the results of ANOVA was in the order X1 2 > X4 2 > X3 2 > X2 2 > X1X2 > X2 > X1. Under the optimum conditions found (13.17 mL HP, 17.26 mL acetic acid, 0.15 g TOAB and 1.5 g phosphotungstic acid), an S removal (followed by extraction one time at 166.7 g L–1 NaOH (caustic soda solution)/oil 1:1 vol ratio for 2 min) of 68.85% was achieved from diesel oil containing 5044 ppmw total S at 50 °C in 5 min under 20 kHz frequency, 700 W power and 40% amplitude direct US. After a certain amount of PTC, the mass transfer was slowed down due to the formation of a thick turbid layer in the mixture, thus leading to a reduction in UAODS. A similar trend of sulfur removal to the trend with PTC has been also observed for the transition metal catalyst, but due to the large volume of phosphotungstic acid and the small surface area of the particles. As a result of the screening analysis, it was stated that as the viscosity of diesel fuel, which has a kinematic viscosity of 3.96 cSt at 40 °C, decreases by max 20% after UAODS process, the relevant independent variables have no effect on the viscosity, and therefore the kinematic viscosity as a dependent variable was not taken into account.In a study [237] investigating the mechanism of the UAODS system in the presence of different catalysts (phosphotungstic acid, acetic acid and formic acid), it has been underlined that the desulfurization reaction is based on the ionic mechanism (caused by the transport of the peroxo-metallate anion and the anion of peracids from the aqueous phase into the interface by the lipophilic cation of PTC) in the presence of phase transfer catalyst, whereas in the absence of PTC, the desulfurization reaction is based on the radical-based mechanism (caused by the formation of active oxygen radicals such as acetyl radical CH3CO· and hydroperoxy radical HO2· by resulting in decomposition of peracids and HP by the collapse of cavitation bubbles formed). It was found that the sulfur removal efficiencies achieved at 1.8 bar for all three catalysts were lower than the desulfurization performances at atmospheric pressure, mainly due to the reduction in microconvection intensity within the mixture under high pressure, resulting in lower mass transport. In this study, in contrast to the other two studies [164,233] in which n-hexane and toluene were used as solvents, it was reported that as n-decane has a high boiling point and therefore has a very low vapor pressure, no reducing species such as H2 and CO, which reduces oxidizing species, were formed as a result of ultrasonic cavitation at atmospheric pressure. At n-decane (organic phase)/HP (aqueous phase) volume ratio of 10, a maximum desulfurization of about 74% with a rate constant of 0.0155 min−1 was performed using 60 mg L–1 TBAB, 4 mL FA and 2 mL HP at 50 °C in 90 min under 35 kHz and 70 W indirect US for the model fuel containing 100 ppmw DBT in n-decane. Excessive use of PTC prevented mass transfer, decreasing UAODS relatively. The excess of the transition metal catalyst acts as an emulsion in the mixture by covering the emulsion droplets with a thin film and creating a barrier in the mass transport of the oxidant into the interface, thus causing the UAODS yield to be levelled off.The optimum conditions, which led to a sulfur removal of 60.75% without extraction, found for the batch reactor in the study [236], were applied to the continuous tube-type flow-through sonoreactor [238] by scaling up 2.5 times and under direct US with two transducers operating at a frequency of 20 kHz and a sonication power of 48 W each, a sulfur removal efficiency of 80.79% was achieved from final gas oil containing 5044 ppmw total S using 30 mL HP, 45 mL acetic acid, 0.375 g TOAB and 3.75 g phosphotungstic acid at equal feed and outlet flow rates in 5 min. It was explained that this higher conversion compared to that in the batch reactor is due to the lack of temperature control (hence leading to an increase in the temperature of the mixture as a result of cavitation under US) in the continuously operating sonoreactor and the fact that every fluid element does not reside for exactly 5 min as in the batch reactor (i.e., resided for 5 min on average). The kinematic viscosity of the relevant gas oil decreased by 9.40% within 5 min under the UAODS conditions, while a 13.5% reduction in kinematic viscosity was achieved by using US alone in the same minute. It has been noted that US gives off some of its energy to split HP and peracetic acid into their radicals under oxidation conditions, while under US alone, it converts the gas oil into lighter fractions by giving off its energy to cleave the C - C and C - S bonds. However, for the cases of US alone and UAODS, no significant change was observed in kinematic viscosity at treatment times of 15 min, compared to the kinematic viscosity before the treating of gas oil. In the absence of acetic acid, besides final gas oil containing 5044 ppmw total S, other feedstocks (atmospheric gas oil with 10,700 ppmw total S, atmospheric kerosene with 4980 ppmw S, Isomax gas oil with 181 ppmw total S) were subjected to oxidation reaction under direct US with 48 W max power and it was stated that the UAODS efficiency is in the order atmospheric kerosene > atmospheric gas oil > final gas oil > Isomax gas oil and the sulfur removal from high-sulfur gas oils is higher. As for kerosene, since lighter fractions as well as the small number of condensed aromatic sulfur compounds (thus lower specific gravity, lower kinematic viscosity, and lower boiling range of kerosene, compared to gas oils) were present, the best desulfurization improvement has been achieved.In a study [239] where crude oil containing 2133 ppmw total S was desulfurized and upgraded (simultaneous extraction and oxidation process) under 40 kHz indirect US, 65.28% S removal was achieved with 200 ppm oxidant, 60 ppm demulsifier dosage and distilled water at 65 °C in 10 min and it was determined that the physical properties of the treated crude oil have improved (ie, decrease in density, decrease in kinematic viscosity at 20 °C, increase in cetane number, decrease in 10% carbon residue on residuum/%).At optimum conditions (17 min, 180.3 mmol HP and 25 ppm FeSO4) found by applying RSM based on central composite design (CCD) in which HP amount, catalyst (FeSO4) amount and time are selected as independent variables, a 90% desulfurization of gas oil [240] containing 9500 ppmw total S was performed by three-stage UAODS process (followed by extraction three times at a volume ratio of methanol/oil 4:5 at room temperature for 2 min each after every UAODS reaction) using isobutanol as PTC in the presence of acetic acid (ie, in acidic medium where the catalyst is active at pH less than 3) at 62 °C under 24 kHz and 400 W direct US. In the presence of TOAB as PTC instead of isobutanol, 21.99% sulfur removal from gas oil was performed by a one-step UAODS process under the same conditions, while in the presence of isobutanol, a 67.70% reduction in total sulfur was achieved by one-step UAODS (followed by extraction). Moreover, it was stated that isobutanol is very cheap, can be mixed into the fuel and burned, and it has economic viability as it does not require separation after UAODS reactions. After the oxidation reactions, the extractions with methanol were carried out under US and the sulfur removal was the same as that obtained by the extraction under stirring, thus demonstrating that ultrasound has no effect on extraction in this study. According to the F-test of the regression model, it was revealed that the effect of time variable and time × HP interaction on UAODS is not of importance.At the optimum conditions (16.4 min sonication time, 122.1 mg TOAB, organic phase/aqueous phase 29.7 mL/10.3 mL volume ratio and 204.8 ppm Fe(VI) for BT, 29.5 min sonication time , 111.6 mg TOAB, organic phase/aqueous phase 16.2 mL/23.8 mL volume ratio and 245.3 ppm Fe(VI) for DBT) found by applying RSM based on BBD for which the ultrasonication time, TOAB amount, organic phase/aqueous phase volume ratio and Ferrate concentration in ppm unit are selected as independent variables, a sulfur removal of 88.3 and 91.8%, respectively, was obtained using 0.1 N acetic acid (pH = 4) from two model fuels (500 ppmw BT in toluene and 500 ppmw DBT in toluene) at 70 °C [241]. The optimum conditions found for BT and DBT were individually applied to diesel fuel containing 1428.6 ppmw total S, resulting in 85.7% BT and 91% DBT reduction in diesel oil. It was explained that these lower desulfurization yields compared to model fuels is due to the presence of many different sulfur compounds in diesel fuel that make oxidation difficult. The effect of different amounts of Ferrate and TOAB on UAODS was also investigated under 20 kHz frequency, 500 W and 40% amplitude direct US. When the ferrate concentration increased to a certain value, sulfur removal gave a maximum and after a certain value, sulfur removal decreased. This was attributed to the fact that as the ferrate concentration increased, the pH of the aqueous phase slightly increased (i.e., more basic medium), thus leading to a decrease in the oxidation capacity of the ferrate in basic medium (lower reduction potential (+0.72 V) in basic medium [242]). However, the standard reduction potential [243] of ferrate in acidic medium is + 2.20 V. With the excessive use of TOAB, the sulfur removal decreased, which has been attributed to the slowing of mass transfer due to turbidity of the mixture and to sterically prevention of electrophilic oxidation of sulfur compounds by the high concentration of alkyl groups. According to ANOVA results, it was reported that OP/AP volume ratio, PTC × (OP:AP) volume ratio interaction and PTC2 have the greatest effect on UAODS for BT, whereas OP/AP, US time × PTC interaction, US time × Ferrate concentration interaction, PTC × ferrate concentration, PTC2, (OP:AP)2 and (Ferrate conc.)2 have the greatest effect on the sulfur removal for DBT. It was determined that the amount of PTC for both model sulfur compounds is not important to UAODS. It has been pronounced that potassium ferrate has higher oxidation capacity and higher stability than HP and HP decomposes thermally at high temperature despite its lower cost, which is another important advantage of this process. Moreover, thermal decomposition [244] of potassium ferrate occurs above 198 °C. The oxidation mechanism is based on the formation of protonated Fe(VI) as a reactive complex [245–247] (which is much stronger oxidant than FeO4 2 −) by reaction of ferrate with acetic acid and, subsequently the transfer of the complex into the organic phase (where organic sulfur compounds are oxidized) by binding to the lipophilic cation of the phase transfer agent.By applying the Pareto-optimal analysis-based fuzzy logic model [248] in which US time, TOAB amount, organic phase/aqueous phase volume ratio and ferrate concentration are selected as four independent variables to maximize the sulfur reduction and, also US energy consumption, TOAB amount and the Ferrate amount are selected as three independent variables to minimize the operating cost, in the presence of acetic acid (pH = 4) at 70 °C under 20 kHz direct US with 200 W power output (500 W, 40% amplitude), it was reported that a conversion of 93.79% was achieved per operating cost of $ 0.830 at the optimum conditions (15.86 min US time, 107.7 mg TOAB, 30 mL:10 mL organic phase/aqueous phase volume ratio and 100 ppm ferrate concentration) for 500 ppmw BT, while a conversion of 88.36% was achieved per $ 0.769 operating cost at the optimum operating conditions (10 min US time, 100.1 mg TOAB, organic phase/aqueous phase volume ratio 16.96 mL/23.04 mL and 300 ppm ferrate concentration) for 500 ppmw DBT. It was shown that the desulfurization efficiencies obtained in this study are comparable with two sonoreactors in series in the previous studies [228,231], whereas the operating cost in this study is lower than that in the continuous sonoreactors connected in series, hence having the potential to be applicable for scaling up purposes.UAODS is performed at relatively much lower temperatures (i.e., in the range of room temperature to 90 °C), atmospheric or near atmospheric pressures, and generally shorter times than HDS. Process efficiency in UAODS is very important in terms of commercial applicability. In addition, US power intensity [125], defined as the power transferred to the liquid per surface area of the ultrasonic probe, and amplitude are important. It is beneficial to use low-amplitude ultrasound from the point of lower power and lower electricity consumption.As mentioned before, reaction and ultrasonic parameters have a very important effect on desulfurization. Increasing the amount of PTC up to a certain value improves UAODS by allowing more PTC-oxidant complexes to transfer into the organic phase and then ODS decreases slightly as a result of the slowing down of mass transfer between the aqueous-organic phase in the liquid mixture due to the formation of a thick turbid layer above an optimum amount of PTC [236,237,241]. As known, the reaction rate constant increases exponentially with increasing temperature according to Arrhenius equation, consequently increasing the reaction rate as well [249]. Nevertheless, above an optimum temperature, the collapse intensity of cavitation decreases as more solvent vapors will accumulate in cavitation bubble [120,250,251] in addition to decomposition of HP into water and oxygen, thus decreasing UAODS yield. Temperature can be increased unless the collapse intensity of the cavitation bubble reduces the total reaction rate [200]. Above an optimum reaction volume, sulfur removal decreases due to the lower ultrasonic power density [78,196]. With increasing HP concentration (i.e., a more concentrated HP solution) up to a certain value in aqueous phase, UAODS usually increases due the formation of more HO· radicals than HP [65,126]. Above an optimum concentration, HP can have a scavenging effect on hydroxyl radicals [157]. The sulfur removal increases up to a certain ultrasonic intensity, whereas dense bubble clouds, which show the cavitation shielding effect, will accumulate near the probe above a certain intensity [184]. Therefore, UAODS yield can decrease at high intensities and consequently, an optimum US intensity is required. Although generally, dissolved gases such as helium and oxygen in liquid mixture act as nucleation sites, facilitating the formation of the cavitation bubble, reaction rates change depending on the solubility, the thermal conductivity and the specific heat of the gases used [91,200]. However, dissolved gas above a certain concentration in cavitation bubble can cushion the collapse of the cavitation bubble, consequently causing a lower collapse intensity [252,253]. Therefore, it is necessary to find the optimum dissolved gas concentration in liquid mixture to increase the UAODS reaction rates unless the dissolved gas quantity decreases the cavitation effect. Pressure can have two opposite effects. As pressure increases, the intensity of the cavitation bubble implosion increases [254]. However, above an optimum pressure, much less bubbles, which can have almost no impact on overall reaction rate, can be produced due to increasing cavitation threshold of the liquid mixture [200]. The effect of pressure on sulfur removal varies as shown in Table S1 in the Supplementary Information and the boiling point of the solvent in the organic phase or the boiling range of fuel becomes crucial. For low boiling point solvents such as hexane, toluene, it is observed that sulfur removal increases with increasing pressure at relatively low operating temperature [164,232], while sulfur removal decreases with increasing pressure at high operating temperature [233]. For high boiling point solvents, it was reported that sulfur removal decreases with increasing pressure at relatively high temperature [197,237]. These differences observed in sulfur removal at high pressures can be attributed to a decrease or increase in the collapse intensity of cavitation bubbles. Nonetheless, much more effort is needed to establish a clear relationship between pressure and temperature in terms of cavitation intensity. In summary, in order to maximize total UAODS reaction rate, it is necessary to consider in combination the effects of reaction and ultrasonic parameters on UAODS yield.Desulfurization process efficiency (DPE = UAODS yield/MR(H2O2/S)) can be defined as the UAODS yield per molar ratio of reactants used (i.e., the molar ratio of hydrogen peroxide to sulfur). The less the amount of HP, the larger the quantity of fuel used to remove sulfur and the higher the UAODS yield, the higher the process efficiency. Figs. 4, 5 and 6 show DPEs calculated using heterogeneous catalysts, homogeneous catalysts in the absence of PTC and homogeneous catalysts in the presence of PTC, respectively. The operating conditions of UAODS reactions with heterogeneous catalyst, homogeneous catalyst in the absence of PTC and homogeneous catalyst in the presence of PTC are given in Tables S2, S3 and S4, respectively. From the three figures, it can be seen that the DPEs under indirect US (ultrasonic bath) are mostly lower than the DPEs under direct US. This low process efficiency can be attributed to the fact that the intensity of the indirect US (in this case, the ultrasonic wave generated by the transducer passes first through the walls of the sample container and then through the liquid) is much lower compared to the intensity of ultrasound in direct contact with liquid using the ultrasonic probe [255]. Also, in an ultrasonic bath, ultrasonic wave cannot propagate equally in all directions into each fluid element in a liquid, thus resulting in heterogeneous dissipation [256–258].It can be seen that in the case of using heterogeneous catalysts, the DPEs are generally higher than DPEs with and without PTC using homogeneous catalysts. These high DPEs can be due to both the adsorption of sulfur compounds on the catalyst surface and the oxidation of sulfur compounds by forming an active oxidizing complex caused by HP on the surface, as well as the adsorption of oxidized sulfur compounds. There are many advantages of using solid catalyst in liquid under US irradiation: solid particles function as nucleation sites to form cavitation bubbles, thus causing free radicals to increase further. Sonication results in an increase in surface area by reducing the particle size of solid catalysts and inactive catalyst becomes reactive as a result of desorption of adsorbed sulfones (passivating surface coating) due to the surface cleaning caused by liquid jet streams which are formed by implosion of cavitation bubbles [259]. In addition, more collision occurs between reactants and catalysts due to microstreaming [250] and agglomeration of catalysts is prevented [260]. Moreover, the high heat generated by the collapse of cavitation bubbles near solid catalysts can propagate inside catalyst, consequently leading the reaction rate to be higher and it is emphasized that the largest sonochemical effect occurs in macropores >50 nm in diameter [261]. On the other hand, too many catalyst particles can attenuate US waves propagating through liquid [125]. Therefore, an optimum catalyst loading is necessary in UAODS reactions.There is an exception in the case of using potassium ferrate in Fig. 6. As potassium ferrate is a stronger oxidant in acidic environment than HP and the active complex consisting of ferrate and acetic acid has higher oxidation power than ferrate alone, DPEs are very high.DPEs for acetic acid-HP and formic acid-HP in Fig. 5 are generally higher than those for the phosphotungstic acid-HP system in Fig. 6, which is due to the small molecular size of acetic acid [262] (ca. 0.4 nm) and formic acid [263] (ca. 0.3 nm), thus alkyl substituted aromatic sulfur compounds do not cause steric hindrance. The reason that DPEs are lower in the case of using phosphotungstic acid-HP system in the presence of PTC in Fig. 6 compared to DPEs in the case of using homogeneous catalysts without PTC is that the alkyl groups adjacent to the sulfur atom of compounds such as 2,5-DMT, 4-MDBT and 4,6-DMDBT in fuel lead to the steric hindrance due to bulky size of the oxidizing polyoxoperoxo complex composed of phosphotungstic acid and HP. However, when organic acids such as formic acid and acetic acid are used in combination with phosphotungstic acid, DPE increases considerably by creating a synergistic effect due to the polyoxoperoxo complexes and peracids formed [236,238]. The reason for using PTCs in the case of phosphotungstic acid is the transfer of the formed polyoxoperoxo complex anion to the organic phase, otherwise DPE without PTC may be low. Also, phosphotungstic acid decomposes as pH increases from 1 to 8.3 [264] and thus an acidic medium is favorable to the UAODS reactions. Since phosphotungstic acid is thermally stable [265] up to 400 °C, it can form stable polyoxoperoxo complexes with HP and hence ODS can be performed at relatively higher temperatures, which are below 100 °C, compared to the temperatures in the case of acetic acid and formic acid. Performic acid [266] and peracetic acid [267] undergo dramatically thermal decomposition, especially at temperatures of 45 °C and above.Formic acid and acetic acid have the capacity to extract sulfur compounds and peracids formed as a result of emulsification by US effect can easily be transferred into the organic phase or the organic-aqueous phase interface. Therefore, it can be deduced that PTC has no significant effect on DPE. In the studies in Fig. 6, it is seen that PTC is used in addition to phosphotungstic acid. The reason for using PTC may be due to the low desulfurization obtained by using phosphotungstic acid in the absence of PTC.In Fig. 4, modified Metal-organic Framework (MOF) was used in the study where DPE of 49, 43.5 and 35.5% was obtained. The reason for the high DPE can be both the entrapment of phosphotungstic acid into amino-functionalized MOF with large surface area and pore volume (hence aromatic sulfur compounds are effectively adsorbed and oxidized on phosphotungstic acid@TMU-17-NH2), and the simultaneous extraction of oxidized sulfur compounds using acetonitrile. In addition, ultrasonic synthesis, which is more environmentally friendly and performed at lower reaction time at room temperature than solvothermal process carried out at high temperature, may have contributed to high desulfurization as MOFs synthesized under ultrasound have generally higher surface area, lower particle size, higher crystallinity, more uniform morphology and size distribution compared to those obtained by conventional preparation methods.Reactor configurations also affect DPE. In Fig. 5, the high DPE of 23.57 is due to the nozzle, through which the aqueous phase consisting of FA and HP flows in a very low amount (0.71 mL min−1), placed just below the tip of ultrasonic probe, thus causing an increase in sulfur removal by generating active radicals in this efficient region and dispersing the aqueous phase more homogeneously into the organic phase.In ODS, ionic liquids have also been tried instead of the aqueous phase. However, their synthesis is generally high cost and it is difficult to transport them due to their high viscosity. In addition, as more US power is needed to fully emulsify the high viscosity ionic liquid phase and organic phase, the operating cost will increase due to electrical energy consumption. Moreover, since the ionic liquid loses its activity after a certain recycle, its regeneration will also lead to an additional cost. Therefore, the use of ionic liquids in continuous processes is not practical.In the studies, one of the biggest problems of UAODS is fuel loss during extraction and/or adsorption process to remove oxidized sulfur compounds after oxidative treatment. During the separation processes, other polar hydrocarbons in fuel pass into the extractant phase or are adsorbed on the adsorbent. Although it has been shown in laboratory and pilot studies that the physicochemical properties of the fuel after the UAODS process change in acceptable ranges according to the fuel specifications for petroleum fractions, how these properties will change in large-scale industrial production is a separate research topic. In addition, the ultrasonic probe must be replaced with the new one as the tip surface erodes by pitting in long service life [125,268], otherwise it becomes inoperable.One of the biggest reasons for the widespread use of HDS is that it has a high fuel recovery as well as a very little negative effect on fuel properties. In addition, hydrotreatment of diesel consisting of paraffinic, aromatic and naphthenic components saturates the aromatic compounds in the diesel, resulting in an increase in the cetane number [269].After UAODS, how to eliminate the waste sulfones generated and accumulated is an environmental issue. Elemental sulfur, which is mainly used for sulfuric acid production [270], can be produced by the reaction of SO2 with H2S generated in HDS units after the waste sulfones are converted to SO2 as a result of thermal decomposition [271] by burning them in high temperature furnace operating at 1093–1427 °C in the Claus process [272,273] or by pyrolysis [274].In a study [275] evaluating the desulfurization process economics by using Aspen Plus simulation, it has been shown that the UAODS process is not cost-effective for fuels containing high sulfur (i.e., in the range of several thousand ppmw) due to high chemical consumption to drastically reduce the sulfur content of fuel and very high amounts of extraction solvent required to separate the huge amounts of sulfones formed, therefore it is not competitive with HDS. Therefore, detailed research taking fuel loss into account is still needed to achieve cost savings and high sulfur removal in the UAODS process by using low amounts of reagents, performing reactions at the lowest possible temperature in the shortest possible time and using the most efficient extraction solvent in the lowest possible amounts.Concluding remarks and future directions can be presented as follows: - In order to increase the sulfur removal per power density consumed as well as to reduce the process cost, one continuous-flow sonoreactor or two continuous-flow sonoreactors in series can be used at low flow rate of the aqueous phase feed and, short retention times. At high conversions, continuous sonoreactors can be connected in parallel to treat more fuel. - Desulfurization can be increased by the addition of heterogeneous catalysts to continuous sonoreactors connected in series. - Potassium ferrate with a much higher reduction potential than HP under acidic conditions can be activated by HCl, HNO3, H2SO4, HClO4 and HCOOH instead of CH3COOH. To reduce the process cost, UAODS reactions can be carried out using potassium ferrate in acidic medium in the absence of relatively expensive PTCs. - Low temperatures in the range of 20–40 °C favor UAODS reactions since the decomposition of performic and peracetic acid increases drastically above 40 °C in the case of homogeneous catalysts. To observe the change of concentration of peroxycarboxylic acid over time, the reactions of HP and carboxylic acids (i.e., HCOOH or CH3COOH) can be carried out at different temperatures, different times and various molar ratios in the absence of both PTC and organic phase under US irradiation and consequently, peroxyformic acid or peroxyacetic acid (HCOOOH or CH3COOOH) concentration at any time t during the reaction can be readily determined by titrimetric analysis. Eventually, the time, at which peroxyformic acid or peroxyacetic acid concentration is maximum, is found for each temperature. Therefore, UAODS reactions can be performed at those times, thus reducing the process cost due to short reaction times and increasing the sulfur removal efficiency. Alternatively, UAODS reactions can be performed at different temperatures and by taking an aliquot of the aqueous phase at certain times during the UAODS reaction for each temperature, the change of the concentration of the peroxycarboxylic formed can be followed by titrimetric analysis. Consequently, a relationship between the sulfur removal and peroxycarboxylic acid concentration can be established and sono-oxidative desulfurization reaction conditions can be optimized. - Indirect ultrasonic application in UAODS reactions is not as effective as direct US application from the point of view of DPE. In order to increase the sulfur removal per power density consumed as well as to reduce the process cost, one continuous-flow sonoreactor or two continuous-flow sonoreactors in series can be used at low flow rate of the aqueous phase feed and, short retention times. At high conversions, continuous sonoreactors can be connected in parallel to treat more fuel.Desulfurization can be increased by the addition of heterogeneous catalysts to continuous sonoreactors connected in series.Potassium ferrate with a much higher reduction potential than HP under acidic conditions can be activated by HCl, HNO3, H2SO4, HClO4 and HCOOH instead of CH3COOH. To reduce the process cost, UAODS reactions can be carried out using potassium ferrate in acidic medium in the absence of relatively expensive PTCs.Low temperatures in the range of 20–40 °C favor UAODS reactions since the decomposition of performic and peracetic acid increases drastically above 40 °C in the case of homogeneous catalysts. To observe the change of concentration of peroxycarboxylic acid over time, the reactions of HP and carboxylic acids (i.e., HCOOH or CH3COOH) can be carried out at different temperatures, different times and various molar ratios in the absence of both PTC and organic phase under US irradiation and consequently, peroxyformic acid or peroxyacetic acid (HCOOOH or CH3COOOH) concentration at any time t during the reaction can be readily determined by titrimetric analysis. Eventually, the time, at which peroxyformic acid or peroxyacetic acid concentration is maximum, is found for each temperature. Therefore, UAODS reactions can be performed at those times, thus reducing the process cost due to short reaction times and increasing the sulfur removal efficiency. Alternatively, UAODS reactions can be performed at different temperatures and by taking an aliquot of the aqueous phase at certain times during the UAODS reaction for each temperature, the change of the concentration of the peroxycarboxylic formed can be followed by titrimetric analysis. Consequently, a relationship between the sulfur removal and peroxycarboxylic acid concentration can be established and sono-oxidative desulfurization reaction conditions can be optimized.Indirect ultrasonic application in UAODS reactions is not as effective as direct US application from the point of view of DPE.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.ultsonch.2021.105845.The following are the Supplementary data to this article: Supplementary Data 1

Recently, environmental pollution has increased significantly due to petroleum-based fuels widely used in vehicles. This environmental pollution is mainly due to the acidic SO2 gas generated by the combustion of fuels and emitted into the atmosphere. SO2 gas causes not only acid rain but also corrosion of metal parts of engines in vehicles. In addition, it functions as a catalyst poison in catalytic converters in exhaust system. Due to these damages, strict regulations have been introduced to reduce the amount of sulfur in fuels. As of 2005, the permissible amount of sulfur in diesel fuels in Europe and America has been limited to 10 and 15 ppm by weight, respectively. Due to the decreasing oil reserves in the world, high viscosity petroleums containing high sulfur and heavier fractions (i.e., low-quality oils) are increasing, thus making desulfurization difficult and leading to high costly process. Since time and economic loss are very important today, these two terms have to be reduced to a minimum. Recently, ultrasound wave in ODS shown as an alternative to HDS is utilized to further increase desulfurization in shorter times. Ultrasound wave locally creates high temperatures and high pressures (hot-spot theory) in liquid, causing the desulfurization reaction to accelerate further. In this review, the advantages and difficulties of oxidative desulfurization, the economics of ultrasound-assisted oxidative desulfurization are summarized and recommendations for improving the process are presented.

The extensive combustion of fossil carbon is most probably responsible for the growing concentration of greenhouse gas CO2 in the atmosphere. The concerns about global warming turned attention towards the production of biofuels by upgrading non-edible and waste vegetable oils and animal fats [1–8]. The most widely used production method of diesel range bio-oil, generally referred to as biodiesel, is catalytic transesterification of latter renewable triglycerides by lower alcohols [1,6,7]. However, the biodiesel cannot fully replace conventional diesel oil, because of its lower energy density, higher viscosity, moderate oxidation stability, and limited compatibility with fossil fuel [4,6]. A better alternative of triglyceride upgrading is deoxygenation via hydroprocessing that is providing a mixture of hydrocarbons. The hydrocarbon mixture is second generation biofuel, often referred to as biogasoil or green diesel. Biogasoil has comparable or even better fuel properties than conventional diesel fuel [6,7,9].Catalytic triglyceride HDO can be carried out at moderate temperature (200 – 400 °C) and hydrogen pressure (<50 bar). The applied catalysts are those, routinely used in the petroleum industry for hydroprocessing, such as, sulfided cobalt and nickel molybdate catalysts, supported noble (Pt, Pd, Ru) and non-noble metals (Ni, Cu, Co), metal phosphides, metal oxides, etc. [3,4,6,10]. Application of monometallic transition metal catalysts are very common for the deoxygenation of fatty acids and triglycerides [3,6,10]. Palladium based catalysts are especially preferred because of the peculiar ability of the palladium metal to activate hydrogen for reaction [10]. Nevertheless, there is a general agreement that the catalyst support does not only provide high surface area to stabilize high metal dispersion, but it also has significant contribution to the HDO activity and selectivity by its acid-base property [4,6,10]. Supports, having strong Brønsted acidity, such as H-zeolites, are less favored because they initiate cracking of long chain paraffins and condensation reactions, producing coke precursors and coke that deactivates the catalyst [6,10]. Therefore, supports of mild-to-moderate acidity, such as activated carbon, TiO2, ZrO2, SiO2, and Al2O3 were found suitable for HDO catalysts [4,6,10]. The most often used support is γ–Al2O3 [4,8].The hydroconversion of triglycerides to paraffins proceeds in the consecutive steps of ester bond hydrogenolysis, giving carboxylic acid, and deoxygenation of the acid to paraffin. The deoxygenation reaction is the rate determining step of paraffin formation [3,4,6,11]. Former studies showed that the deoxygenation reaction of the carboxylic acid intermediate over Pd catalysts follows pathways resulting in the formation of CO (hydrodecarbonylation, HDCO), CO2 (hydrodecarboxylation, HDCO2), and H2O (H2-reduction of oxygen, HDH2O). Over supported metal catalysts the HDCO was found to be the major reaction route, whereas the HDCO2 and HDH2O were reaction routes, which had only minor contribution to the HDO reaction [7,8]. The mechanisms of these HDO pathways are not fully understood. It was suggested that the HDCO reaction proceeded through direct hydrogenolysis of the carboxylic acid to paraffin and formic acid (C–C bond scission), which reaction step was followed by quick decomposition of HCOOH to CO and H2O [6,8,11]. Deoxygenation on the HDH2O route was proposed to proceed by deoxygenation of carboxylic acid to aldehyde intermediate [2,3,5,8]. Accordingly, formation of aldehyde from carboxylic acid involves hydrogenation/dehydration reactions (hydrogenation of C = O bond to CH–OH followed by H2O formation involving C–O bond scission). The surface-bound aldehyde intermediate is then further hydrogenated to paraffins by releasing either H2O or CO, corresponding to routes HDH2O and HDCO, respectively [3,5].In the present study, alumina-supported Pd catalysts were prepared and studied to learn more about the mechanism of triglyceride HDO reaction. The effect of support phosphatization on the catalyst structure, acid-base properties, and activity was investigated. The HDO activity was tested using tricaprylin and valeric acid model compounds. Quasi–operando DRIFTS investigation provided insight in the chemistry of surface intermediate formation during the catalytic reaction and permitted to come to important conclusions, concerning some mechanistic details.Alumina-supported palladium catalyst was perepared by impregnating 10 gramms of γ–Al2O3 (Ketjen CK-300, Alfa Aesar) by 10 cm3 of aqueous Pd(NH3)4(NO3)2 (product of Strem) solution. The concentration of the solution was adjusted to get catalyst of 0.5 wt% Pd content. The sample was dried at 110 °C for 16 h. To decompose the metal precursor salt the sample was calcined. It was heated first at a heating rate of 2 °C min−1 to 150 °C, kept at this temperature for 1 h, and then the temparature was raised at a heating rate of 4 °C min−1 to 350 °C. The catalyst was kept at this tempreture for additional 4 h. The obtained catalyst sample was designated as Pd/Al2O3.The phosphatized-alumina-supported Pd catalyst samples were prepared following the same procedure as above, except that the alumina support was phosphatized first to different extents. Supports with 1.0, 2.5, and 5.0 wt% phosphorous content were prepared by impregnating 10 g of γ–Al2O3 with 10 cm3 of a solution, containing calculated amount of phosphoric acid. The impregnated samples were dried at 110 °C for 16 h then calcined in air at 550 °C for 4 h to generate the phosphatized alumina supports. The thus obtained phosphatized γ–Al2O3 catalyst supports are designated as Al2O3–1P, Al2O3–2.5P and Al2O3–5P, respectively. These supports were used to prepare catalysts of 0.5 wt% Pd content. The corresponding supported Pd catalysts were designated as Pd/Al2O3–1P, Pd/Al2O3–2.5P, and Pd/Al2O3–5P.The P and Pd content of the catalyst samples were determined by using Inductively Coupled Plasma Optical Emission Spectroscopic (ICP-OES) method (Spectro Genesis ICP-OES apparatus).Nitrogen adsorption isotherms of the catalyst samples were determined at −196 °C using an automatic, volumetric adsorption analyzer (The “Surfer”, product of Thermo-Fisher Scientific). The sample was dehydrated before the measurement at 250 °C under high vacuum (10-6 mbar) for 4 h. The SSA of the catalyst samples was determined by the BET method, whereas the pore-size distribution was calculated using the Barett-Joyner-Halenda (BJH) method.The XRPD measurements were carried out using a Philips PW 1810/3710 X-ray powder diffractometer in a Bragg-Brentano parafocusing arrangement applying monochromated Cu Kα (λ = 1.5418 A) radiation.The dispersion of Pd in the catalysts was determined by CO pulse chemisorption method. About 100 mg of the sample was placed into a quartz microreactor (I.D.: 4 mm) and reduced in situ in hydrogen flow at 450 °C for 1 h. It was flushed then by He flow at 450 °C for 30 min and cooled to room temperature in the He flow. In 3 min intervals carbon monoxide pulses of 10 µL volume were injected sequentially into the He flow, passing through the catalyst bed. The CO concentration of the reactor effluent was monitored using thermal conductivity detector (TCD). The TCD signal was processed by computer. The introduction of CO pulses was continued until the chemisorption sites were saturated. After calibration the molar amount of chemisorbed CO was calculated from the areas of the TCD signals.The CO2-TPD measurement was used to characterize the basicity of the supports. About 150 mg of the sample was placed into a quartz microreactor (I.D.: 4 mm) and activated in O2-flow at 550 °C for 1 h. The sample was then flushed with N2 for 15 min at 550 °C, evacuated at the same temperature for 30 min and cooled to room temperature. Adsorption of CO2 was carried out by contacting the sample with CO2 gas at 13.3 kPa for 15 min. The system was flushed by He and the temperature of the reactor was ramped up at a rate of 10 °C min−1 to 700 °C. The CO2 concentration in the gas flow was monitored by TCD.A Nicolet 6700 FT-IR (Thermo Scientific) instrument was used in transmission mode to record the spectra of the surface species present on the neat supports and catalysts and obtained from adsorption of compounds. Self-supporting wafer of the examined sample having a “thickness” of 5–10 mg cm−2 was placed into the sample holder of a stainless steel spectroscopic cell equipped with CaF2 windows and a furnace section for in situ activation of the sample either in atmospheric gas flow or under high vacuum. Spectra were taken by averaging 512 scans at a nominal resolution of 2 cm−1.The acidity and basicity of the supports were studied by determining the spectra of adsorbed pyridine (Py) or CO2, respectively. Prior to Py adsorption the sample was pretreated at 450 °C under high vacuum (10-6 mbar) for 1 h then the temperature was lowered to 200 °C and the sample was contacted with Py vapor at 5 mbar for 30 min. The sample was cooled then to 100 °C. The Py vapor was removed from the cell by successive evacuation at temperatures 100, 200, 300, 400, and 450 °C for 30 min at each temperature. After each evacuation a spectrum was recorded at room temperature. The spectrum of the wafer, recorded before Py adsorption, was subtracted from each spectrum to obtain the spectrum of the adsorbed species only.A procedure, similar to that described above, was followed to determine the spectra of the species obtained from adsorption of CO2. Wafer of the activated sample was contacted with CO2 at 15 mbar at room temperature for 30 min. The spectrum of the adsorbed species from CO2 was measured after successive evacuation at room temperature, 100, 200, 300, and 400 °C under high vacuum for 30 min at each temperature.The electronic state of Pd in the catalysts was characterized by analyzing the FTIR spectrum of adsorbed species formed from CO. Prior to CO adsorption, the Pd-containing catalyst wafer was reduced at 450 °C in H2 stream for 1 h. The catalyst was contacted with CO gas at 5 mbar at room temperature for 20 min then the spectrum of the carbonyl species formed was recorded after 20 min evacuation under high vacuum at room temperature. Each absorbance spectra were scaled to 5 mg cm−2 wafer thickness to allow quantitative comparisons.Spectra were recorded using a Varian NMR System spectrometer operating at 600 MHz 1H frequency (242.74 MHz for 31P and 156.26 MHz for 27Al) with a Chemagnetics 3.2 mm narrow-bore triple resonance T3 probe in double resonance mode. The 31P direct polarization spectra were recorded (160 transients) at 20 °C with 12 kHz of spinning rate and 300 s repetition delay. The 27Al – 1H cross polarization spectra were recorded (12000 transients) with 0.5 ms contact time, 5 s of repetition delay at 20 °C with 12 kHz of spinning rate. For both experiments SPINAL 1H decoupling was used. As chemical shift reference ammonium dihydrogen phosphate (δiso = 0.81 ppm with respect to 85 wt% H3PO4 solution) for the 31P and sodium aluminate (δiso = 79.3 ppm) for the 27Al measurements was used.Hydroconversion of tricaprylin was investigated using a high pressure fixed-bed flow-through microreactor system. The catalytic reactor (I.D.: 8 mm) was filled with 2.0 g of catalyst using its 0.315–0.630 mm sieve fraction. Prior to the catalytic run, the catalyst was reduced in situ in 50 cm3 min−1 flow of H2 at 450 °C for 2 h at atmospheric pressure, then the temperature was lowered to the desired reaction temperature (300 or 350 °C) and the pressure was increased to 21 bar total pressure. The tricaprylin reactant was fed into the reactor using a high-pressure syringe pump (ISCO) at a weight hourly space velocity (WHSV) of 4 g t r i c a p r y l i n g c a t a l y s t - 1 h - 1 , whereas the H2/tricaprylin molar ratio was 20. The product mixture was cooled to room temperature and the liquid products were separated from the gas products in a reflux condenser downstream of the reactor. The effluent gas leaving the condenser through a back pressure regulator valve contained mainly H2, and CO (as main product), CO2, and minor amount of CH4. The effluent was analysed on-line using a gas chromatograph (GC) equipped with a TCD detector and a 60/80 Carbonex-1000 (L 15.0ft × OD 1/8″) stainless steel column. The liquid product from the presurized collection vessel was first transferred into a closed atmospheric vessel. During this transfer, the pressure of the vessel reached the system pressure, which was then released via a transfer line connected to a syringe, where the gas, expanded to atmospheric pressure, was collected. Note that this latter expanded gas contained all of those products, which were in liquid phase under system pressure but appeared as gases at atmospheric pressure (mainly propene, and some ethane in H2). The liquid sample, drained now from atmospheric pressure, was analyzed by GC equipped with flame ionization detector (FID) using CP-FFAP CB (L 25.0 m × ID 0.32 mm × df 0.3 μm) capillary column, whereas the expanded gas products were analysed by GC-TCD-FID using ShinCarbon ST (L 2.0 m × ID1/8in. × OD 2.0 mm) column. Samples were taken in every hour after the steady state was reached (after one hour time of stream). At high tricaprylin conversions, when the reactant was converted almost completely into paraffins, the liquid product split into two clear, colorless hydrocarbon and water phases. Since the lower aqueous phase contained only a negligible amount of organic components, in such cases only the composition of the upper organic phase was analyzed. In some cases, when the tricaprylin conversion was not complete, and/or was not completely converted into paraffins, a single phase liquid product was obtained, which was analysed for the organic components. In some cases, when unknown components were also formed, GC–MS equiped with Rxi-5Sil MS (L30.0 m × ID0.25 mm × df0.25 μm) column was used for peak assignments.Quasi-operando DRIFT spectroscopic investigation of the adsorbed species formed from carboxylic acid intermediate under catalytic conditions is crucial in order to reveal structure–activity relationships of the catalyst in the HDO reaction of triglycerides or fatty acids. The experiments were carried out using an FT-IR spectrometer (Thermo Nicolet iS10) equipped with a Collector II diffuse reflectance mirror system and a flow-through DRIFT spectroscopic reactor cell (I.D.: 5 mm, height of catalyst bed ∼ 4 mm) filled with about 20 mg of powdered catalyst sample. The design of the cell allows carrier gas or gas phase reactant mixture to flow through the catalyst bed in the sample cup. The reactant is introduced into the cell by switching the carrier gas flow to a gas saturator containing the reactant at room temperature. Due to the experimental difficulties related to the very low vapor pressure of caprylic acid, valeric acid (VA) was used as model compound in these experiments. First, the catalyst was pre-treated in situ in a 50 cm3 min−1 H2 flow at 450 °C for 2 h and then the spectrum of the activated catalyst powder was collected (512 scans at a nominal resolution of 2 cm−1) at the desired reaction temperatures (300 or 350 °C). The reaction of VA was initiated by switching the H2 or He flow (50 cm3 min−1) to the gas saturator. The thus obtained H2/VA or He/VA mixture contained 258 ppm VA. The spectrum taken in the presence of the reacting gas mixture at 300 or 350 °C was corrected with the spectrum of the pure catalyst at the same temperature. Since the contribution of the gas phase spectrum was found negligible under the applied conditions, the thus obtained difference spectrum practically reflects the bands of surface species formed (positive bands) or consumed (negative bands) in the VA adsorption/reaction process.The effluent gas leaving the reactor cell was continuously monitored by online mass spectrometer (MS; VG ProLab, Thermo Scientific) following the characteristic masses of the major reaction products: butane (m/z = 58, C4H10 +), pentane (m/z = 72, C5H12 +), CO (m/z = 28, CO+), and CO2 (m/z = 44, CO2 +). All signals were corrected for the contribution of other reaction products giving a fragment at the same m/z value.The measured Pd and P contents of the catalysts, listed in Table1 , are in good agreement with the values that follow from the applied conditions of catalyst preparation.The nitrogen adsorption isotherms, shown in Fig. S1, are characteristic for mesoporous oxides. They are classified as type IV isoterms, having H2 type hysteresis loop. The SSA of the catalysts decreased as their phosphorous content was increased (Table1). The SSA of the Pd/Al2O3–5P catalyst is about 40% lower than that of the Pd/Al2O3 catalyst. These results suggest that phosphate groups can block some pores of the alumina support and thereby decrease the SSA.XRPD patterns of the phosphatized catalysts and that of the parent γ–Al2O3 support are shown in Fig. 1 . The diffractograms of the support and all the catalysts were similar, i.e., no new crystalline phase could be detected (Fig. 1). The results suggest that the size of the Pd or PdO particles on the support is well below the detection limit of the XRPD method (the diameter was less than about 5 nm).The Pd dispersion (D Pd) was obtained as the ratio of the number of surface Pd atoms and the total number of Pd atoms in the catalyst. The molar amount of chemisorbed CO was taken to be equivalent with half of the molar amount of surface Pd atoms [12].Assuming spherical particle shape and that the three low-index planes are in equal proportions on the polycrystalline surface of the face-centered cubic crystals of the metal, the mean Pd particle size (d Pd) was calculated from the dispersion by the equation (1) d P d = 6 v a D P d where v is the volume occupied by a single Pd atom in the bulk of metal (1.47·10–2 nm3), and a is the average surface area occupied by one Pd atom (7.93·10–2 nm2) [13]. The Pd dispersions and the mean particle sizes, listed in Table1, barely changed with the phosphorous content of the catalysts.The FT-IR spectra measured for the phosphate-free and phosphatized γ–Al2O3 supports in the range of stretching vibration of surface OH groups are shown in Fig. 2 . The assignment of the alumina OH bands, first given by Knözinger and Ratnasamy [14], was later refined by Busca et al. [15–17]. The νOH bands above about 3700 cm−1 were assigned to terminal OH groups. The band at 3792 cm−1, appearing as a weak shoulder in the spectrum of pure γ–Al2O3 (Fig. 2a) was attributed to νOH vibration of groups linked to tetrahedral aluminum ions. The band at 3773 cm−1 belongs to OH groups next to a coordinately unsaturated tetrahedral aluminum atom, i.e. next to a Lewis acid site. In a very recent paper [18] the assignment of the band at about 3780–3770 cm−1 was made more accurate. The band was shown to stem from hydroxyl groups bound to octahedrally coordinated surface Al3+ ions (O5AlVI-OH sites) that transforms to pentacoordinated Al3+ surface sites (“bare” O5AlV sites) upon thermal dehydroxylation.The relatively broad band, centered at 3730 cm−1 (Fig. 2a) was assigned to OH groups bound to octahedrally coordinated Al3+ ions [15–17]. The presence of a coordinately unsaturated octahedral aluminum atom adjacent to an OH group, results in OH band, shifted to the 3740–3700 cm−1 frequency region [15–17]. However, we could not distinguish these bands because of the band broadening and strong overlapping (Fig. 2a). The OH bands below about 3700 cm−1 are attributed to different bridging OH species [15–17]. Thus, the broad band centered approximately at 3670 cm−1 can be assigned to bridging OH species, whereas the very broad feature around 3590 cm−1 can be attributed to triply bridging OH species or OH groups in hydrogen bonding interactions [15–17].The intensity drop of the OH bands (Fig. 2, b-d) clearly indicates OH consumption in a reaction with phosphoric acid. The reaction has been described as sort of surface acid-base neutralization reaction resulting in the formation of surface phosphate species and water [19–21]. The formation of phosphate species is clearly indicated by the appearance of the band at 3677 cm−1, assigned to the νOH vibration of P–OH species (Fig. 2, b-d) and a strong νP=O vibrational band of the phosphate groups in the range of 1150–1250 cm−1 frequency (not shown). Interestingly, the characteristic νOH band of the terminal OH groups on tetrahedral aluminum ions at 3792 cm−1 gained in intensity as the concentration of the surface phosphate species increased (Fig. 2, a-d), suggesting that a surface reaction between phosphoric acid and alumina surface is not a simple acid-base neutralization reaction.The surface basicity of the phosphate-free and phosphatized γ–Al2O3 supports was characterized by the carbonate-like surface species obtained from CO2 adsorption. The IR spectrum of these species was recorded and the spectral features were assigned on the basis of the available literature [17,21,22] (Fig. 3 ). The CO2 uptake of the supports resulted mainly in the formation of two types of bicarbonate species as indicated by the appearance of the asymmetric νO-C=O vibrations at 1645 cm−1 for both B1 and B2 type bicarbonate species and the symmetric vibrations at 1454 cm−1 of the B1 and 1482 cm−1 of the B2 type bicarbonate species (Fig. 3A). The corresponding νOH bands of the bicarbonate species appear as positive bands at 3610 (B2 type) and 3595 cm−1 (B1 type, as a shoulder) in the range of the O–H vibration frequencies (Fig. 3B). The most intense negative νOH bands at 3773 and around 3700 cm−1 (Fig. 3B) suggest that mainly those OH groups were involved in the formation of bicarbonates, which have a coordinately unsaturated tetrahedral or octahedral aluminum atom in their neighborhood. Note that the involvement of these hydroxyl groups in the CO2 adsorption could be just due to their location near to Lewis acid sites [16]. The lack of negative band at 3677 cm−1, where the νOH band of the phosphorus–bound hydroxyls appears, implies that the P–OH groups do not participate in bicarbonate formation (Fig. 3B). In harmony with the conclusion of other authors [21,23], this result permitted for us to conclude that phosphatization does not generate basic centers for CO2 uptake, i.e., the P–OH group is stronger acid than the CO2.The CO2 uptake on the supports results also in the formation of bidentate chelating carbonate that gives weak IR band at 1670 cm−1 (shoulder) and monodentate carbonate bands at 1542and 1404 cm−1 (shoulders) frequencies (Fig. 3A) [17,22]. Formation of these carbonate species probably takes place with the involvement of coordinately unsaturated aluminum cation and oxide anion pairs, which oxide ions are Lewis base sites on the alumina surface [16,17,21]. It is important to note that the strength of the characteristic absorption bands of the carbonate species is inversely proportional to the phosphate loading (Fig. 3A, a-d). These results suggest that formation of phosphate species is accompanied not only by consumption of basic surface OH groups, but also by elimination of Lewis base sites.The concentration and distribution of basic sites on the parent and phosphatized γ–Al2O3 supports were determined by CO2-TPD measurements. The TPD curves in Fig. 4 indicate the presence of at least three overlapping component bands. According to Wang et al. [22], a peak appearing around 80 °C is due to the decomposition of bicarbonate species formed on weak basic sites, whereas the component peaks observed around 160 and 250 °C can be attributed to decomposition of chelating bidentate carbonate species formed on medium strength basic sites and monodentate carbonate species formed on strong basic sites, respectively. These assignements are clearly supported by the observed thermal stability of different carbonate species (Fig. S2). An additional high temperature peak, attributed to bridging bidentate carbonate species formed on strong basic sites, may also appear at around 325 °C temperature. Therefore, the CO2-TPD curves shown in Fig. 4 were resolved by peak fitting using four component peaks. The peak fitting process resulted essentially in three major component peaks with a maximum around 85, 135, and 220 °C, representing weak, medium strength and strong basic sites (the fourth component peak had a negligible intensity on each sample, therefore it was ignored). The total amount of different basic sites and their distribution was calculated from the area under the corresponding curves using the result of a previous calibration measurement (Table S1). The concentrations of the weak-to-moderate strength basic sites and the strong basic sites are listed in Table2 . Results show that introduction of phosphate groups significantly decreased the concentration of all types of basic centers on the surface of the alumina support, which dropped by more than 90% for the support, having the highest phosphorous content.Adsorption of Py on the Lewis acid sites of alumina results in the formation of coordinately bonded Py giving absorption bands in the range of 1630–1590 (8a band) and 1460–1430 cm−1 (19b band) [16,17,25]. The pair of bands observed at 1623 and 1456 cm−1 in Fig. 5 a can be assigned to Py adsorbed on tri-coordinated Al3+ cations (tetrahedral Al cations with coordinative unsaturation), which represent the strongest acid Lewis sites of alumina. The second pair of bands at 1615 and 1451 cm−1 can be also attributed to Py adsorbed on Al ions with coordinative unsaturation, which are most probably in octahedral coordination [16,17,25] and represent weaker acid Lewis sites than those of the coordinately unsaturated tetrahedral aluminum cations.The spectra of adsorbed CO2 support above identification of Py sorption sites. Bands at 2360 and 2345 cm−1, which belong to the so called Σ u + mode of linearly coordinated (end-on adsorbed) CO2 [17,21,26], were found to develop in the presence of CO2 gas in the IR cell (Fig. S3). These bands are most probably bands of CO2, bound to strong and weak, coordinately unsaturated Lewis acid Al centers in tetrahedral and octahedral coordination, respectively.The intensity of Py bands is lower for the supports having higher phosphate concentration (Fig. 5, a-d), suggesting that formation of surface phosphate affected the concentration of both types of Lewis acid sites.The bands of linearly adsorbed CO2 are also weaker for the phosphatized supports, indicating that surface phosphate eliminates Lewis acid sites of alumina (Fig. S3, a-d).Evacuation at 400 °C resulted in the total desorption of Py from the weaker Lewis acid sites (see the intensity drop of the bands at 1615 and 1451 cm−1), whereas the strongest Lewis acid sites still withheld Py (see the bands at 1623 and 1456 cm−1) (Fig. 5, dashed lines). Assuming that the total number of Lewis sites and the number of strong Lewis sites are proportional to the band intensities observed after evacuation at 100 °C and 400 °C, respectively, the corresponding concentrations of Lewis acid sites were determined by using the extinction coefficient given in ref. [24]. The concentrations of the weak-to-moderate strength Lewis acid sites and the strong Lewis acid sites are listed in Table2.The type of surface phosphate species was detected by 31P MAS NMR. The spectrum of the Al2O3–1P sample, having the lowest P content, shows peaks at −10 and –22 ppm (Fig.6A, b), which peaks can be assigned to phosphorous in monomeric and polymeric phosphate species, respectively [22,23,27,28]. At higher phosphorous contents, both peaks were stronger, while the peak of the polymeric species gained even more intensity (Fig. 6 A, b-d). In line with expectations [23], the higher surface phosphate concentration favors the formation of polymeric species via condensation of P–OH groups.The changing local environment of aluminum atoms on the surface of alumina upon phosphate modification was characterized by 1H and 27Al CP/MAS NMR (Fig.6B). The 1H spectra reflect the local environment of those Al atoms which are near to protons at the surface or near to surface-attached OH groups [23,29]. On the parent alumina support 27Al resonance peaks can be observed at 14, 38, and 75 ppm (Fig.6B, a), which can be attributed to octahedral (AlVI), pentagonal (AlV), and tetrahedral (AlIV) surface aluminum atoms [27–30]. Note that pentagonal aluminum atoms (AlV) are often observed in high surface area transition aluminas in minor concentrations and their presence is associated with oxygen defects adjacent to aluminum nucleus or substitution of lattice oxygen in octahedral symmetry by hydroxyl groups [30,31]. When the alumina surface was modified with increasing amount of phosphate, it was clearly visible that relatively broad component peaks developed at lower chemical shifts near to these peaks at about 54, 26, and 5 ppm (Fig. 6B, b-d) indicating the changes in the local environment of the corresponding surface Al atoms due to the formation of Al-O-P bonds [27,32]. Results show that all types of surface Al atoms were involved in the formation of Al-O-P bonds, suggesting the non-selective binding of the phosphate to the alumina surface. This observation is in line with the spectral changes found in the νOH region (Fig. 2) indicating the consumption of all the different types of OH groups upon phosphating the alumina surface.It is generally accepted that the carbonyl bands of CO adsorbed on highly dispersed Pd catalysts appear in the spectral range below and above about 2000 cm−1, attributed to bridging and linearly-bound CO, respectively [12,33,34]. The FTIR spectrum of the species formed from CO adsorption on alumina-supported Pd catalysts are shown in Fig. 7 . Carbonyl bands are clearly discernible at about 1860, 1945, 2050, and 2085 cm−1. Following the band assignments of Lear et al. [34,35], the broad band at about 1860 cm−1 can be assigned to µ3 hollow-bonded CO on Pd [111] planes or µ2 bridge-bonded CO on Pd [100] planes, whereas the band near to 1945 cm−1 can be attributed to the µ2 bridge-bonded CO on Pd [100] facets and CO, bridge bonded to particle edges. The linear CO peaks at around 2050 and 2085 cm−1 can be ascribed to CO bound to Pd [111]/[111] and Pd [111]/[100] particle edges, and CO bound to particle corners, respectively [34,35]. These bands are all present both on the non-phosphatized and phosphatized-alumina-supported Pd catalysts, although some deviation from the published relative intensities, are apparent (Fig. 7). In particular, the linear features relative to the bridge-bonded features became more pronounced at increasing phosphorous content of the support, which suggest somewhat greater contribution of the Pd particle edges and corners to the adsorption. Note that phosphating the alumina support hardly affected the metal dispersion and the Pd particle size in the catalysts (Table1).Results of catalytic hydroconversion of tricaprylin (TC) are shown in Fig. 8 . The organic liquid product contained the unreacted TC (if any) and caprylic acid, propyl caprylate, 1-octanol, octyl caprylate, heptane, octane, and some other minor products, mainly octanal, 8-pentadecanon, 9-nonanone, and dicaprylates formed by the hydrogenolysis of only one ester bond of TC. Caprylic acid and propyl caprylate is formed by the hydrogenolysis (HYS) reaction of three or two ester bonds, respectively, whereas octyl caprylate could have been formed by esterification of caprylic acid by octanol. Heptane and octane were produced via deoxygenation of caprylic acid. Octanol is a possible intermediate of paraffin formation [11]. The effluent gas contained mainly propane (from HYS reaction) and CO. Minor amounts of CO2, ethane, and methane could be also detected. The dominance of heptane over octane and CO over CO2 in the liquid and gas phase product mixture, respectively, suggests that hydrodecarbonylation (HDCO) is the main deoxygenation route, whereas hydrodecarboxylation (HDCO2) and oxygen hydrogenation (HDH2O) represent minor reaction routes [6,7].At the reaction temperature of 300 °C the conversion of TC was low (18.5%) on the Pd/Al2O3 catalyst, but reached virtually 100% on the Pd/Al2O3–2.5P and Pd/Al2O3–5P catalysts (Fig. 8, left side). When the reaction temperature was raised to 350 °C, all the phosphatized-alumina-supported Pd catalysts showed high activity in the HYS reaction resulting in 100% TC conversion (Fig. 8, right side). In line with earlier findings [11], these results suggest that caprylic acid intermediate was formed by facile HYS reaction from TC, which was followed by consecutive, rate-limiting deoxygenation (mainly HDCO) reaction of the intermediate. Interestingly, the yield of the paraffin products (heptane and octane) dramatically increases with the phosphorous content of the alumina support reaching nearly 100% on the Pd/Al2O3–5P catalysts (Fig. 8, right side). These results clearly suggest that phosphatization of alumina surface resulted in the change of catalyst structure so that the rate of the hydrodeoxygenation (mainly HDCO) reaction was significantly enhanced.The carboxylate species formed from adsorption of valeric acid, and their reactivity was investigated under catalytic conditions by quasi-operando DRIFT spectroscopy. The results obtained for the Pd/Al2O3 and Pd/Al2O3–5P catalysts are presented in Figs. 9 and 10 , respectively.Molecularly adsorbed carboxylic acid could not be observed under the conditions of experiments, i.e., the characteristic νC=O band of valeric acid expected to appear at ∼ 1780 cm−1 could not be detected. The negative bands in the νOH region (Figs. 9-10, Section A) and positive bands in the νO–C–O region (Figs. 9-10, Section B) clearly suggest that the adsorption of the carboxylic acid resulted in the consumption of surface OH groups and in the simultaneous formation of surface carboxylate species [36,37]. This surface reaction (dissociative adsorption) is described as the deprotonation of the carboxylic acid by the combination of acid hydrogen with a surface hydroxyl group to produce surface carboxylate species and H2O [36–38]. Results shown in Fig. 9A and 10A indicate that practically all types of OH groups on the support can be involved in the formation of carboxylate species, including the P–OH groups of the phosphatized support (Fig. 2). Note that phosphatization resulted in consumption of surface Al–OH groups and formation of new P–OH groups (Fig. 2). Both the remaining Al–OH and the new P–OH groups are available for the adsorption of carboxylic acid and their total number determines the surface concentration of the carboxylate groups.The position of the asymmetric and symmetric νO-C-O stretching bands appearing over and below about 1500 cm−1, respectively, in addition to the difference between their peak positions (Δν = νas-νs) are indicative of the bonding structure of carboxylate species [36,37,39,40]. The frequency of the νas vibration and the corresponding Δν value were shown to increase in the following order: chelating bidentate < bridging bidentate (≈ free ionic) < monodentate carboxyl species. The intense pair of bands observed at 1575 and 1470 cm−1 (Δν = 105 cm−1) for the Pd/Al2O3 sample (Fig. 9B) and a similar pair of bands at 1585 and 1470 cm−1 (Δν = 115 cm−1) for the Pd/Al2O3–5P sample (Fig. 10B) can be assigned to the asymmetric and symmetric vibrations of chelating bidentate carboxylate species [36,37,39,40]. A second type of carboxylate species gives an intense asymmetric νO-C-O stretching band at 1650 cm−1 (Fig. 9B). The identification of its symmetric pair is, however, difficult due to the appearance of overlapping C–H deformation vibrations in the frequency range below 1500 cm−1, most probably due to the appearance of the δas(CH3), βs(CH2), and δs(CH3) vibrations of the –CH3 and –CH2- groups of the hydrocarbon chain (the δs(CH3) vibration is clearly discernible at around 1350 cm−1) [36,40,41]. However, we found a band at about 1390 cm−1 that showed parallel of intensity with that of the 1650 cm−1 band, if reaction conditions were varied. It was substantiated that a band at about 1390 cm−1 is the pair of the 1650 cm−1 band, stemming from symmetric νO-C-O stretching vibration (Fig. 9B, 10B, and S4, S5). The relatively high frequency of the asymmetric νO-C-O vibration band (1650 cm−1) and the large frequency separation from the corresponding symmetric vibration band (Δν = 1650 – 1390 = 260 cm−1) clearly suggest that this second type of carboxylate group can be identified as monodentate carboxylate species bonding to surface aluminum atom [39,40]. Note that the concentration of the monodentate carboxylate species is much lower over the Pd/Al2O3–5P sample than over the Pd/Al2O3 sample (Fig. S4, Sections C and D). Phosphatization of γ–alumina surface eliminated mainly those sites, where monodentate carboxylate species could have been formed.Upon contacting the catalysts with He/VA mixture at 300 °C gradually carboxylate bands developed (Fig. S4) until their surface concentration reached steady state (Fig. 9B and 10B, spectra a). Virtually the same steady state concentrations were reached at 350 °C (Fig. 9B and 10B, spectra b). As expected, no reaction products were formed in the absence of H2.When the reactant flow was changed to H2/VA, the intensity of the bands at 1575–1585 cm−1 and 1470 cm−1 decreased, suggesting that mainly the concentration of the corresponding bidentate carboxylate species decreased, whereas the surface concentration of the monodentate carboxylate species (bands at 1650 and 1390 cm−1) hardly changed. The consumption of the bidentate carboxyl species was accompanied by the appearance of the main deoxygenation reaction products, such as, butane, pentane, CO and CO2. Higher reaction temperature resulted in a higher reaction rate and, therefore, in higher product concentrations in the effluent gas (Fig. 9C and 10C). The dominance of butane and CO in the product mixture suggests that the main route of VA deoxygenation is the HDCO reaction. The results suggest that the surface of the Pd/Al2O3 catalyst is covered by more reactive bidentate carboxylate species and less reactive monodentate carboxylate species. In contrast, the surface coverage of the Pd/Al2O3–5P catalyst by the more active carboxylate is substantially higher, and it is much lower by the less active species than the corresponding coverages of the Pd/Al2O3 catalyst.Product formation was accompanied by recovery of surface OH groups as indicated by the intensity drop of some negative OH bands (Fig. 9A and 10A). Because the reaction of carboxylic acid and OH groups is accompanied by release of H2O, the recovery of OH groups had to involve C–O bond hydrogenolysis. Mainly lower-frequency OH groups recovered, showing that the less basic OH groups were involved in the formation of reactive carboxylates [16,17].The effect of phosphatization on the alumina surface seems to be twofold: it consumes surface Al–OH groups (Fig. 2) and also reduces the concentration of Lewis acid sites (Table 2) and consequently the concentration of the Lewis acid (Al⊕) – Lewis base (O⊝) pair sites. Note that latter oxygen atoms can behave as Brønsted base (proton acceptor) or Lewis base (electron pair donor) sites depending on the nature of the adsorbate [42].The reaction of phosphoric acid with surface OH groups is well documented [20,21]. It is considered to be an acid-base reaction, as shown in Scheme1 A. The reaction results in the formation of surface phosphate groups and water. Phosphate species formed on adjacent OH groups can condensate via P–OH groups to form polymeric phosphate species [23], which appeared as dominating species at higher (≥2.5 wt%) phosphate loadings (Fig. 6A). It is important to note that phosphatization results not only in the appearance of P–OH groups but also in the formation of non-reactive terminal AlIV-OH groups (Fig. 2). A similar phenomenon was observed in a former study and was related to the reaction between a bridging OH group (Al–O(H)–Al) and phosphoric acid giving a terminal OH group and a [H2PO4]– ligand attached to a surface Al atom [20]; however, it was not clarified how the charge neutrality was preserved in this process.The present study evidenced that formation of new terminal AlIV–OH groups was accompanied by the elimination of strong acid Lewis sites (i.e., coordinately unsaturated tetrahedral Al sites) and their charge balancing Lewis base (oxide ion) pair sites. We rationalize these observations by the reaction shown in Scheme1B. The strong Lewis acid sites were suggested to become reversibly reconstructed by establishing a weak bond with a nearby oxide ion and thus they could appear more as a distorted tetrahedral ion than as a tri-coordinated one (see left side of Scheme1B) [16]. However, the very weak coordination bond in the strained species can be easily broken in the presence of a base or an acid [16,43]. Our results substantiate that phosphoric acid reacts with these Lewis acid (Al⊕) – Lewis base (O⊝) pair sites resulting in the formation of terminal AlIV-OH species and [H2PO4]– groups completing the coordination sphere of coordinately unsaturated tetrahedral Al sites (right side of Scheme1B). Note that latter oxide ions behave as proton acceptor Brønsted base sites in the process, whereas the [H2PO4]– groups cannot be distinguished from those formed with the involvement of surface OH groups (vide supra). The process proposed here clearly indicates how the charge neutrality can be maintained during the formation of the terminal OH groups, and also accounts for the elimination of Lewis acid (Al⊕) –Lewis base (O⊝) pair sites (Scheme1B). This is in line with the suggestion of DeCanio et al. [43] for the bonding of fluoride ions to strong Lewis acid sites upon HF treatment of alumina support.The quasi-operando DRIFTS investigation revealed the formation of two types of surface carboxylate species under catalytic conditions (Figs. 9-10). Bidentate carboxylate species were formed via acid-base reaction between fatty acid and a surface hydroxyl groups to produce stable surface carboxylate and H2O as shown in Scheme2 A [36–38]. It is most likely that the generated water desorbs from the surface at the reaction temperatures (300 and 350 °C) applied here during the DRIFTS investigations [37]. This process shows close resemblance to that observed for the surface reaction with phosphoric acid (see above and in Scheme1A). The second type of carboxylate could be identified as monodentate carboxylate group [39,40]. It is rational to think that these latter species were formed in a similar process over Lewis acid (Al⊕) – Lewis base (O⊝) pair sites than that we proposed for the surface reaction with phosphoric acid (cf. Scheme1B and 2B). The result is a monodentate carboxylate group, in which one of the oxygen forms an ester-like bond with a surface aluminum atom, while the other oxygen forms an H-bond with the neighboring OH group (right side of Scheme2B) [39,44]. The reactive adsorption of carboxylic acid on strong Lewis acid (Al⊕) – Lewis base (O⊝) pair sites is clearly supported by the good correlation between the concentration of strong Lewis acid sites (determined by Py adsorption) and the integrated absorbance values of the asymmetric νO-C-O band at 1650 cm−1, assigned to monodentate carboxylate species (Table S2).In relation to the catalytic properties, the most important consequence of phosphatization is that it reduces the concentration of Al⊕– O⊝ pair sites, where the less reactive monodentate carboxylate species are formed. The DRIFTS investigations showed that both types of carboxylates are strongly-bound surface intermediates, but the bidentate carboxylate species are more susceptible to deoxygenation reaction leading to paraffin product than the monodentate carboxylate intermediate (Figs. 9 to 10). The dominance of the bidentate carboxyl species on the phosphatized alumina support explains the substantially higher deoxygenation (mainly HDCO) activity of the phosphate catalyst as compared to the Pd/Al2O3 catalyst (Fig. 8). Note that phosphatization did not affect noticeably either the particle size, or the morphology of the Pd particles, therefore the enhanced deoxygenation activity and paraffin selectivity can be attributed to changes in the surface properties of the alumina support. Nevertheless, the reaction between the active surface intermediate and hydrogen most probably should take place at the metal/support interface where activated hydrogen is available for the reaction.We found that preferentially weak base OH groups were recovered during the HDO reaction (Fig. 9A and 10A). The recovery of surface OH groups must involve the scission of a carboxylate C–O bond. The most likely product is aldehyde (Scheme2C), which is considered as an important intermediate of the HDO reaction of carboxylic acids [2,5,8]. Note that the process depicted in Scheme 2C is the reverse process than that observed for surface carboxyl formation from aldehyde on metal oxides [36]. Our results suggest that the bidentate carboxylate species, formed in reaction with weak base hydroxyls, are more ready to react with hydrogen than the carboxylates formed in reaction with strong base surface sites. In this context, replacement of OH groups by P–OH groups favorably affects the HDO activity.Paraffin formation proceeds mainly via hydrodecarbonylation (HDCO) of the aldehyde intermediate, whereas oxygen reduction reaction (HDH2O) of aldehyde represents a minor reaction route. The paraffin chain of fatty acids is shortened in HDCO but it is preserved in HDH2O reaction [2,5,7]. This latter reaction was suggested to proceed via primary alcohol intermediate and requires hydrogenation, dehydration, and/or hydrogenolysis steps [2,8]. One can speculate that latter reaction goes on a less complex reaction pathway in which the cleavage of both C–O bonds in the surface carboxylate species occurs, however, the justification of such reaction route still requires further investigation.This study provides insights into the structure – activity relationships, determining the triglyceride HDO activity of alumina-supported Pd catalysts. The Pd/γ-A2O3 catalyst showed relatively good activity in the ester bond hydrogenolysis of triglyceride, but poor activity in the consecutive deoxygenation of the obtained carboxylic acid intermediate to paraffin product.Surface modification of the γ-alumina support with phosphate resulted in surprisingly high activity enhancement of the catalyst in the later rate-determining step of paraffin formation. The phosphatization of the alumina surface resulted in (i) partial elimination of basic Al–OH groups and the concomitant formation of weak acid P–OH groups on the alumina surface, (ii) a decrease in the number of Lewis acid (Al⊕) – Lewis base (O⊝) pair sites, whereas (iii) it did not affect noticeably either the particle size, or the morphology of the Pd particles on the support.Quasi-operando DRIFTS investigation under catalytic conditions revealed that both surface Al–OH and P–OH groups serve as sites for fatty acids to form bidentate carboxylate groups, which can further react with hydrogen to form paraffin products. If the OH groups, involved in the carboxylate formation are less basic, the formed carboxylate groups are more ready to react with hydrogen.Carboxylic acids can also react with strong Lewis acid (Al⊕) – Lewis base (O⊝) pair sites giving monodentate carboxylate species, having low reactivity. Since phosphatization significantly reduced the number of these pair sites, the concentration of the less reactive monodentate carboxylate groups decreased substantially. The dominance of reactive bidentate carboxylate groups on the phosphatized alumina surface explains the substantially higher HDO, mainly hydrodecarbonylation (HDCO) activity of the phosphatized-alumina-supported Pd catalyst than the HDO activity of the Pd/γ–Al2O3 catalyst.The product formation from the bidentate carboxylate surface intermediate is accompanied by OH group recovery, suggesting that the deoxygenation reaction must start with the hydrogenolysis of a carboxylate C–O bond. This reaction leads to the formation of aldehyde that must be the intermediate of deoxygenation to paraffin.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 of the project by the Economic Development and Innovation Operative Program of Hungary, GINOP-2.3.2-15-2016-00053: Development of liquid fuels having high hydrogen content in the molecule (contribution to sustainable mobility). The Project is supported by the European Union. Thanks are also due to the Interreg V-A Slovakia - Hungary Cooperation Program, SKHU/1902, (Project No: SKHU/1902/4.1/001) and to the Project (Project No. VEKOP-2.3.2-16-2017-00013) supported by the European Union and the State of Hungary, co-financed by the European Regional Development Fund. This work was also supported by the TKP2020-IKA-07 project financed under the 2020-4.1.1-TKP2020 Thematic Excellence Programme by the National Research, Development and Innovation Fund of Hungary.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcat.2021.08.052.The following are the Supplementary data to this article: Supplementary data 1

The mechanism of catalytic hydrodeoxygenation (HDO) of fats, vegetable oils, and fatty acids was studied using alumina-supported Pd catalysts and tricaprylin and valeric acid as model reactants. The chemistry of fatty acid/catalyst interaction was studied by quasi-operando Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS). The Pd/γ–Al2O3 catalyst showed good activity in the hydrogenolysis reaction of the ester bonds to convert tricaprylin to caprylic acid, but they were of poor activity in the consecutive hydrodeoxygenation (HDO) of the acid to paraffin. The surface modification of the support alumina by phosphate groups significantly increased the HDO activity of the Pd catalyst and, consequently, the paraffin yield. The activity change was accounted partly for the partial replacement of the weak base Al–OH groups by weak acid P–OH groups but mainly for the partial elimination of Lewis acid (Al⊕) – Lewis base (O⊝) pair sites on the surface of the support. Both surface Al–OH and P–OH groups were shown to participate in the reaction with carboxylic acid and formed bidentate surface carboxylate species, which further reacted with hydrogen to give paraffin. Carboxylates of less basic surface sites were found to be more prone to HDO reaction than those of strong base sites. Monodentate carboxylates, formed on Al⊕ O⊝ pair sites were of low reactivity. Phosphatizing eliminated most of the Lewis type acid-base pair sites, therefore, reactive bidentate carboxylates represented the most abundant surface intermediate (MASI) during the HDO reaction of triglyceride. The hydroxyl coverage of the carboxylated surface was shown to become somewhat higher under steady-state reaction conditions. The increased hydroxyl coverage implies that C–O bond hydrogenolysis of the surface carboxylate proceeds, regenerating OH groups and forming aldehyde that could be intermediate of paraffin formation.

The emergence of antidepressants has effectively addressed symptoms for depression patients in recent years. Nefazodone (Nefa)–a novel third generation antidepressant agent and phenyl piperazine derivative–has been widely used in the treatment of depression. Several studies have reported the recent abuse of Nefa, leading to its release into natural water courses, which can cause a serious impact on the antipredator behavior or alter the reproductive behavior of aquatic species [1–3]. However, there are few reports on the degradation of Nefa. It is therefore essential to develop useful treatments to eliminate Nefa in aqueous media.Photocatalysis technology has recently received significant attention because it is environmentally friendly, low cost, and convenient, and has been shown to effectively remove antibiotics and many halogenated organic pollutants [4–7]. The selection of suitable photocatalysts for superefficient photocatalytic degradation of pollutants under visible light irradiation is therefore important. Graphitic carbon nitride (g-C3N4) has been shown to be highly effective for removing organic contaminants and offers benefits such as a tunable bandgap, low toxicity, and easy modification [8‒11]. Doping modification is one of the most effective strategies for further enhancing the photocatalytic activity of g-C3N4 [12,13]. Among the metal-free dopants, phosphorus (P) plays an important role in modifying the structure of g-C3N4, which reduces the band gap energy, increases the separation efficiency of photogenerated charge carriers, and enhances the optical response [14]. For example, P can be directly doped into g-C3N4 frameworks to give a coral-like porous tube morphology [15]. In addition, transition metal phosphides can serve as cocatalysts to improve the activity and stability of g-C3N4 by constructing heterojunctions [16,17]. However, the roles of P-doping in g-C3N4–based photocatalysts do not have been adequately studied.A previous study, in which a co-doped g-C3N4–based photocatalyst for tetrachlorobisphenol A removal was prepared using P source calcination and transition metal layered double hydroxides (Co and Ni LDHs) as soft-templates, focusing on the effect of P atoms on the electron structure of Co and Ni transition metals [18]. However, the role of P doping in Co and Ni-loaded carbon nitrides and the effects of different P-doping strategies were not discussed in depth. In this study, four different photocatalysts are prepared to investigate the role of P doping in Co and Ni-loaded carbon nitride photocatalysts. The morphology and structure of the as-prepared samples are characterized. The optical performance is tested, including the optical absorption and photogenerated charge carrier transfer. The photocatalytic activity of the samples is evaluated in terms of Nefa degradation under visible light irradiation. Corresponding degradation mechanisms and routes are proposed based on analysis of the active species and intermediate products.Prior to synthesis of the composites, CN and CoNi LDH were prepared in accordance with previously reported methods [18]. Four different photocatalysts were then synthesized using a P annealing treatment with NaH2PO2 (1 ​g) as a P source (PH3) and denoted PA-1, PA-2, PA-3, and PA-4. For PA-1, the PH3-treated CN (0.2 ​g) and PH3-treated CoNi LDH (0.01 ​g) were mixed by ultrasonication and calcined at 350 ​°C for 2 ​h (5 ​°C/min) under a N2 atmosphere in a tube furnace. After cooling to room temperature, PA-1 was washed with ultrapure water for several times and dried in an oven at 60 ​°C. PA-2 and PA-3 were prepared using a similar process, but pristine CN and CoNi LDH were used as precursors, respectively. For PA-4, pristine CN (0.2 ​g) and CoNi LDH (0.01 ​g) were mixed and placed in a tube furnace. NaH2PO2 was then introduced at the upstream side and the calcination conditions were the same as that for PA-1. A schematic diagram of the synthetic methods is shown in Fig. 1 .The morphology, structure, and optical performance were characterized by scanning/transmission electron microscopy (SEM/TEM), aberration-corrected high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM), X-ray photoelectron spectroscopy (XPS), electron spin-resonance spectroscopy (ESR), UV-vis diffuse reflectance spectroscopy (DRS), room-temperature photoluminescence emission spectroscopy (PL), and electrochemistry. The concentration and degradation products of Nefa were analyzed by liquid chromatography and high-resolution mass spectrometry. The details of the characterization are shown in the Supplementary Information.A 100-mL solution containing Nefa (10 ​mg/L) and photocatalyst (20 ​mg) was added to a quartz reactor. Prior to the photocatalytic experiments, the quartz reactor was placed in a dark environment for 30 ​min to reach saturation adsorption. The light source for photocatalytic degradation was a 300 ​W Xenon lamp (PLS-SXE300, Beijing Perfectlight) with a 420-nm cut-off filter. A given volume of solution was sampled at regular intervals, and the samples were measured following passage through a 0.22-μm syringe filter.Different calcination strategies were used to determine the role of P doping in Co and Ni-loaded carbon nitride photocatalysts. The SEM images (Fig. S1) show that all samples had the inherent porous morphology of CN, which provided a large surface area. PA-1 and PA-3 exhibited clear agglomeration due to π-π stacking, indicating that the introduced PH3 may destroy the inherent porous morphology of CN and result in the agglomeration. TEM images of pure CN and the samples are shown in Fig. S2 and Fig. 2 a–d. For PA-1, PA-2, and PA-3, a lattice spacing of 0.21 ​nm related to the metal phosphides (CoNiP x ) was observed [19,20], while PA-4 exhibited bright spots attributed to Co and Ni single atoms based on the findings of a previous study [18]. The observations indicate that different calcination strategies could change the local environment of single-atom catalysts. In addition, the EDS elemental mapping images in Fig. 2e show that C, N, Co, Ni, and P were well distributed in PA-4.The structure of the samples was analyzed by XPS. The N/C of CN, PA-1, PA-2, PA-3, and PA-4 was determined to be 1.23, 1.20, 1.12, 1.13, and 1.16, respectively. The reduced ratios for the annealed samples compared with pure CN indicate that the direct contact between CN and the P source led to the formation of more nitrogen defects, resulting in improved optical performance. The formation of nitrogen defects in PA-4 also provided a favorable environment for anchoring single atoms [21]. Three conventional peaks located at 284.8 ​eV (adventitious carbon), 286.4 ​eV (C–N), and 288.2 ​eV (N–CN) were observed in the C1s XPS spectra (Fig. 3 a). The N1s XPS spectra show that all samples exhibited three peaks at 398.6, 399.6, and 401.6 ​eV corresponding to bi-coordinated (N2C), tri-coordinated (N3C), and NH x groups in the heptazine framework, respectively (Fig. 3b) [22,23]. The spectra of the PH3-treated CN (P–CN) control are shown in Fig. S3. The component ratios in P–CN show similar values to that of CN, implying a slight influence of P treatment on the pure CN. Compared with the ratios of N–CN (39.2%) and N2C (27.9%) in the CN, the ratios in all the P-doped samples were reduced, indicating that the nitrogen defects are due to the cracking of N–CN structure. In the P2p XPS spectra (Fig. 3c), all the P-doped samples showed two peaks at 129.0 and 133.8 ​eV related to Metal–P and P–N/C, respectively [24–26]. The greater intensities of Metal–P and P–N/C for PA-4 imply that more P-doping structures were formed by the concurrent P annealing process in the presence of CN and CoNi LDH. The loading amounts of Co and Ni in the P-doped samples are listed in Table S1. For PA-1 and PA-2, metal phosphides were first prepared using the CoNi LDH soft-templates by effective P-annealing reactions, which led to no available single atom precursors. For PA-3, the introduced P first reacted with carbon nitride to give P–CN, which changed the coordination environment of the single atoms. In PA-4, the introduced P was able to concurrently react with CoNi LDH and carbon nitride, allowing the etching of the CoNi LDH soft-templates, while also doping into the coordination structure to form single atoms.The presence of nitrogen defects was also revealed by ESR spectroscopy. The g value at 2.004 is attributed to N vacancies (Fig. 3d) [27]. The ESR intensity of all samples, PA-1, PA-2, PA-3, and PA-4, increased markedly compared with that of CN. PA-4 generated more N vacancies than the other samples owing to its synthesis process, which is consistent with the observed N/C ratios. For PA-4, the CN and CoNi LDH precursors were first mixed and then subjected to the PH3 annealing process, which retained the porous structure of CN, and provided more P sources to react with the precursors and generate more active sites.The optical absorption characteristics of the as-fabricated samples were measured by DRS. Fig. 4 a shows that the absorption edge of CN was 542 ​nm. Compared with CN, the P-doped samples exhibit red-shifts (>550 ​nm), and the absorption intensity was boosted in both UV and visible light regions, indicating that the addition of P significantly promoted the optical absorption capacity [28,29]. PL spectra of the as-prepared samples were acquired to determine the separation of photogenerated carriers. Fig. 4b shows that all P-doped samples exhibited a lower PL signal than that of the pure CN, suggesting that the recombination of electron-hole pairs was impeded by the introduction of P [30]. As anticipated, PA-4 had the highest charge separation efficiency, indicating that the formation of single atoms might make the electron cloud shift to the metal active sites and inhibit the recombination. The transient photocurrent response and electrochemical impedance spectra (EIS) were also determined. Fig. 4c shows that PA-4 exhibited the best photocurrent response intensity–approximately 4 times to that of CN. The arc radius of the P-doped samples was much smaller than that of CN (Fig. 4d). Among the samples, PA-4 showed the smallest radius related to the lowest charge transfer resistance, which might be due to the formation of N vacancies and more P-doping induced single atom sites.The photocatalytic activity of the as-fabricated samples was determined by analyzing the removal of Nefa from water under visible light irradiation. Fig. 5 a shows that Nefa was not eliminated without photocatalysts under visible light conditions. All P-doped samples exhibited an improved catalytic performance compared with CN. PA-4 showed the highest activity among all samples. Its degradation efficiency reached 99.9% within 40 ​min, which is faster than that of the pure CN (50%). For PA-1, PA-2, and PA-3, the enhanced activity might be resulted from the construction of heterojunctions from P-doped CN and CoNiP x [31,32]. In PA-4, the reaction of P with CoNi LDH and CN generated single atoms, which altered the structure of the carbon nitride framework and accelerated the charge carrier transfer [33]. Additionally, the corresponding kinetics were well matched with the pseudo-first-order model, and the k values of CN, PA-1, PA-2, PA-3, and PA-4 were found to be 0.021, 0.032, 0.049, 0.046, and 0.065 ​min−1, respectively (Fig. 5b). As shown in Fig. 5c, the degradation rate of Nefa decreased by 2.73% after five experiment cycles, indicating that PA-4 exhibited good catalytic stability.To determine the photocatalytic degradation mechanism of PA-4 for Nefa, the main active species were investigated: superoxide radicals (O2 •−), electrons (e−), hydroxyl radicals (•OH), and holes (h+). ρ-Benzoquinone (BQ), AgNO3, tertbutyl alcohol (TBA), and sodium oxalate (SO) were used as scavengers for O2 •−, e−, •OH, and h+, respectively [34]. As shown in Fig. 5d, the degradation of Nefa was completely suppressed in the presence of BQ, indicating that O2 •− plays a major role in the photocatalytic process. Similarly, the degradation efficiency was negatively affected by the addition of AgNO3, suggesting that e− also plays an essential role. The addition of TBA and SO slightly limited the degradation efficiency, implying that •OH and h+ ​play minor roles. The ESR spectra also indicated the presence of the active species mentioned above (Fig. 6 ).Analysis of the intermediate products (Fig. S4) allowed possible Nefa degradation routes over PA-4 under visible light irradiation to be proposed (Fig. 7 ). P1 and P2 are believed to be generated following hydroxylation reactions. These intermediate products are then thought to be broken into smaller units such as ρ-hydroxy-(m-chlorophenyl) piperazine (P3), 1-(m-chlorophenyl) piperazine (P4), m-chloroaniline (P5), and piperazine (P6) by N-dealkylation reactions [35,36]. Another route is related to the removal of chlorobenzene from the Nefa structure and hydroxylation reaction. The piperazine structure is then disrupted to generate P9 and P10. All the intermediate products are then further oxidized and finally generate CO2, H2O, and Cl− among other products.Four P-doped Co and Ni-loaded carbon nitride photocatalysts were prepared using different P-doping strategies. Characterization results showed that all the P-doped samples presented N defect structures, which enhanced their optical absorption performances and inhibited photogenerated charge carrier recombination compared with pure CN. Among the P-doped samples, PA-4–prepared using a concurrent P annealing process in the presence of CN and CoNi LDH–presented single atom structure, which effectively improved its photogenerated charge carrier transfer. Additional structure characterizations indicated that the introduced P in PA-4 was able to etch the soft-templates and get doped into the coordination environment to form Co and Ni single atom photocatalysts, which inhibited photogenerated charge carrier recombination and improved the photocatalytic activity. All the active species, O2 •−, •OH, and photogenerated electrons and holes, contributed to the degradation of Nefa. PA-4 exhibited a catalytic removal of 99.9% for Nefa degradation within 40 ​min under visible light irradiation, which was faster than that of pure CN (50%).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. 52100076). The authors thank Suqian Ningbiao Technology Testing Co., Ltd. for the characterization support.Information associated with this article can be found in the Supplementary material.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.efmat.2022.05.001.

To understand the role of phosphorus doping in Co and Ni-loaded carbon nitride photocatalysts, four P-doped samples are prepared using different strategies. Morphological characterization shows that Co and Ni single atoms were prepared using a concurrent P annealing process in the presence of carbon nitride and CoNi layered double hydroxides (PA-4). In addition, structural characterization indicates that the introduced P can etch CoNi soft-templates and be doped into the coordination environment. The PA-4 structure is believed to enhance the photogenerated charge carrier transfer. The as-prepared PA-4 samples exhibit better photocatalytic activity for nefazodone (Nefa) degradation in water (99.9% within 40 ​min) than other P-doped samples. Quenching experiments indicate that O2 •−, •OH, and photogenerated electrons and holes contribute to the degradation of Nefa. Analysis of the intermediate products suggests that the degradation routes primarily involve hydroxylation reactions, N-dealkylation reactions, and piperazine cracking. The findings provide an alternative strategy for the preparation of P-doped Co and Ni-loaded carbon nitride photocatalysts for contaminant degradation and elucidate the role of P doping.

Direct N-acylation of amines through the use of carboxylic acid, most particularly, acetic acid as an acylating agent to form amide linkage is a challenging tasks in synthetic chemistry [1]. This is interesting though not surprising because the amide functionality is an essential key part of many biologically as well as industrially important organics, such as peptides [2] polymers [3] natural products, and pharmaceuticals [4]. Moreover, N-acylation is also an important synthetic target in organic chemistry for the protection of amino groups during the multistep transformations [5].The most popular route of N-acylation or amide synthesis involves the use of activated carboxylic acid derivatives, such as acid chlorides, anhydrides, and esters [6]. For example, Goodreid et al. [7] reported direct amidation reaction between metal carboxylate salt of the corresponding acid and amine. Though this protocol gives a good quantitative yield (61%) within 2 h, it involves the tedious step of carboxylate salt formation by carboxylation of lithiated terminal alkyne. The cross-coupling reaction between aryl esters and aniline to form amide linkage is also reported [8]. This again is a tedious two-step protocol, where carboxylic acids require conversion into the corresponding esters. The direct amidation between aryl chloride or bromide and amines has also been reported [9]. In this reaction protocol, activation of halides is required to be carried out by expensive palladium-based organometallic catalyst. Thus, all these conventional substrates have serious limitations, such as corrosion effects, high cost, hygroscopicity, tedious workup, and low atom efficiency [10]. The one-step (direct) reaction between amines with carboxylic acids would therefore be both economically and environmentally a benign route to amide synthesis.In addition to homogeneous catalysts, a large number of heterogeneous catalysts are used for amide formation reaction. In the direct amidation of indoles by using ZnCl2 catalyst, expensive electrophilic reagent N-[(benzenesulfonyl)oxy] amides are used for the selective 3-amidation of indole [11]. Here, a series of primary amides were prepared by amidation reaction between twenty-five different carboxylic acids and urea by using ZrOCl2·8H2O and cerium ammonium nitrate as heterogeneous catalysts under solvent-free condition [12]. The problems with the aforementioned two protocols are the need for microwave irradiation, and reproducibility. Another class of heterogeneous Lewis acid catalysts for direct amidation reaction are metal triflates, but they are difficult to synthesize, and therefore of high cost. The synthesis of N-(pyridine-2-yl) amides by a reaction between aldehydes and 2-aminopyridines was carried out under mild reaction conditions by Cu(OTf)3 catalyst [13]. A long chain peptide and protein linkages were degraded into amide fractions by Sc(OTf)3 as heterogeneous catalyst [14]. Tris (methoxyphenyl) bismuthanes was used to activate primary carboxylic acids, and later on coupled with a series of amines and alcohols to yield the corresponding amides and esters, respectively [15]. The protonated zeolite Y (Zeolite HY) catalyst was used by several workers for direct amidation reaction [5,16,17]. In this protocol, the structural modification of zeolite catalyst is an essential step to obtain amides in good yield. Direct amidation co-catalyzed by Ag/Al2O3 and Cs2CO3 between alcohols and amines yielded thirteen different amides, but this method operates only in the presence of harmful toluene solvent under refluxing condition [18]. Though the above-discussed materials are ingenious and effective catalysts for amide synthesis, all possess some practical constraints, which includes high cost, toxicity, corrosive nature, difficulty in separation, non-recyclability, harsh reaction conditions, polluting behavior, and low thermal stability [19].Bare metal oxides, such as Nb2O5, ZnO, Al2O3, TiO2, etc., have also been used as heterogeneous catalysts in direct amidation reaction. Metal oxide catalysts are the most suitable green, reasonable catalysts for direct amidation. The zinc oxide in the form of nanofluid was used as a pseudo-homogeneous catalyst for the direct amidation between aliphatic carboxylic acids and primary amines under solvent-free conditions. A tedious protocol is needed in this reaction to prepare a new reaction media, and the catalyst is recovered in the form of ZnO nanoparticles, which again needed to be converted into nanofluid ZnO for further reaction cycles [20]. Lewis acidic Nb2O5, as a basic group tolerant heterogeneous catalyst is used for the direct amide synthesis. But this reaction protocol followed continuous azeotropic refluxing of the reaction mixture with toluene for a longer time (30 h) [21]. Other metal oxides with high Lewis acidity, like Al2O3 [22,23] and TiO2 [24] were also found to be efficient for direct amidation between non-activated carboxylic acids and amines. However, most of them show low tolerance towards basic functionalities present either on acid or on the amine substrates and they also utilize hazardous solvents in the reaction protocol.For further enhancement of catalytic activity, the metal oxides were combined with other organic or inorganic compounds to form binary or ternary mixed metal oxide systems. For example, the thiol modified binary Pt/TiO2 was used as a heterogeneous catalyst for the selective hydrogenation of nitroarenes to the corresponding amines [25]. Vitamin B12-TiO2 was used as an efficient oxygen controlled catalyst for the preparation of esters and amides from trichlorinated organic compounds just by irradiating the reaction mixture with light [26].The introduction of Brønsted acidity along with Lewis acidity in the metal oxide heterogeneous catalyst proved to be a promising practice for researchers in the intervening years. Such Brønsted acidity was introduced by anchoring acidic groups, like sulfonic [27] and phosphoric [28] groups onto the surface of the catalyst. Synthesis of amides from fatty acids as well as benzoic acid derivatives was carried by using nanosulfated TiO2 as a solid acid catalyst [6]. The direct amidation between carboxylic acids and amines was carried out with sulfated tungstate as a green solid acid catalyst [10]. The acid-functionalized TiO2-based binary nanocomposites with improved Lewis as well as Brønsted acidities having synergetic effect of constituent metal oxides have been effective in various organic transformations. For example, sulfated Fe2O3/TiO2 nanocomposite was used as an efficient visible active photocatalyst [29]. The ultrasensitive sulfated graphene/TiO2 nanocomposite was studied by some workers for the detection of global antioxidant capacity [30]. Ryoo et al. reported the catecholic chelation of the benzene disulfonate with titanium ion of the binary SiO2/TiO2 composite [19]. The anchoring of two –SO3H groups on the robust network of the composite surface plays a crucial role in the acylation of ethanol by using acetic acid at ambient temperature of 80 °C. The esterification of bio-based organic acid, like levulinic acid, with ethanol in the presence of sulfated TiO2 and sulfated ZrO2/TiO2 composite was studied [31]. The report well established that the sulfated ZrO2/TiO2 composite with a large number of sulfonic groups anchored on its surface showed high catalytic activity towards the esterification of levulinic acid as compared to the bare sulfated TiO2 catalyst. Such promising catalytic performance of sulfated TiO2 based mixed metal oxide nanocomposites is encouraging research on such materials.Herein, we report sulfated binary TiO2/SnO2 nanocomposite as a green heterogeneous catalyst for the direct amide formation reaction between various amines and acetic acid. The overall reaction protocol involves no use of any coupling agent or solvent.All chemicals purchased were from S D Fine-Chem Limited, Mumbai, India and of AR grade.The bare TiO2 and bare SnO2 NPs were prepared as per our previously reported synthetic protocol [32].In our reported work [32] TiO2/SnO2 NC (with 4:1 wt% of TiO2 and SnO2 respectively) showed higher catalytic activity as compared to its single and binary counterparts. In the present investigation, this catalyst is further acid-functionalized to enhance its catalytic activity. A typical synthesis of TiO2/SnO2 NC (with 4:1 wt% of TiO2 and SnO2, respectively) was carried out by the wet impregnation method. During the synthesis of SnO2 NPs, the amount of TiO2 NPs to the desired stoichiometric ratio was added, and the resulting mixture was stirred vigorously at 60 °C for 3 h. Finally, all the samples were calcined at 450 °C for 3 h, to yield white colored NCs.For sulfation, 2 g of TiO2/SnO2 NCs was taken into a three-necked round bottom flask kept in an ice bath. Then, 2 mL of Chlorosulfonic acid was added dropwise into the flask by using a pressure-equalizing funnel fitted on the middle neck of the flask. The other two openings of the flask were packed by corks in order to avoid contact with atmospheric moisture. The resulting mixture was stirred for 2 h to assure proper anchoring of the sulfonic groups onto the surface of TiO2/SnO2 NCs. Then the mixture was removed from the flask and washed thoroughly with deionized water several times to remove unreacted chlorosulfonic acid. The white colored product was dried at 100 °C on hot plate and calcined at 300 °C in a furnace for 3 h.The crystallographic information of sulfated TiO2/SnO2 NCs was derived by X-ray diffractometry (XRD) with Ni-filtered Cu Kα radiation of 1.54056 Å (X’ pert PRO, Philips, Eindhoven). High-resolution transmission electron microscopy (HRTEM) with selected area electron diffraction (SAED) imagery was obtained (JEOL-3010 and Tecnai G2 F20). The chemical composition of the samples was determined by X-ray photoelectron spectroscopy (XPS, VG Multilab 2000, Thermo VG Scientific, UK) with a monochromatic Mg-K (1253.6 eV) radiation source. The UV–visible DRS spectra of samples was recorded by UV–visible spectrophotometry (LabIndia 3092). Ammonia-Temperature programmed desorption (NH3-TPD) analysis was used to elucidate the total acidity of sulfated sample (Micromeritics, AutoChem II 2920 chemisorption analyzer, USA, equipped with thermal conductivity detector (TCD)). BET analysis (N2 adsorption and desorption isotherms) was carried out (Quantachrome Nova Win instrument). 1H NMR and 13C NMR spectra of the compounds were recorded by Bruker AC-300 spectrometry using tetramethylsilane as an internal standard. The IR spectral analysis of the compounds was done by Perkin-Elmer FT-IR 783 spectrophotometer.The direct amidation reactions were performed in 25 mL round bottom flask under reflux at 115 °C and constant stirring (250 rpm). Equimolar volumes of glacial acetic acid (1.0 mL, 17 mM) and the corresponding amine (17 mM) were taken into the flask. Then, 50 mg of catalyst was added into the mixture. The progress of the reaction was evaluated by thin layer chromatography. The formation of the product was confirmed through FT-IR, 1H NMR, and 13C NMR measurements. After completion of the reaction, 10 mL of ethyl acetate was added into the reaction mixture. The mixture was centrifuged to remove the catalyst and the filtrate was washed with NaHCO3 (3 × 10 mL), and finally with deionized water to obtain the product. Fig. 1 shows the powder X-ray diffractograms of pure TiO2/SnO2 NC and sulfated TiO2/SnO2 NC (with 4:1 wt% of TiO2 and SnO2, respectively). In both the diffractograms, the reflections of TiO2 (2θ = (25.22, 48.44, 54.27, 55.49) °, etc.) were well-matched with the anatase TiO2 reflections (JCPDS #21-1272). Meanwhile, the reflections of SnO2 (2θ = (26.56, 34.12, 52.03)°, etc.) were well-matched with the tetragonal cassiterite SnO2 reflections (JCPDS #41-1445). The overlapping of representative peaks of TiO2 and SnO2 reveals the formation of the well-intermixed composite. The XRD pattern of sulfated NC appearing like the non-sulfated NC confirms the uniform dispersion of sulfonic groups onto the surface of the sulfated NC [33]. A slight decrease in the intensities of characteristic peaks was observed in the sulfated NC due to the surface-adsorbed sulfonic (-SO3H) groups which is in line with earlier reports [34].The presence of surface-adsorbed –SO3H groups onto the surface of NCs was ascertained by FT-IR analysis (Fig. 2 ). The FT-IR spectrum of non-sulfated NC (Fig. 2 a)) exhibited broadband and sharp band at (3,400 and 1,632) cm−1, which were attributed to the stretching and bending vibrations of surface adsorbed –OH groups of water molecules, respectively [32]. These bands are also present in the FT-IR spectrum of sulfated NC (Fig. 2 b)). The characteristic Ti–O bending vibration band at 1,400 cm−1 and Ti–O stretching vibration band in the range (400 to 900) cm−1 are evidenced in both non-sulfated as well as sulfated NCs, which is in line with the previous reports [32]. The Sn–O stretching bands in the range (750 to 455) cm−1 for SnO2 were mixed and embedded in the broad peak (M−O stretching vibration band region, where, M = Ti or Sn) [32]. The –SO3H groups anchored onto the surface of NC, confirmed by the existence of the four characteristic bands in the range (1232 to 956) cm−1 (Fig. 2 b)) [35].The morphology and particle size of the sulfated TiO2/SnO2 NC were determined by using TEM, HRTEM, and SEAD patterns. The bright-field TEM image (Fig. 3 a)) of the sample shows a close aggregation of particles having an average size of between (10 and 25) nm, which is in good agreement with the crystallite size values obtained by XRD analysis. The HRTEM image (Fig. 3 b)) clearly shows intense lattice fringes corresponding to the anatase TiO2 with ‘d’ value 0.352 nm for (101) planes and cassiterite SnO2 with ‘d’ value 0.352 nm for (110) planes; which further supports the good polycrystallinity of the sulfated TiO2/SnO2 NC [32]. The SAED pattern of the sample (Fig. 3 c)) focused on some more structural details. In this pattern, the observed diffraction rings are indexed and are perfectly matched with anatase TiO2 and cassiterite SnO2 [32]. The visibly sharp reflections of rings in the pattern clearly confirm the high polycrystallinity of the sulfated TiO2/SnO2 NCs. Fig. 4 shows the elemental composition and oxidation states of various elements in the sample that were investigated by using XPS technique.The survey spectrum represented in Fig. 4 a shows the presence of all expected elements with their characteristic peak positions. Fig. 4 b shows the XPS high-resolution spectrum of oxygen (O 1s). In the spectrum, the main peak at binding energy 530.8 eV is deconvoluted into two shoulder peaks, which indicates that the oxygen is present in several chemical states according to the measured binding energy. The lower binding energy peak at 530.8 eV is attributed to oxygen bonded to the metal –M–O (M = Ti, Sn), while the peak at 531.9 eV is attributed to adsorbed hydroxyl species (–O–H) [36]. The Ti 2p spectrum (Fig. 4 c)) reveals spin–orbit splitting of Ti 2p1/2 and Ti 2p3/2 core level states at the positions (465.3 and 459.5) eV, respectively. The spin–orbit splitting difference is 5.8 eV, which clearly confirms that the Ti element in the sample possesses + 4 oxidation state that is contributed by TiO2 [36]. In the high-resolution XPS spectrum of tin (Sn 3d) (Fig. 4 d)), the two symmetric peaks appearing at binding energies (487.3 and 495.8) eV are assigned to the lattice SnO2. The spin–orbit splitting difference between the Sn 3d5/2 and Sn 3d3/2 levels (8.5 eV) matches the standard spectrum of Sn (+4 oxidation state) bonded with O in SnO2 lattice [37]. Fig. 4 e shows the high resolution XPS spectrum for S 2p. The single broad peak situated at the position 169.5 eV corresponds to S element in the sample having + 6 oxidation state, which is contributed by the surface adsorbed sulfonic groups [29].The acidic strengths of non-sulfated TiO2/SnO2 and sulfated TiO2/SnO2 NCs were studied by NH3–TPD (Fig. 5 ) technique in which 10% NH3–He was used as a probe molecule to elucidate the total acidity of catalysts.NH3–TPD analysis of the non-sulfated TiO2/SnO2 NCs shows the presence of two desorption peak maxima at 231.9 and 313.2 °C. Similarly, sulfated TiO2/SnO2 NCs shows two desorption peak maxima at slightly higher temperatures at (245.8 and 340.8) °C. In sulfated TiO2/SnO2 NCs, a broad desorption peak area in the medium temperature range (200 to 400) °C is assigned to NH3 adsorbed on acid sites having medium strength [38]. Table 1 shows that for non-sulfated and sulfated TiO2/SnO2 NCs. The total acidity values obtained are (0.17405 and 0.21951) mmol/g, respectively († Electronic supplementary information, ESI). Thus, it can be concluded that the sulfated NC is more acidic than non-sulfated NC.The surface area and pore structure of the non-sulfated and sulfated TiO2/SnO2 NCs were measured by using the nitrogen adsorption–desorption isotherms and pore-size distribution obtained by BET analysis as shown in Fig. 6 a and b, respectively. According to the Brunauer–Deming–Deming–Teller (BDDT) classification, both the isotherms correspond to type IV isotherms, typically signifying the mesoporous nature of the prepared samples. The observed hysteresis loop situated in between medium relative pressure values (P/P0 = (0.4 to 0.8)), suggested the mesoporous nature of the samples [39]. For non-sulfated TiO2/SnO2 NC (Fig. 6 a)), the pore size distribution shows pores with the average size of 5.62 nm, and the specific surface area 71.04 m2.g−1. In the case of sulfated TiO2/SnO2 NC (Fig. 6 b)), the pore-size distribution shows pores with an average size of 5.52 nm, which is in good agreement with the nature of type IV isotherm. The BET analysis also revealed that the specific surface area of sulfated TiO2/SnO2 NC is 61.684 m2.g−1, which is higher than its bare counterparts, and comparable with that of the non-sulfated TiO2/SnO2 NC. A slight decrease in the surface area of sulfated NC in comparison to the non-sulfated NC was attributed to the surface adsorbed –SO3H groups in the case of sulfated NC. A significantly high surface area of sulfated NC could result in high catalytic activity during amide formation reaction.As a primary objective, the sulfated TiO2/SnO2 NC was tested as a heterogeneous catalyst for the direct amide formation reaction. In this direction, various experimental conditions required for direct amidation were optimized (see Figs. S1 and S2 of the † Electronic supplementary information, ESI). In a typical procedure, glacial acetic acid was used as an acylating reagent as well as a solvent for the reaction. In a 25 mL RB flask, glacial acetic acid (1.0 mL, 17 mmol) and aniline (Table 2 , entry 1) (1.6 mL, 17 mmol) were taken, and 50 mg of prepared catalyst was added with constant stirring at the temperature of 115 °C. The progress of the reaction was monitored by thin-layer chromatography. The reaction is completed within 2 h with a 95% yield. It is confirmed that the product is not formed in the absence of a catalyst. Scheme 1 shows the representative amide formation reaction.The well-optimized reaction protocol used for the direct amidation of aniline was also used for the amidation of various derivatives of aniline with electron-donating and electron-withdrawing groups to establish the general applicability of the sulfated TiO2/SnO2 NC as a heterogeneous catalyst for the direct amidation. Table 2 shows that equimolar amounts of aniline derivative and acetic acid were treated with each other; excellent yields of (65 to 97) % of the corresponding amides were observed in (2 to 6) h under solvent-free condition. Moreover, to compare catalytic selectivity and activity, the turnover number (TON) and turnover frequency (TOF) of reactions were calculated and are given in Table 2. The TON and TOF values clearly show that the electron-donating groups (–CH3, –OCH3) favor amidation reaction, while in contrast, the electron-withdrawing groups (–NO2, –Cl) harm amidation reaction.After the usual reaction, all the products obtained were pure (confirmed by spectral data and melting points), and did not require additional efforts to purify them. The products were characterized by 1HNMR, 13CNMR, and IR spectroscopic techniques. All the spectroscopic data obtained were found to be identical with the literature data of known compounds († ESI). N-phenylacetamide (1a). White lustrous solid; m. p. = 113 – 114 °C; 1H NMR (300 MHz, CDCl3) δ 2.15 (3H, s, –CH3), 7.10–7.12 (1H, d, Ar-H), 7.27–7.32 (2H, t, Ar-H), 7.52–7.55 (2H, d, Ar-H), 8.77 (1H, s, –NH-); 13C NMR (75.47 MHz, CDCl3) δ 24.33, 120.36, 124.29, 128.88, 138.16, 169.47; IR cm−1 (KBr) N-(-2-methylphenyl) acetamide (1b). Colorless solid; m. p. = 110 – 111 °C; 1H NMR (300 MHz, CDCl3) δ 2.20 (3H, s, –CH3), 2.26 (3H, s, Ar-CH3), 7.07–7.11 (1H, m, Ar-H), 7.18–7.23 (1H, m, Ar-H), 7.28 (1H, m, Ar-H), 7.71–7.74 (1H, m, Ar-H); 13C NMR (75.47 MHz, CDCl3) δ 17.78, 24.22, 123.62, 125.39, 126.71, 129.54, 130.47, 135.64, 168.44; IR cm−1 (KBr) N-(-4-methylphenyl) acetamide (1c). White solid; m. p. = 149 – 150 °C; 1H NMR (300 MHz, CDCl3) δ 2.15 (3H, s, –CH3), 2.32 (3H, s, Ar-CH3), 7.10–7.28 (1H, d, Ar-H), 7.46–7.49 (2H, d, Ar-H), 9.79 (1H, s, –NH); 13C NMR (75.47 MHz, CDCl3) δ 20.85, 24.45, 120.12, 129.44, 133.92, 135.37, 168.46; IR cm−1 (KBr) N-(-4-methoxylphenyl) acetamide (1d). Colourless solid; m. p. = 128 – 130 °C; 1H NMR (300 MHz, DMSO) δ 2.00 (3H, s, –CH3), 3.70 (3H, s, -O-CH3), 6.84–6.87 (2H, d, Ar-H), 7.37–7.58 (1H, m, Ar-H); 13C NMR (75.47 MHz, DMSO) δ 24.21, 55.56, 114.22, 121.02, 132.95, 155.48, 168.24; IR cm−1 (KBr) N-(-3-nitrophenyl) acetamide (1e). Yellow solid; m. p. = 150 – 152 °C; 1H NMR (300 MHz, DMSO) δ 2.08 (3H, s, –CH3), 7.53–7.58 (1H, m, Ar-H), 7.82–7.88 (2H, m, Ar-H), 8.59 (1H, m, Ar-H), 10.42–10.44 (1H, d, –NH); 13C NMR (75.47 MHz, DMSO) δ 24.42, 113.43, 117.96, 125.31, 130.49, 140.76, 148.35, 169.60, 169.98; IR cm−1 (KBr) N-(-4-nitrophenyl) acetamide (1f). Yellow solid; m. p. = 215 – 216 °C; 1H NMR (300 MHz, DMSO) δ 2.10 (3H, s, –CH3), 6.57–6.72 (1H, m, Ar-H), 7.78–7.94 (2H, m, Ar-H), 8.17–8.20 (2H, m, Ar-H), 10.56 (1H, s, –NH); 13C NMR (75.47 MHz, DMSO) δ 24.64, 112.83, 118.99, 125.37, 126.81, 136.10, 142.44, 145.86, 156.12, 169.84; IR cm−1 (KBr) N-(-4-chlorophenyl) acetamide (1g). Colourless solid; m. p. = 177 – 179 °C; 1H NMR (300 MHz, DMSO) δ 2.03 (3H, s, –CH3), 7.31–7.33 (2H, m, Ar-H), 7.58–7.61 (2H, m, Ar-H), 10.08 (1H, s, –NH); 13C NMR (75.47 MHz, DMSO) δ 24.40, 120.95, 126.99, 128.98, 138.69, 168.93; IR cm−1 (KBr)As part of the systematic study, we also explored the reusability of the catalyst as an important criterion of any heterogeneous catalyst. After each cycle of amidation reaction, the solid catalyst was removed by ultrafiltration and washed thoroughly with deionized water. Then the catalyst was dried in oven to remove adsorbed organic moieties (if any) as well as water at 110 °C. The equimolar amounts (17 mM) of aniline and acetic acid were further treated with each other in the presence of a dried catalyst for the next cycle of amidation reaction. Fig. 7 shows that the catalyst retained its activity even after four successive cycles. Thus, it has been established that the catalyst has good stability, and therefore can be efficiently recycled and reused for repeated cycles of amidation reaction with an appreciable recovery of product yield.In summary, we have prepared sulfated TiO2/SnO2 NC by using chlorosulfonic acid as a sulfating agent, and investigated various physicochemical properties of the as-prepared NC using different instrumental techniques. This NC was then explored as a green heterogeneous catalyst for direct amidation reaction between diverse primary amines and acetic acid. The sulfated TiO2/SnO2 NC catalyzed amidation protocol offers a number of advantages, such as recyclability of the catalyst without loss in its activity, easy work-up, large-scale availability of catalyst, appreciable-to-high product yields of (65 to 97) % in short reaction time of (2 to 6) h, solvent-free condition, and easy separation of catalyst from the reaction mixture by filtration. All of these important outcomes of the study contribute to making this process more advanced, economical, and green, in terms of the environmental aspects. S.M. Patil: Data curation, supervision, writing - reveiw and editing. S.A. Vanalakar: Data curation, supervision, writing - reveiw and editing. S.A. Sankpal: Visualization, Investigation. S.P. Deshmukh: Writing - review & editing. S.D. Delekar: 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.SMP is thankful to the University Grants Commission, New Delhi, India for financial assistance under the award of UGC-FIP [F. No. 36-40/14 (WRO)], which is gratefully acknowledged.Supplementary data to this article can be found online at https://doi.org/10.1016/j.rechem.2021.100102.The following are the Supplementary data to this article: Supplementary data 1

Heterogeneous binary TiO2/SnO2 nanocomposite (with 4:1 wt% of TiO2 and SnO2, respectively) catalyst was prepared by a sol–gel method, and further sulfated by chlorosulfonic acid. X-ray diffraction technique revealed the structure of the nanocrystalline catalyst to be tetragonal. Fourier transformed infrared technique elucidated the presence of surface-anchored sulfonic (–SO3H) groups on the catalyst. Morphological details of the catalyst were obtained by transmission electron microscopy. Elemental analysis of the catalyst was carried out by X-ray photoelectron spectroscopy. NH3-TPD technique was used to elucidate the surface acidity of the catalyst. The active surface area and mesoporosity of catalyst were studied by the BET method. Thereafter, this sulfated catalyst was utilized for the direct amidation reaction between a series of amine derivatives and acetic acid. The reaction gives excellent product yield within 120 min, and at a relatively moderate temperature of ~115 °C.

As an important class of chemicals, carbonate has many applications in chemical industry such as organic synthesis, printing and dyeing, electrochemistry and petrochemical sectors [1–4]. Diphenyl carbonate (DPC) is one of the versatile carbonates, and is often used as a key monomer for the green production of polycarbonates (PC) to replace the hazardous phosgene. It is also widely employed in the synthesis of various organic compounds and used as starting materials for the fine chemical and pharmaceutical industries like plastic plasticizer, solvent carrier, medicine and agriculture, and among others [5–8]. At present, there are several routes to produce DPC, mainly including phosgene method, oxidative carbonylation and transesterification methods [9–12]. Among these routes, using toxic phosgene as the raw materials in the process has a hazardous threat to the environment and human health, and such a route needs a large amount of NaOH to neutralize the produced inorganic acid, resulting in a great deal of waste salt and waste water. For oxidative carbonylation route, noble metal Pd catalysts have to be used in the synthetic process that suffers from the high cost of the production. Fortunately, the transesterification route is considered to be green and economic for the synthesis of DPC (Scheme 1 ), because the employed raw material is no toxicity, no pollution, inexpensive, and readily available. Besides, only CH3OH is the by-product in this process.Generally, this reaction goes through two-step process [13,14]: (i) methyl phenyl carbonate (MPC) of the intermediate is produced by the transesterification of DMC and phenol (Scheme 1-(1)), and (ii) the further disproportionation of MPC (Scheme 1-(2)). Nevertheless, there are several problems for this reaction. Firstly, DMC is the raw material and is also the product, which makes this reaction slower. Secondly, the equilibrium constants K1 (6.3 × 10−5, 25 °C) and K2 (0.19, 25 °C) are very small in the absence of catalyst even at elevated temperature, leading to the very low yield of DPC. Therefore, the key to these problems is to develop efficient catalytic system. In the past years, various catalytic systems and reaction technologies have been reported for this process, but most of them have been focused on the first-step transesterification process of DMC and phenol since it is the rate-controlling step. Actually, to improve the total yield of DPC and the reaction efficiency, the second-step disproportionation of MPC is also very important. In recent years, we have devoted to the catalytic researches of MPC disproportionation by using titanium/tin organic compounds and metal/mixed metal oxides [15–19], but the activity and selectivity are not high enough, and high reaction temperature and long reaction time have to be required. In addition, these catalysts all contain transition metals as active centers which are highly toxic and environmentally harmful, and affect the purity of DPC as well. Metal-free catalysts would be an attractive alternative for replacing metal-containing catalysts, but no work has been reported in this area. Consequently, it is great of significance to design and develop metal-free catalysts toward clean DPC synthesis under facile operation condition.In recent years, ionic liquids have attracted growing attention in the application of solvents and/or catalysts for chemistry and chemical industry due to their green and tunable properties [20–22]. Thus, ionic liquids may be considered to be a novel metal-free catalyst for the synthesis of DPC. However, their utilization in homogenous catalysis brings disadvantages of using ionic liquid in large quantities and difficulty in their separation after reaction. Importantly, ionic liquids can be chemically grafted on solid materials to share the excellent catalytic activity of homogeneous catalyst and the easy-separation feature of heterogeneous catalyst [23–25]. Under these circumstances, the selection of solid supports is of critical importance. Mesoporous silica materials, especially SBA-15, have plenty of mesopores, high thermal stability and high specific surface area, which enable them to be good candidates for the grafting of ionic liquids.In this work, we designed a series of ionic liquids with different functional active groups as metal-free catalysts, which were immobilized on SBA-15 by chemical bonding grafting. These catalysts were then used for the synthesis of DPC (Fig. 1 ), and the effect of reaction parameters as well as the recyclability of the catalysts was studied in detail. It was found that [SBA-15-IL-OH]Br displayed outstanding catalytic performance, and MPC conversion of 80.5% and DPC selectivity of >99% could be achieved with a low catalyst loading under mild reaction conditions. The TOF value was thrice higher than that by using previously reported traditional transition metal-based catalysts. Moreover, the surface properties of the catalysts were thoroughly analyzed by elemental analysis, FT-IR, XRD, TG, SEM-EDS, TEM and BET techniques to explore the relationship between structure and performance.Firstly, 1-(3-(triethoxysilyl)propyl)-1H-imidazole was synthesized according to the modified literature procedures [26,27], and the synthesis process was shown in Scheme 2 . In a typical procedure, imidazole (13.6 g) (3-chloropropyl)triethoxysilane (48.2 g) and toluene (400 mL) were added into a round-bottomed flask under nitrogen atmosphere and then the mixture was stirred under refluxed condition for 24 h. The reaction was monitored through TLC. Then, triethylamine (20.2 g) was dropwise added to the solution and the resulting mixture was refluxed for another 2 h. After completion of the reaction, the mixture was filtrated off and washed three times with toluene (3 × 20 mL). Then the solvent was evaporated at reduced pressure and the obtained product was purified through column chromatography with neutral alumina. Finally, the resulting solution was removed by rotary evaporation to give a colorless viscous liquid (Compound 1). The structure of Compound 1 was confirmed by 1H NMR, 13C NMR and FT-IR (Figs. S1–S2, Supplementary Information).In the second step, SBA-15 material was synthesized by following a previously reported procedure with some modifications [28,29]. For this purpose, triblock copolymer P123 (5.0 g) was dissolved in 2 mol L−1 of aqueous HCl (180 mL) until transparent solution was formed. Then, TEOS (12.0 g) was slowly added into the solution under stirring. The mixture was kept at room temperature for 24 h under vigorous stirring, and then aged in a Teflon-lined autoclave at 100 °C for 24 h. The as-made sample was recovered by filtration and washed extensively with ethanol and deionized water, and dried overnight in an oven at 80 °C. Finally, the obtained white solid was calcined at 500 °C for 5 h in air under static conditions.For the synthesis of compound 2, 1.0 g of SBA-15 was dispersed in 50 mL of dry toluene by ultrasonic stirring for 30 min under nitrogen atmosphere, and then 2.0 g of Compound 1 (see Scheme 2) in 5 mL of dry toluene was added, and the reaction mixture was stirred and refluxed at 90 °C for 24 h through the condensation reaction between the surface silanol groups (Si–OH) of SBA-15 and the ethoxyl groups (–OCH2CH3) of organotriethoxysilanes [30,31]. After cooled down to room temperature, the reaction mixture was filtered and washed three times with absolute ethanol, and then the obtained white solid was Soxhlet extracted with dichloromethane. The hybrid SBA-15 was recovered and dried at 60 °C under vacuum for 24 h (Compound 2, see Scheme 2). The structure of Compound 2 was verified by FT-IR, SEM-EDS, 29Si MAS NMR and 13C MAS NMR measurements (Figs. S3–S5 and Table S1).The synthesis of ionic liquid-SBA-15 hybrid catalysts with different functional groups including –OH, –COOH, –CH3, –SO3H and –NH2 was illustrated in Scheme 3 . The OH-functionalized ionic liquid hybrid SBA-15 was displayed in Scheme 3-(1). Briefly, compound 2 (1.0 g) was suspended in 50 mL of dry toluene to form a uniform dispersion by sonication in a round-bottomed flask. Then, 1.02 g of 3-bromo-1-propanol was added to the solution and the reaction mixture was stirred for 24 h under reflux conditions. After reaction, the mixture was cooled down to room temperature, the produced solid was filtered and washed with toluene repeatedly, followed by drying at 60 °C under vacuum overnight to afford OH-functionalized ionic liquid hybrid SBA-15. For convenience, the obtained catalyst was denoted as [SBA-15-IL-OH]Br. Besides, the ionic liquids with –OH group and different alkyl chain lengths on the imidazolium cations were also prepared (Scheme S1), and the synthetic procedures were similar to that of [SBA-15-IL-OH]Br. The only difference was that the corresponding material was used to replace 3-bromo-1-propanol, which was designated as [SBA-15-IL-xC-OH]Br where xC is the carbon atom number in alkyl chain of the IL (x = 2, 4, 6 and 8).Based on the similar procedure except that the starting material was 4-bromobutyric acid, the COOH-functionalized ionic liquid hybrid SBA-15 shown in Scheme 3-(2) was prepared. The obtained catalyst was denoted as [SBA-15-IL-COOH]Br. For the preparation of CH3-functionalized ionic liquid hybrid SBA-15, SO3H-functionalized ionic liquid hybrid SBA-15 and NH2-functionalized ionic liquid hybrid SBA-15 shown in Scheme 3-3,3-4 and 3-(5), the procedures were presented in Supplementary Information (Section S.2-S.4). The obtained catalyst was denoted as [SBA-15-IL-CH3]Br [SBA-15-IL-SO3H]Br and [SBA-15-IL-NH2]Br, respectively.In addition, OH-functionalized ionic liquid hybrid SBA-15 with BF4 −, PF6 −, HSO4 − and OH− anions were also prepared for comparison according to the procedures reported in literatures [32,33], and the procedures were shown in Scheme S2. The obtained catalysts were designated as [SBA-15-IL-OH]BF4 [SBA-15-IL-OH]PF6 [SBA-15-IL-OH]HSO4 and [SBA-15-IL-OH]OH, respectively.Powder X-ray diffraction patterns (XRD) were recorded on a DX-2700B diffractometer with the Ni-filtered CuKα radiation (1.5418 Å). FT-IR spectra were acquired by Thermo Nicolet 380 spectrometer using KBr pellet technique. Scanning electron microscopy (SEM) and Energy dispersive X-ray (EDS) analysis were performed with a JSM-7500F instrument. Transmission electron microscopy (TEM) measurements were carried out on JEM-1011 apparatus with a field-emission gun operating at 200 kV. N2 adsorption–desorption profiles were determined on a 3H–2000PS2 adsorption instrument. Element analysis was performed on an Elementar Vario EL cube (EA) to determine the chemical composition of the samples. Thermal gravimetric (TG) analysis was performed using a Netzsch Thermoanalyzer STA 449C analyzer under N2 atmosphere. 1H NMR and 13C NMR spectra were recorded on a Bruker 600 MHz spectrometer. 13C MAS NMR and 29Si MAS NMR measurements were performed on a Bruker Advance III 400 MHz NMR spectrometer.MPC disproportionation was performed in a three-necked, round-bottomed flask under magnetic agitation. In a typical experiment, 150 mmol of MPC and catalyst (0.4–1.2 g) was added into the three-neck flask under N2 atmosphere. Afterwards, the reaction mixture was heated at the desired temperature. During the reaction progress, DMC product was distilled out by a liquid dividing head attached to a receiver flask to break the equilibrium limit thus shifting the reaction towards the production of target product DPC. Upon completion, the reactor was cooled down to room temperature and the catalyst was filtered off from the mixed solution, washed with acetone and then dried under vacuum. The filtrate was detected by GC–MS HP-6890/5973 instrument with HP-5 capillary chromatography packed column. The products were quantitatively analyzed with 7890A gas chromatograph equipped with flame ionization detector (FID) and a DB-35 capillary chromatography packed column.The FT-IR spectra of the samples were showed in Fig. 2 to examine the structure of catalysts. As shown in Fig. 2(a), for parent SBA-15, the intensive absorption peaks at 3442 and 1635 cm−1 were ascribed to the stretching and bending vibrations of surface O–H groups, and the typical peaks at around 1099, 806, and 458 cm−1 were attributed to the asymmetric stretching, symmetric stretching and bending modes of Si–O–Si condensed silica network [34], respectively. Besides, the peak at 971 cm−1 was associated with the bending vibration of framework Si–OH group [35]. Clearly, the FT-IR spectra of functional ionic liquids hybrid SBA-15 in Fig. 2(b-f) exhibited the respective SBA-15 characteristic peaks, but the peak at 971 cm−1 weakened and even disappeared, which indicates that the bonding interactions of the surface Si–OH of SBA-15 with –OCH2CH3 of the functionalized ionic liquids took place during the preparation process, resulting in the removal of the Si–OH. At the same time, some new peaks were observed obviously, for instance, two new characteristic peaks at 3013–2829 cm−1 could be seen corresponding to the CH2 asymmetric and symmetric stretching vibrations of the propyl chain of ionic liquids, and they all exhibited four typical peaks at around 1596–1395 cm−1 corresponding to the stretching vibrations of imidazolium ring [36–38]. These results suggest that functionalized ionic liquids with different functional groups were successfully grafted onto the surface of SBA-15 through chemical bonding.The XRD patterns of the samples were showed in Fig. 3 . It can be seen from Fig. 3(A), the small-angle patterns for both of the parent SBA-15 and ionic liquids hybrid SBA-15 exhibited three remarkable diffraction peaks in low angle region. The intense diffraction peak at 0.99° was assigned to the diffraction of the (100) plane of the mesoporous structure with a remarkable long-range ordering degree, and the other two weak peaks at 1.59 and 1.82° corresponding to (110) and (200) reflections were attributed to the two-dimensional hexagonal planes of the mesoporous structure [32,39,40]. However, compared to the parent SBA-15, the diffraction peak (100) for ionic liquids hybrid SBA-15 only became broader and weaker slightly, which was probably resulted from the occupation of the porous channel and the partial blockage of small pores by ionic liquids. This phenomena was consistent with the unit cell parameter calculated by 2d100/(3)1/2 in Table 1 , where the unit cell volume a0 increased from 10 to 10.2. Fig. 3(B) displayed the wide-angle XRD patterns of the samples. For parent SBA-15, the strong and broad peak at 23° was evidently observed, which was attributed to the amorphous nature of mesoporous silica, and the characteristic peaks of ionic liquids hybrid SBA-15 in the wide-angle ranges were similar to that of parent SBA-15. As a result, it could be inferred that the ordered mesoporous structure of SBA-15 remained perfect after grafting of the functionalized ionic liquids, which was further evidenced by TEM analysis below.Structural properties of the samples were evaluated by N2 adsorption–desorption technique. As shown in Fig. 4 , both of the SBA-15 and the hybrid SBA-15 exhibited type IV isotherms with clear H1 hysteresis loops, which is a characteristic feature of the typical mesoporous silica material according to the IUPAC classification scheme [41,42]. The increase of steep step in adsorption volume for parent SBA-15 in the ranges of 0.65–0.92 P/Po corresponded to the characteristic of capillary condensation in the mesopore, however as for hybrid SBA-15, they appeared in the range of 0.57–0.72 P/Po. The pore size distributions in Fig. S6 displayed that the parent SBA-15 exhibited homogeneous distribution with average pore diameter of 6.6 nm. However, average diameter of the hybrid SBA-15 by ionic liquids decreased to around 5.7 nm. At the same time, it was noted that BET surface area and pore volume of the samples were declined evidently after hybridization of ionic liquids (Table 1), suggesting that the grafting of ionic liquids was in the pores, and the pore wall thickness was increased as compared with that of the parent SBA-15. These results suggested that the functionalized ionic liquids were mainly grafted inside the pore channels of SBA-15.The surface morphology and microstructure of the samples were investigated by SEM-EDS and TEM techniques. As shown in Fig. 5 (a), the parent SBA-15 exhibited long rod-like morphology and smooth surface with uniform and homogeneous particle size of around 1–3 μm. After hybridization of the functionalized ionic liquids, the surface morphology of SBA-15 with regular appearance in Fig. 5(b) and Fig. S7 was not obviously changed, but the particles were better dispersed, suggesting that the hybrid SBA-15 had good structure integrity and morphology. The EDS spectra of the ionic liquid hybrid SBA-15 were shown in Fig. S8, and the existence of the corresponding elements in their structure further confirmed that functionalized ionic liquids were well bonded to SBA-15, consistent with the FT-IR analysis.To investigate the pore structure of these catalysts, TEM was used to probe the mesopore architecture of the particles. As shown in Fig. 5(c), the parent SBA-15 displayed a two-dimensional hexagonal network and long-range mesopore architecture with parallel and perpendicular directions, and the pore diameter and wall thickness were 6.8 nm and 3.5 nm, respectively. After hybridization, the pore diameter decreased and the wall thickness increased, but the pore channels were not blocked (Fig. 5(d) and Fig. S9). Moreover, the values of wall thickness and pore diameter observed by TEM were in good agreement with those from XRD and BET investigation (Table 1). It may also be noted that the hexagonal order channels of SBA-15 were still kept well after hybridization.TG analysis was employed to investigate the thermal stability of the samples, and the TG curves were displayed in Fig. S10. For the parent SBA-15, the weight loss below 100 °C could be assigned to the desorption of adsorbed moisture, and the weight loss over 100 °C was negligible. As we can see, the ionic liquids hybrid SBA-15 could endure the temperature of about 250 °C, the weight loss in the range from 250 to 450 °C was related to the removal of ionic liquid components from the silica surface, suggesting that these catalysts could be used for the reactions investigated in the present work owing to their high thermal stability. From the TG analysis, the ionic liquid contents were calculated as 0.68, 0.67, 0.64, 0.62 and 0.59 mmol g−1 for SBA-15-IL-CH3, SBA-15-IL-OH, SBA-15-IL-COOH, SBA-15-IL-SO3H and SBA-15-IL-NH2, respectively, which agreed with the results of element analysis shown in Table S2.The ionic liquids-SBA-15 hybrid catalysts with different functional groups were used for the synthesis of DPC, and the results of catalytic reaction were given in Table 2 . It was shown that SBA-15 support was almost inactive for the synthesis of DPC due to the absence of active sites (Table 2, entry 1). However, hybridization of the functionalized ionic liquids on SBA-15 displayed excellent catalytic activity in real catalytic conditions (Table 2, entries 2–6), and the major by-products were anisole and phenol. The order of the activity was as follows [SBA-15-IL-SO3H]Br > [SBA-15-IL-NH2]Br > [SBA-15-IL-COOH]Br > [SBA-15-IL-OH]Br > [SBA-15-IL-CH3]Br, which was presumedly resulted from the influence of different active groups [43,44].Among these catalysts [SBA-15-IL-COOH]Br and [SBA-15-IL-NH2]Br showed high catalytic activities with more than 81% MPC conversion (Table 2, entries 4 and 5), but gave the lower DPC selectivity (about 90%). In this case, high by-product yields were ascribed to the decarboxylation of MPC and the methylation of MPC by DMC into anisole and phenol catalyzed by strong acidic or basic sites on the functional groups, because ionic liquids with these groups were strong active sites in some catalytic reactions [39,45,46]. Therefore [SBA-15-IL-SO3H]Br with stronger acidic –SO3H group exhibited much higher MPC conversion (Table 2, entry 6). On the contrary [SBA-15-IL-CH3]Br revealed the higher DPC selectivity but a relatively lower MPC conversion due to the fact that –CH3 group was weak acid site (Table 2, entry 2). It is interesting to note that [SBA-15-IL-OH]Br with moderately acidic strength of –OH group showed the best catalytic performance regarding both the conversion and selectivity, giving 80.5% MPC conversion and 99.6% DPC selectivity (Table 2, entry 3). Compared with the other catalysts investigated here, the high catalytic activity of [SBA-15-IL-OH]Br may be attributed to the large surface area, superior pore size and abundant active sites [26,47,48]. Furthermore, the effect of the alkyl chain length of the cation with the same functional group and the counterpart anion was studied (Table 2, entries 7–10), revealing the decrease activity with the increase of the chain length, which suggested that the alkyl chain played an important role in the catalytic activities as reported in literatures for other reactions [49–52]. Meanwhile [SBA-15-IL-2C–OH]Br using 50 mg was tested in the reaction, and 4.6% DPC yield was obtained, indicating that the reaction was carried out in no stoichiometry regime. In addition, the effect of anionic structures of the ionic liquids was also examined for [SBA-15-IL-OH]Br catalyst, and BF4 −, PF6 −, HSO4 − and OH− were chosen for this purpose. It was found that the anionic structures had a strong impact on the catalytic activities (Table 2, entry 3 and entries 11–14), and at the same time the high yields of by-products were also obtained. The catalytic activity of these anions decreased in the order: HSO4 − > OH− > Br− > BF4 − > PF6 −. By contrast, SBA-15-IL-OH with Br− anion showed the best catalytic performances (Table 2, entry 3), and was thought to be more suitable for the target catalytic reaction.The grafted content of ionic liquids had great influence on the yield of DPC, and the appropriate content could ensure that the catalyst had more active sites. Therefore, the effect of the grafted content of ionic liquid on the catalytic activity of [SBA-15-IL-OH]Br was examined in no stoichiometry and the result was displayed in Fig. 6 (a). It could be seen that the conversion of MPC increased from 40.1% to 80.6% as the ionic liquid loading increased from 4% to 12%, while the selectivity was not dependent on the loading content. When the loading content was above 12%, the conversion reduced gradually. In this case, the specific surface area was greatly declined and the pore channels of SBA-15 was also partially blocked as confirmed by BET measurements (Fig. S11 and Table S3). Thus, the appropriate loading content was 12%.The effects of different parameters such as reaction temperature, catalyst dosage and reaction time were also investigated by using [SBA-15-IL-OH]Br as an example. As shown in Fig. 6(b), the catalytic activity was sensitive to the variation of the reaction temperature. The conversion of MPC increased from 40.3% to 81.1% as the reaction temperature increased from 150 °C to 170 °C and the selectivity of DPC was almost 99% because the reaction was an endothermic process. However, the selectivity dropped to 91.9% when the reaction temperature was further raised to 190 °C, some by-products such as anisole and phenol were produced at such a high reaction temperature due to the decomposition of MPC decarboxylation and the methylation of DMC, which was similar to the results reported previously in literatures [14,17]. Thus, the suitable reaction temperature was 170 °C for this reaction system.The effect of catalyst dosage on the reaction was illustrated in Fig. 6(c), increasing the catalyst dosage accelerated the conversion and [SBA-15-IL-OH]Br exhibited increased catalytic activity. When the catalyst usage increased from 0.4 to 0.8 g, the MPC conversion increased from 34.3% to 80.6% with a DPC selectivity of 99.5%, which could be attributed to the increase of the catalytic active sites. When catalyst usage was further reached to 1.2 g, the conversion increased to 81.6% but the selectivity was significantly reduced. This was plausibly correlated with the mass transfer between the catalyst and the reactants. When the catalyst was excessive, the heat transfer resistance could lead to more by-products, thus the selectivity decreased. Therefore, the optimal catalyst amount was 0.8 g.The change of MPC conversion and DPC selectivity with reaction time was shown in Fig. 6(d). The MPC conversion increased from 36.2% to 80.4% as the reaction time increased from 1 h to 2 h, and the selectivity was almost 99.5%. Continue to extend the reaction time to 3.5 h, the conversion only increased to 82.6%, but the selectivity decreased a little. Thus, this result illustrated that the reaction could be nearly completed and reached the equilibrium within 2 h.Based on the above results, the optimal condition could be summarized as follows: a loading of 12% ionic liquid, reaction temperature of 170 °C, catalyst amount of 0.8 g and reaction time of 2 h. Under these optimal conditions, conversion of 80.5% for MPC and yield of 80.2% for DPC were achieved. Furthermore, the catalytic performances of SBA-15-IL-OH and other heterogeneous catalysts reported in the literatures were compared in Table 3 . It can be seen that [SBA-15-IL-OH]Br may work at relatively mild reaction condition, and exhibited a good catalytic performance. And the TOF value up to 174 h−1 was obtained, which is three times higher than the highest value (58 h−1) reported in literatures by using traditional transition metal-based catalysts. Thus, it is worth noting that [SBA-15-IL-OH]Br is an efficient catalyst to avoid the use of metal-based chemicals.The hot filtration test is conducted to examine the heterogeneity of the ionic liquid hybrid SBA-15. This test was carried out in the presence of 0.5 g [SBA-15-IL-OH]Br at 170 °C for 1 h. Subsequently, the catalyst was removed by filtration and the reaction was continued for another 4 h in the absence of the catalyst. It was found that the yield of the product DPC in the filtrate was not changed (Fig. S12), indicating that [SBA-15-IL-OH]Br was a stable heterogeneous catalyst.It is well known that the recyclability of the catalyst plays an essential role in practical application. In order to examine the reusability of [SBA-15-IL-OH]Br, reusable experiments were performed under the same reaction conditions. After completion of each reaction, the catalyst may be readily separated by filtration and the recovered catalyst was washed with acetone, dried under vacuum at 60 °C for 12 h, and then used for the next run. As shown in Fig. 7 , the catalyst could be reused for at least six times without obvious decrease of activity, and the conversion and selectivity were remained 77.6% and 99.4%, respectively. Besides, after the sixth cycle, the recovered catalyst was subjected to XRD, FT-IR and TEM measurements. It is noted from Figs. S13 and S14 that [SBA-15-IL-OH]Br still exhibited the consistent characteristic peaks after six cycles. Moreover, the morphology and pore channel of the recovered catalyst had no significant change compared with the fresh one (Fig. S15). These results further verify that the catalyst possesses remarkable stability and reusability in this reaction, indicating the potential in the future application.Based on the results obtained in this work and reported in the literatures [47,48,51], a possible mechanism is proposed for the process over [SBA-15-IL-OH]Br, as shown in Scheme 4 . First, the carbonyl group of MPC is activated by the hydrogen bonding interactions between C2–H of the imidazolium cation and carbonyl group of the MPC (step 1). At the same time, the H atom of OH of the imidazolium cation coordinates with the O atom of carbonyl group of the MPC through a hydrogen bond interaction, which cooperatively activates the reactant molecules. Subsequently, the O atom of methoxyl group of another MPC molecule attacks the activated carbonyl group by nucleophilic addition reaction, thereby forming an interior ring carbonate complex (step 2). Finally, the carbonate complex is opened up to yield the desired DPC via an intramolecular nucleophilic substitution and catalyst regeneration to complete the catalytic cycle (step 3). In the catalysis process, there are more than one active site in the catalyst, they can active multiple MPC molecules simultaneously by hydrogen bonding interactions, which make the catalysis more efficient, thereby improved the catalytic activity and yield of DPC.DFT calculations were also performed to understand the possible mechanism of the reaction. The geometries of reactants, intermediates and products were fully optimized without any restriction at the theoretical level of M06–2X/6-311g (d,p) [47,51]. Frequency calculations were further carried out to confirm that these stationary points are real minimums on potential energy surfaces. The result of calculation was shown in Fig. S16. The hydrogen bond of O--H1 (between C2–H of the imidazolium cation and carbonyl group of the MPC) was formed with a bond length of 3.446 Å (Intermediate I) after complexation of [SBA-15-IL-OH]Br and MPC, and simultaneously the hydrogen bond between O--H2 (between H atom of OH of the imidazolium cation and the O atom of carbonyl group of the MPC) was formed with a bond length of 2.097 Å, which displayed a stronger interaction with MPC than O--H1. The Gibbs free energy of intermediate I to the reactants is −21.9 kcal mol−1, which indicates a spontaneous process. This suggests that MPC can be activated cooperatively by imidazolium C2–H and hydroxyl. Subsequently, the C--O bond was formed with a bond length of 3.090 Å by intra-molecular nucleophilic substitution (Intermediate II), and the bond length of the O--H1 was elongated to 5.064 Å in this process that made the cleavage of carbonate complex much easier, thus accelerating the reaction. The Gibbs free energy of intermediate II is slightly higher than intermediate I, which could be remedied by the energy released from the C–O cleavage in the reactant. These results demonstrate that the hydrogen bond interactions between [SBA-15-IL-OH]Br and MPC indeed played an important role in the promotion of the reaction investigated here.In summary, ionic liquids hybrid SBA-15 with different functional groups were prepared and used as metal-free catalysts for the effective and solvent-free synthesis of DPC. It was found that [SBA-15-IL-OH]Br with –OH group showed the best catalytic performance with 80.5% MPC conversion and 99.6% DPC selectivity. Notably, the TOF value up to 174 h−1 was obtained, which was three times higher than the best value reported in literatures by using traditional metal-containing catalysts. This catalytic system could be easily isolated and recovered from the reaction mixture by filtration, and reused for six recycling without remarkable loss of catalytic activity. The findings reported here provides a new platform based on ionic liquids-SBA-15 hybrid materials as green metal-free catalysts to accomplish the highly efficient synthesis of DPC in the absence of solvent, revealing a remarkable potential for application in the future.There are no conflicts to declare.We gratefully acknowledge the financial support from the National Natural Science Foundation of China (No. 21808048 and U1704251), Training Plan for University's Young Backbone Teachers of Henan Province (2021GGJS121), Program for Science & Technology Innovation Talents in Universities of Henan Province (23HASTIT014), Postgraduate Education Reform and Quality Improvement Project of Henan Province (YJS2022KC22), Project funded by China Postdoctoral Science Foundation (No. 2018M632782), Project funded by Postdoctoral Research Grant in Henan Province (No. 001802030), Key Project of Science and Technology Program of Henan Province (No. 222102230109, 212102310330 and 182102210050), and the Science Research Start-up Fund of Henan Institute of Science and Technology (No. 2015031).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.2021.02.010.

Diphenyl carbonate (DPC) is one of the versatile carbonates, and is often used for the production of polycarbonates. In recent years, the catalytic synthesis of DPC has become an important topic but the development of a highly active metal-free catalyst is a great challenge. Herein, a series of ionic liquids-SBA-15 hybrid catalysts with different functional groups have been developed for the synthesis of DPC under solvent-free condition, which are effective and clean instead of the metal-containing catalysts. It is found that in the presence of [SBA-15-IL-OH]Br catalyst, methyl phenyl carbonate (MPC) conversion of 80.5% along with 99.6% DPC selectivity is achieved, the TOF value is thrice higher than the best value reported by using transition metal-based catalysts. Moreover, the catalyst displays remarkable stability and recyclability. This work provides a new idea to design and prepare eco-friendly catalysts in a broad range of applications for the green synthesis of carbonates.

The progressive climate change is a globally relevant issue, and accordingly to the Intergovernmental Panel on Climate Change (IPCC) reports, humanity has so far caused a rise in global temperature of about 1°C. To avoid drastic adverse effects on biodiversity, the melting of ice caps, and further rise of sea levels, a value of 1.5°C related to the pre-industrial level should not be exceeded (Abram et al., 2019). Regarding the potent greenhouse gas CO2, the average global atmospheric concentration exceeded 400 ppm in 2016, which is the highest level ever recorded. In addition, the worldwide fossil fuel emissions of CO2 increased by more than 2% in 2018 (Abram et al., 2019). Therefore, strategies are needed that prevent further increase of the CO2 concentration in our atmosphere. The options discussed in this context are the capturing of carbon dioxide as well as its direct conversion. Along this line, the electrochemical reduction of CO2 via appropriate catalysts into value-added products is a promising strategy. The obtainable products include C1 compounds, such as CO, CH4, HCOOH, or CH3OH, as well as multicarbon products, e.g. acetic acid and ethylene, and C2+ alcohols (ethanol and propanol). This review focusses on the formation of multicarbon alcohols to summarize recent developments, which moved the electrochemical CO2 reaction a few steps closer to an industrial realization. Multicarbon alcohols are important target products in electrochemical CO2RR, as they are valuable basic chemicals for the chemical industry, can be used for energy production or as fuel additives (Jouny et al., 2018). An additional route for ethanol production thus further diversifies the feedstock for various products.With regard to the current state of the art, only CO and HCOOH are commercially viable. However, market analyses show that by further development of catalysts, electrodes, and cells and with it a consequent reduction in energy and product separation costs, higher alcohols are promising products for the future (Jouny et al., 2018; Somoza-Tornos et al., 2021). They have a larger market potential than CO and formic acid (Jouny et al., 2018). According to Jiao and coworkers, a yield of at least 62% should be achieved for n-propanol, and 77% for ethanol at −0.7 V to become economically feasible. The current densities should be in the range of 200–400 mA cm−2 (Jouny et al., 2018). A more detailed techno-economic analysis of the CO2RR products can be found in Review Electrochemical CO2 reduction - The macroscopic world of electrode design, reactor concepts & economic aspects.Multicarbon alcohols are formed during carbon dioxide reduction reaction (CO2RR) according to the following reaction equations: 2 CO 2 + 9 H 2 O + 12 e − → CH 3 CH 2 OH + 12 OH − 3 CO 2 + 13 H 2 O + 18 e − → CH 3 CH 2 CH 2 OH + 18 OH − Both the mechanism for the formation of C2+ alcohols and catalysts that enable the selective electrocatalytic CO2RR to C2+ alcohols are considered. Furthermore, the influence of process conditions and techno-economic considerations are also explained in more detail.In addition, it can be highly effective to couple electrochemical CO2 reduction with biocatalyzed methods. Schmid and coworkers achieved an FE of almost 100% for the conversion of CO2 to butanol and hexanol with a fermentation following the CO2RR using bacterium Clostridium autoethanogenum and C. kluyveri (Haas et al., 2018).In terms of the electrocatalytic conversion, different types of electrolysers are described in literature. Basically, they can be divided into three main types: liquid-phase, gas-phase, and solid-oxide electrolyser cell (Kibria et al., 2019). The oxygen evolution reaction (OER) usually takes place at the anode of the electrolyzers and the CO2RR at the cathode. One ubiquitous and dominating problem with CO2RR in general is the competing, parasitic reduction of water to H2 (hydrogen evolution reaction, HER) (Lv et al., 2018b; Albo et al., 2019; Gabardo et al., 2019; Gao et al., 2019; Martić et al., 2019, 2020; Xiang et al., 2019; Chang et al., 2020; Dutta et al., 2020; Kim et al., 2020b; Song et al., 2020; Wei et al., 2020; Zhang et al., 2020c; Herzog et al., 2021; Wang et al., 2021), which occurs in the same potential range as the CO2 reduction. Thereby, the Faraday efficiency for the formation of hydrogen in CO2RR with the target product ethanol is typically reported to be above 30% (Kim et al., 2020b). However, especially for the CO2RR to higher alcohols, the preferential formation of ethylene is a further problem and numerous studies focus on the selectivity inversion between ethylene and ethanol (Gu et al., 2021; Kim et al., 2021; Santatiwongchai et al., 2021; Wang et al., 2021).To indicate the selectivity of a catalyst or electrode, the so-called Faraday efficiency (FE, Equation 1) is given by (Equation 1) F E = z · n · F I · t · 100 (z - number of electrons transferred; n - amount of substance of product; F - Faraday constant, I - current applied; t - reaction time).Particularly for studies that focus on catalyst design and synthesis, H-type cells are widespread, despite the severe limitations of those systems (Burdyny and Smith, 2019). Its name is derived from its H-like form with cathode and anode compartments filled with liquid electrolyte, separated via an ion exchange membrane. The catalyst is usually deposited on glassy carbon or carbon paper and the CO2 is dissolved in the electrolyte. While this setup allows for simple and rapid testing of catalysts, it suffers from mass-transport limitation due to the low solubility of CO2 and can due to carbonate formation not be operated with alkaline electrolytes like KOH, which have been shown to improve CO2RR activity and C2+ selectivity (Carroll et al., 1991; Kibria et al., 2019). The low CO2 solubility and therefore availability limits the maximum current densities in H-type cells to about 100 mA cm−2, rendering them not feasible to be used in industrial CO2RR processes (Weekes et al., 2018). However, as the local conditions and, thus, selectivity are highly dependent on the current density and potential applied, the results obtained in an H-type cell make it difficult to draw significant conclusions about the catalyst performance under industrially relevant conditions. Those limitations demand the use of alternative setups for testing and optimizing of catalysts under realistic conditions at higher current densities (Weekes et al., 2018; Burdyny and Smith, 2019). This means that catalyst testing should be carried out under reasonable conditions like current densities of at least 200 mA cm−2 and stability tests of the catalysts and electrodes used of at least 24 h (Burdyny and Smith, 2019; Martić et al., 2019; Siegmund et al., 2021).Therefore, flow cells or gas-phase electrolysers (by using membrane electrode assemblies) should be used, in which gas and electrolyte streams are continuously supplied and cycled, respectively, to achieve the industrially relevant current density of >200 mA cm−2 (Weekes et al., 2018; Li et al., 2019b; Martić et al., 2019).The following brief overview describes the mechanistic background of the formation of multicarbon alcohols during the electrochemical CO2 reduction.In addition to general considerations on the mechanism of C-C coupling at the beginning of the chapter, various mechanisms found for diverse catalysts are further presented with only few catalysts being addressed here as examples. A more detailed discussion of the different catalysts and their operating principles is given in chapters "Structural properties and crystal orientations" ff. For CO2RR, copper plays a special role here, because it can form a variety of products and is the only metal capable of forming higher hydrocarbons and oxygenates. The diversity of possible products obtained by copper catalysts illustrates the complexity of the reduction reaction (Hori, 2008; Kuhl et al., 2012; Nitopi et al., 2019). For systematic optimization, a comprehensive understanding of the underlying reaction mechanism is fundamental.In the electrochemical CO2 reduction process, an initial electrochemical transfer of H+/e− to CO2 occurs. The resulting intermediate can bind to the electrode surface either via oxygen or via carbon. In the former case, formation of HCOOH can be expected, whereas in the latter case CO (Figure 1 ) is obtained, making this step crucial for the formation of the products in CO2RR (Cheng et al., 2016; Feaster et al., 2017; Chernyshova et al., 2018). Thereby, CO is widely considered as a key intermediate for further reduced C1 and C2 products, supported by investigations on the reduction of CO as well as in situ measurements (Hori et al., 1994, 1997; Wuttig et al., 2016; Gunathunge et al., 2017; Pérez-Gallent et al., 2017b; Bertheussen et al., 2018; Birdja et al., 2019; Nitopi et al., 2019).The C-C bond formation is the crucial reaction step that separates the pathways for single and multicarbon products. The dimerization of two ∗CO species is commonly considered a key step for the C-C bond formation, resulting in bidentate ∗CO∗CO as intermediate species. Figure 1 shows the proposed mechanistic pathway for the formation of ethanol and n-propanol (Cheng et al., 2021). The subsequent reduction steps to ∗CO∗CHOH or ∗CO∗COH have been considered as possible follow up intermediates (Calle-Vallejo and Koper, 2013; Kortlever et al., 2015; Montoya et al., 2015; Goodpaster et al., 2016; Cheng et al., 2017, 2021; Garza et al., 2018; Hanselman et al., 2018; Jiang et al., 2018; Todorova et al., 2020). Herein, the ∗COCOH intermediate could be observed via in situ IR spectroscopy (Pérez-Gallent et al., 2017a). Furthermore, operando Raman spectroscopy results suggest that the dimerization of ∗CO is competing with the hydrogenation to ∗COH or ∗CHO, which are further reduced to C1 products (Todorova et al., 2020). Along this line, C-C coupling steps via reaction of ∗CHO or ∗COH with CO to ∗COCHO or ∗COCOH also have been postulated (Goodpaster et al., 2016; Xiao et al., 2016; Garza et al., 2018; Jiang et al., 2018). Methylcarbonyl represents the most likely intermediate where a distinction takes place as to whether hydrogenation to ethanol or acetaldehyde occurs or whether further coupling with ∗CO and thus the formation of propanol takes place. In this case, the ∗CO attacks the carbonyl carbon of the acetaldehyde (Chang et al., 2020).Notably, the structure and properties of the (copper) electrodes have a significant influence on the C-C coupling step (Gao et al., 2019; Fan et al., 2020). It has been shown that the selectivity of CO2RR is dependent on the exposed copper facets. For example, Cu(110) and Cu(551) facets promote the formation of C2 products (Hori et al., 2002; Schouten et al., 2012, 2013; Kim et al., 2016). Engineering of catalyst size and morphology has been proven successful in steering the selectivity toward C2 products due to the exposed facets and differences in surface features like defect density, grain boundaries, and overall surface. Furthermore, various studies showed that morphological changes of the catalysts under the chosen process conditions have a significant effect on the product selectivity (Gregorio et al., 2020; Hou et al., 2020). In particular, too large particles as well as high current densities were identified as crucial parameters leading to aggregation and consequently to an altered product selectivity (Huang et al., 2018). The influence of structure on the pursued reaction mechanism was investigated for oxide-derived (OD) copper. It was shown that step square sites (s-sq) support the formation of C2+ alcohols, due to favorable thermodynamics for hydrogenation. In addition, the bond length between CO and the active site was correlated with the observed preferential product formation. For example, ethanol is preferentially formed at s-sq sites, which have the shortest determined bond length of 1.296 Å compared to planar-square and concave square, where ethylene formation preferentially occurs (Cheng et al., 2021). Another way of tuning catalyst selectivity is by adjusting the copper oxidation state. While the increased selectivity and activity of oxide-derived materials has partially been assigned to morphologic effects resulting from the reduction, results indicate that Cu+ and subsurface oxygen species play a role, too (Mistry et al., 2016; Favaro et al., 2017; Xiao et al., 2017b; Luna et al., 2018; Pander et al., 2018; Zhou et al., 2018). Recent results show that for copper-oxide-containing electrodes, reduction of the oxide layer occurs first before product formation due to CO2RR and HER (Löffler et al., 2021). The difference with pure copper electrodes is that the reduction of the oxide leads to the increased occurrence of defects and grain boundaries, resulting in a highly active surface.After C-C coupling, subsequent reduction steps lead to the multicarbon reduction products ethylene and ethanol. The possible intermediates and conceivable branching in the mechanistic pathway are, however, still under debate (Todorova et al., 2020). Bell and coworkers described ∗COCHO as first dimer intermediate followed by reduction to either glyoxal or ∗CO∗CHOH, and depending on the products formed, the reaction pathway proceeds either ethanol or ethylene, respectively. Glyoxal is subsequently reduced to acetaldehyde and ethanol (Garza et al., 2018). Acetaldehyde has been confirmed as an important intermediate toward ethanol formation via in situ NMR spectroscopy as well as mass spectrometry (Bertheussen et al., 2016; Clark and Bell, 2018). Other authors describe (as also can be seen in Figure 1) ∗CO∗COH as the key coupling product, whereby the mechanism then follows a different path via the reduction to ∗CCO. According to Goddard and coworkers, the next intermediate ∗CH∗COH is either dehydrated to form ∗CH∗C, which yields ethylene after another hydrogenation step, or to ∗CHCHOH, which is converted to ethanol via three further hydrogenation steps (Cheng et al., 2017; Xiao et al., 2017a). According to Calle-Vallejo and coworkers, acetaldehyde is the selectivity determining intermediate, which is converted to either ethylene or ethanol after further reaction steps (Calle-Vallejo and Koper, 2013; Hanselman et al., 2018). Contrarily, Asthagiri and coworkers postulated acetaldehyde and the two further hydrogenated species ∗CH2CH2O∗ and CH3CH2O∗ as three possible points where the pathways diverge (Luo et al., 2016). Hirunsit and coworkers mention the dissociation of the C-O bonds as most important for following the pathway either toward ethanol or ethylene formation (Santatiwongchai et al., 2021). Investigations on Cu(100) surfaces have shown that the protonation steps five to seven are decisive and if the C-O bond is about to break later, EtOH will be formed instead of ethylene. To conclude, this work shows that the following intermediates lead to ethanol: ∗CH3CO, ∗CH3CHO, ∗CH3CHOH, and ∗CH3CH2O whereas ∗CH2CH, ∗CCH2, and ∗CHCH lead to ethylene. ∗CHCHOH, ∗CH2CHO, ∗HOCH2CH2O, ∗CH2CH2OH, ∗CH2CHOH, and ∗HOCH2CH2OH are the intermediates which can result in either ethanol or ethylene formation.To increase the selectivity toward multicarbon alcohols, multimetallic catalysts are frequently used. For example, the ethanol to ethylene ratio could be increased by a factor of 12.5 by introducing zinc as a co-catalyst to copper (Ren et al., 2016). This is where the so-called spillover effect occurs. The effect was described not only for Cu-Zn (Ren et al., 2016) but also for Cu-Ag (Dutta et al., 2020; Martić et al., 2020; Ting et al., 2020), Cu-Pd (Rahaman et al., 2020), and for catalysts with Cu nanoparticles and pyridinic nitrogen in N-doped carbon (Han et al., 2020a). One of the mechanisms proposed for bimetallic catalysts is shown in Figure 2 . In this process, CO2 is reduced to CO at Zn, Ag, Pd, or pyridinic N sites, where CO is only weakly adsorbed (Ren et al., 2016; Han et al., 2020a; Rahaman et al., 2020) and CO migration to active copper sites can be achieved. There, CO is bound superiorly and will either be further reduced or undergo further reactions with adjacent ∗C1 and ∗C2 intermediates (Han et al., 2020a). With respect to the Cu-Ag-containing catalysts, the ratio of Cu: Ag is expected to have a direct influence on the product distribution due to an altered electronic structure (Martić et al., 2020). The interaction of copper and silver results in a shift of the Ed value, which represents the location of the center of the d-band, from that of copper at-3.30 eV by −0.56 eV toward that of silver (−5.36 eV). The electronic change results in less binding of CO2RR and HER intermediates, leading to preferential CO formation with FEs ranging from 55% to 68%. The main liquid product was ethanol with about 25% FE at 400 mA cm−2. Furthermore, the selectivity of 34.2% for ethanol in phase-blended Ag-Cu catalysts has been shown to be three times higher than with pure Cu2O (Lee et al., 2017). The authors emphasized the importance of the biphasic boundary for improved ethanol to ethylene selectivity. Upon modification of the distance between CO-producing Ag and Cu sites, increased insertion of CO and consequently formation of EtOH (demonstrated by ∗C2) can be achieved (Figure 3 ).Ag-Cu foams could be activated for ethanol production via a 12 h thermal annealing in air at 200°C. The obtained oxide-derived bimetallic catalyst showed a maximum FE of 33.7% for ethanol at −1.0 V and 6.9% for propanol at −0.9 V vs. RHE, while the formation of those products was negligible without the mentioned thermal treatment of the catalyst (Dutta et al., 2020).In addition to the spillover effect in bimetallic compounds, the combination of Cu nanorods (nr) and NGQ (nitrogen-doped graphene quantum dots) also enables an interesting mechanism. Oxygenated C2 intermediates were stabilized at the NGQ/Cu-nr, and by allowing both Cu-nr and NGQ to form C2 products, the formation of the multicarbon products is promoted by dual active sites. On both components, the existence of ∗CO as intermediate could be detected, but there was no evidence for a spillover or tandem effects (Chen et al., 2020a). Both effects describe the same process from a different point of view. However, while the term spillover effect describes the adsorption of the CO formed and its migration on the catalyst surface, the term tandem effect refers to the catalyst, i.e. that it has different domains on which different reaction steps take place. Therefore, a dual active-site mechanism was suggested, indicating the presence of active sites in NGQ as well as in Cu for the formation of C2+ products. In addition, the catalyst was found to stabilize the intermediate ∗CH2∗CHO, which is crucial for the higher FEs (52.4%) of multicarbon alcohols.Heteroatom-doped nanostructured carbon materials have also been examined as catalysts for the reduction of CO2 to alcohols. Their performance can be tuned via the nature and amount of heteroatom sites as well as the carbon morphology (Wu et al., 2019). A nitrogen- and boron-doped nano diamond catalyst reached a high ethanol selectivity of 93.2% at −1.0 V vs. RHE due to the synergistic effects of the heteroatom sites. The measurements were performed in H-type cells, with a CO2-saturated 0.1 M NaHCO3 electrolyte, and the total current densities were below 2 mA cm−2 (Liu et al., 2017). Because boron has an electron-poor p-orbital, it acts similarly to transition metals with an empty d-orbital and thus represents an active site for adsorption and subsequent reduction of CO as well as for CO2 (Zhu et al., 2021). For nitrogen-doped porous carbons, the high ethanol selectivity of 77% and 78% at −0.56 V vs. RHE has been attributed to synergistic effects between the carbon structure and active sites (Song et al., 2017, 2020). In addition, P-doping of catalysts could be used to adjust the adsorption strength for the CO intermediate. Thus, with P-doping, 2.8 times as much ethanol (15%) could be obtained with Cu0.92P0.08 C2+ product yield (Kong et al., 2021). Likewise, catalysts combining doped nanocarbons and copper catalysts have been described, reporting, e.g. tandem effects of heteroatom and metal sites with up to 64.8% FE for ethanol and 8.7% for propanol at −1.05 V vs. RHE (Song et al., 2016; Karapinar et al., 2019; Han et al., 2020a). However, it must be emphasized that the FEs of over 60% for ethanol achieved herein by different groups obtained under conditions of extremely low current densities between 2 and 16 mA cm−2. Hence, further improvements in systems allowing for higher current densities above 200 mA cm−2 are required to establish an industrial relevant process.In general, many mechanistic insights are obtained using computational methods such as DFT. Here, DFT is often used to show potential pathways for a target-oriented catalyst design and can reveal mechanistic information, e.g. regarding detailed reaction pathways (Li et al., 2020; Malkani et al., 2020; Santatiwongchai et al., 2021). The use of in situ techniques such as isotope labeling or the application of in situ spectroscopy such as XAS (X-ray absorption spectroscopy) or surface-enhanced vibrational spectroscopy methods can further help to complete the mechanistic understanding (Pérez-Gallent et al., 2017a; Malkani et al., 2020; Wang et al., 2020b). Studies of surface reconstruction in copper electrodes during CO2RR were e.g. conducted in 2017 by Waegele and coworkers as well as Koper and coworkers using Raman spectroscopy and Fourier transform infrared spectroscopy (FT-IR) (Gunathunge et al., 2017; Pérez-Gallent et al., 2017a). Furthermore, XAS has already been used to study the electronic as well as the coordinative structure on Cu catalysts during ongoing CO2RR (Xu et al., 2020; Herzog et al., 2021). Xu and coworkers describe in detail the advantages that in situ techniques offer, such as identifying the metals that provide the adsorption sites in the electrocatalytic reaction and analyzing metal-adsorbate interactions (Malkani et al., 2020). This contributes to a broader understanding of the mechanistic processes involved in CO2RR and for a more in-depth discussion on these techniques we refer to such papers. Table 1 provides an overview of recent developments in CO2RR to multicarbon alcohols, including the catalysts and electrolytes used as well as the resulting Faraday efficiencies. Firstly, copper and copper oxide as well as copper-oxide-derived (OD) catalysts are listed, followed by copper-carbon catalysts as well as copper catalysts, which were doped e. g. with boron or modified with halides, catalysts made of copper and another metal, and lastly miscellaneous catalysts, which do not fit in one of the categories mentioned before. The dominant usage of copper can be explained by its ability of producing multicarbon products during the reduction of CO2 (Loiudice et al., 2016; Garza et al., 2018; Karapinar et al., 2019; Malkhandi and Yeo, 2019; Jeong et al., 2020; Lei et al., 2020). The use of Cu electrodes in CO2 reduction experiments allows for the formation of a broad variety of products. Cyclic voltammetry (CV) measurements yielded CO, allyl alcohol, propionaldehyde, n-propanol, acetaldehyde, EtOH, ethylene, and methane in varying amounts and ratios (Clark and Bell, 2018). The table furthermore summarizes the FEs of the respective products. Thereby, it becomes visible that the selective formation of multicarbon alcohols still possesses a challenge. The products marked “C2+” usually contain high amounts of C2H4, which is often the main reason for the high overall FEs. This effect is a result of the fact that ethylene is generally preferred to ethanol formation in copper-based electrodes (Ren et al., 2016).Nevertheless, catalysts of various compositions already achieved FEs above 50% for multicarbon products. Best results were obtained with up to 85% FEEtOH using Ag-graphene-NCF (Nano Carbon Fibers) as the catalyst, but the resulting current density was less than 1 mA cm−2, essentially not allowing any conclusive results on potential applications in larger scale (Lv et al., 2018b). Catalysts made of Cu-N-C (Karapinar et al., 2019), Cu-NPC (Han et al., 2020a), or consisting of micropores in N-doped mesoporous carbon (Song et al., 2020) also reached high Faraday efficiencies above 55% for ethanol. However, all of these catalysts/electrodes were operated at industrially irrelevant current densities of less than 20 mA cm−2. An intriguing question is what the performance or product distribution of these catalysts and electrodes will be at higher current densities. In contrast, higher current densities with simultaneously increased FEs for ethanol were obtained with Cu sputtered on PTFE and NC (FEEtOH 52% at partial current densities of 156 mA cm−2) (Wang et al., 2020b) or N-doped graphene quantum dots on Cu-OD Cu nanorods with a FEC2+ alcohol of 52.4% at a total of 282 mA cm−2 (Chen et al., 2020a).As can also be seen from Table 1, the most frequently used electrolytes are KHCO3, CsHCO3, and KOH. However, because the influences on the resulting selectivity of the catalysts is multifactorial and involves not only the electrolyte but also other aspects such as cell design, membrane, temperature, and other parameters, the influence of those is discussed in detail in the chapter “Process Conditions”.The syntheses of solid electrocatalysts, which are capable of producing ethanol during the electrochemical reduction of CO2, are manifold. In the most common cases, precipitation methods or electrodeposition were used, as well as sputtering of thin films. To further optimize the performance of the catalysts, surface modifications or reconstructions were also frequently carried out, or the catalyst layer was created by means of evaporation (e.g. via chemical vapor deposition).During electrodeposition, the catalyst is plated directly onto a substrate from an electrolyte solution, whereby the substrate is used as a working electrode and the deposition can be galvanostatic or potentiostatic. Electrodeposition has so far been used to coat gas diffusion layers, like carbon paper (Aeshala et al., 2012; Hoang et al., 2017, 2018; Lee et al., 2017; Kong et al., 2021), but also other substrates like metal foams, polished Cu discs, or Cu foil (Dutta et al., 2016, 2020; Ren et al., 2016; Rahaman et al., 2017, 2020; Kim et al., 2020b), which were then often applied in H-cells. Often, these catalyst materials were deposited from sulfuric acid, CuSO4, and other metal-sulfate-containing electrolytes (Aeshala et al., 2012; Dutta et al., 2016, 2020; Ren et al., 2016; Hoang et al., 2017, 2018; Rahaman et al., 2020). In addition, additives such as sodium citrate (Dutta et al., 2020) or citric acid (Kong et al., 2021), 3,5-diamino-1,2,4-triazole (DAT) (Hoang et al., 2017, 2018), as well as lactic acid (Ren et al., 2016; Lee et al., 2017) were added to the electrolyte solution. Sodium citrate was used in the deposition of Ag15Cu85-foam on Cu foil (Dutta et al., 2020). The deposition was realized from silver and copper(II)-sulfate-containing electrolyte at 3 A cm−2. In this process, the competing HER commonly results in the formation of gas bubbles as a geometric template for foam formation. Figure 4 schematically shows the process of deposition of porous copper using the resulting hydrogen as a template. The sodium citrate used should have an impact on the growth characteristics through chemisorption at the cathode surface. However, the electrodeposition of foams on Cu wafers was also successfully carried out without additives using sulfuric acid/CuSO4 solution at 3 mA cm−2 (Dutta et al., 2016). Owing to the mesoporous structure of the resulting Cu foam, there is an increased formation of C2 products such as ethane and ethylene. In the case of the Cu-Ag foams, subsequent calcination at 200°C and the associated formation of Cu2O also led to increased Faraday efficiencies for EtOH and PrOH of up to 33.7% and 6.9%, respectively (Dutta et al., 2020). Calcination was also carried out following the electrodeposition of Cu dendrites on electropolished meshs (Rahaman et al., 2017) and a Cu-Pd foam on Cu foils (Rahaman et al., 2020), to activate the catalyst as this thermal treatment may result in a higher FE for ethanol instead of CO due to segregation of the phases (Dutta et al., 2020). Zeng et al. electrodeposited Cu onto carbon paper and used thermal annealing to dope the Cu with phosphorus at 400°C and under N2 atmosphere using NaH2PO2∙H2O (Kong et al., 2021). The yield of C2+ products was thus increased by 1.9 times, and the FE for EtOH was even 2.8 times higher (15%) than without any doping. Another used additive is (3,5-diamino-1,2,4-triazole) DAT, which acts as an inhibitor for Cu deposition before reaching −0.18 V (vs RHE) (Hoang et al., 2017, 2018). As a result, it was possible to deposit Cu films with a high surface area and activity for CO2RR. It was possible to achieve 5–6 times higher current densities, when DAT was used as an additive in the deposition process than without (Hoang et al., 2018). Lactic acid was also used as an additive as it stabilizes the Cu ions in the solution (Ren et al., 2016; Lee et al., 2017).Another commonly used method for catalyst synthesis is precipitation, and a broad variety of starting materials and products have been used or obtained. In some cases, the syntheses were carried out in the microwave, such as a precipitation reaction for Bi-MOFs (Albo et al., 2019), or in autoclaves as in the synthesis of Cu(OH)F from Cu(II) nitrate in DMF, with the addition of NH4HF2 for 4 h at 160°C (Ma et al., 2020b). In CO2RR, the resulting Cu(OH)F catalyst enabled the formation of C2+ products with FEs up to 65.2% at a maximum current density of 800 mA cm−2. Another example of the use of autoclaves is the preparation of paramelaconite (Cu4O3) from a Cu(II) nitrate-DMF-EtOH mixture after the addition of formic acid and dimethylamine at 130°C (Martić et al., 2019). The catalyst achieved an FE for C2+ products of over 61%. Cu nanoparticles can also be prepared by precipitation over the formation of copper oxides as shown by Jiao and coworkers. They precipitated Cu(OH)2 nanorods from a mixture of aqueous copper nitrate solution with ammonia and converted them to porous CuO by thermal annealing in the following (Lv et al., 2018a). After applicating the nanorods onto a GDL, the reduction to copper nanoparticles was performed at 10 mA cm−2. Another example is the precipitation of Cu(OH)2 followed by thermal annealing under a H2/Ar atmosphere. The catalyst was then partially oxidized by storing it in air before applying it to CO2RR (Shang et al., 2019). This procedure yielded core-shell Cu@Cu2O catalysts, which led to an FE of EtOH of 29% during CO2RR. Another example is the precipitation of Cu nanoparticles from a Cu(II)-containing solution with the addition of NaBH4 (Ma et al., 2016; Wei et al., 2020), which led to an FE for C2+ products of up to 80% (Wei et al., 2020). If a NaBH4 solution is combined with a CuCl2 solution, boron-doped copper can be obtained as a precipitate, which achieves FEs for EtOH up to 27% (Zhou et al., 2018). In addition to pure Cu precipitates, mixed oxides as well as other compounds with several metals have been successfully synthesized via precipitation and used for electrochemical CO2 reduction. For example, a catalyst of graphene oxide, ZnO, and Cu2O was prepared by precipitation and produced up to 30% propanol (Geioushy et al., 2017). Precipitation of Ag2Cu2O3 with aqueous NaOH from a solution containing Cu and Ag nitrate was also successfully carried out under inert conditions, and the Faraday efficiency for this catalyst was about 25% for EtOH (Martić et al., 2020). Another catalyst that produced nearly 30% ethanol, when used in a GDE, is a CuPb-0.7/C (Pb shell thickness is 0.7 nm) catalyst, which was precipitated from a copper acetate, PbCl2, ascorbic acid, diphenyl ether, and oleylamine-containing solution (Wang et al., 2020a). With 25% FE, slightly less ethanol was produced by V-Cu2S nanoparticles, which were prepared using Cu acetylacetone and dodecanethiol (Zhuang et al., 2018).Another method for the preparation of catalysts or electrodes, which has already led to materials providing high Faraday efficiencies at industrially relevant current densities, is sputtering. In most cases, Cu was sputtered from a pure Cu target onto a PTFE membrane (Dinh et al., 2018a; Gabardo et al., 2019; García de Arquer et al., 2020; Li et al., 2020) (pore size 0.45 μm). Subsequently, either carbon black (Dinh et al., 2018a; Gabardo et al., 2019), graphite (Gabardo et al., 2019), Nafion, or a mixture of Nafion and Cu-NPs (García de Arquer et al., 2020) was spray coated onto the sputtered layer. Spray coating of porphyrin-based complexes (FeTTP) onto sputtered copper ultimately resulted in an FE for ethanol of 41% (Li et al., 2020), as did co-sputtering of Cu and Ag onto a PTFE membrane (Li et al., 2019b). The highest EtOH yield was obtained with an FE of 52% by first sputtering Cu and then a layer of N-C onto the membrane (Wang et al., 2020b).Some catalysts were also synthesized via evaporation and vapor deposition onto a substrate. For example, compared to pure Cu, alcohol formation occurred at >265 mV more positive electrode potentials on a polycrystalline Cu foil coated with gold (Carlos G. Morales-Guio et al., 2018). Furthermore, CVD of boron- and nitrogen-doped diamond on a Si substrate was performed and the resulting electrode led during CO2RR to an FE of 93.2% for ethanol, but with current densities below 2 mA cm−2 (Liu et al., 2017).For modifying or reconstructing the surface of Cu foils/substrates, various ways including electrochemical and plasma activation were used. However, the resulting catalysts were always used in H-type cells, which lead to very low current densities. One possibility of surface modification for copper foil is to cyclize it. For example, the FE for ethanol could be increased from 2.2% to 7.7% by cyclizing the foil for three cycles between −1.1 and 0.9 V for 20 mV/s in a 0.1 M KHCO3 solution, containing 4 mM KCl (Schouten et al., 2011). Cyclization in copper nitrate solution led to the formation of single crystal Cu2O nanocubes and an FE for C2+ products of 60% was obtained (Jiang et al., 2018). Another possibility to modify the catalysts surfaces is plasma activation in O2 plasma (FEC2+ 69%) (Gao et al., 2018) or heating a copper substrate in an oven to 1100°C followed by quenching in air, leading to the formation of sponge-like structures and an FE for C2+ products of 70% (Lei et al., 2020). Wet chemical modification of the surface by oxidation with H2O2 and diluted HCl leads to the formation of CuCl on the surface, followed by the formation of Cu2O by immersion in KHCO3 (Kibria et al., 2018). Subsequent electrochemical CO2 reduction then led to FEs for C2+ products above 80%. Also, modification with halides was obtained by immersing Cu foils in solutions containing CuBr2 (Wang et al., 2021). Here, CuBr tetrahedrons formed on the surface which were subsequently immersed and thus uniformly coated in dodecanethiol. The application of the coated catalyst in CO2RR resulted in almost 36% FE for EtOH.Apart from the synthesis routes described so far, various catalyst syntheses can be found which were only used by a few groups including special synthesis routes—e.g. a 4-step organometallic synthesis of co-corroles (48% EtOH, −0.56 V, total 2.5 mA cm−2) (Gonglach et al., 2019), the synthesis of Cu-N-C by low-energy ball milling followed by pyrolysis in an argon stream (55% EtOH, −1.2 V, total 16 mA cm−2) (Karapinar et al., 2019), the impregnation of melamine foam in a silver nitrate-graphene oxide solution followed by calcination (79%–85% EtOH, −0.5 to −0.7 V, total 0.3 mA cm−2) (Lv et al., 2018b), or the preparation of carbon supported Cu catalysts by using an amalgamated Cu-Li method (91% EtOH. −0.7 V, total 1.2 mA cm−2) (Xu et al., 2020).Regarding the synthesis and study of electrocatalysts versus industrial applicability, our group has recently published a perspective article (Siegmund et al., 2021). There we defined the following evaluation criteria: (1) The issue of complexity and price required to synthesize the catalyst: Synthesis routes such as multiple steps synthesis are considered problematic in this regard, as they are accompanied by great complexity, as well as costly purification steps. Precipitation reactions, sputtering, or electrodeposition, on the other hand, are in simple principle and can be carried out in just a few steps. The processes described under “Surface modification” can also be described as predominantly less complex. (2) The issue of producing the catalyst in sufficiently large quantities (Siegmund et al., 2021): E.g. it is possible to sputter large areas without any problems, which is already used for the production of thin-film solar cells (Edoff, 2012). Precipitation reactions are also common processes in industry and have the potential to be carried out on a large scale, as does electrodeposition of metals. However, individual considerations would need to be given to each catalyst synthesis in terms of its scalability. More problematic are synthesis routes which contain discontinuous processes, e. g. evaporation processes. (3) The issue of (long-term) stability of the catalyst materials at relevant current densities (Siegmund et al., 2021): Some catalyst materials mentioned above have already been tested for their stability over longer time periods, e.g. Co-corroles showed stable electrolysis over 140 h, but at very low current densities of −2.5 mA cm−2 (Gonglach et al., 2019). Also, sputtered electrodes were already stable over 150 h electrolysis (at up to 100 mA cm−2) (Dinh et al., 2018a). For many of the catalysts, however, evidence of long-term stability under industrially relevant conditions is lacking, which is urgently needed to evaluate the applicability of the materials.In addition to the composition of the catalyst, its surface morphology and crystal face orientation were determined to be decisive factors in the selective reduction of CO2 to C2+ alcohols and therefore the factors that increase the FE for multicarbon product formation are discussed here.Several studies have already shown that Cu(100) surfaces are more selective for C2+ products, while Cu(111) is more likely to lead to the production of CH4 (Jiang et al., 2018; Wang et al., 2019; Gregorio et al., 2020; Han et al., 2020b; Ting et al., 2020). However, an excess of CO at Cu(111) sites could also lead to EtOH formation. Cu(100), on the other hand, supports the dimerization of ∗CO, which is formed as intermediate (Han et al., 2020b). The selectivity via the surface orientation is also evident when using Cu nanocubes and Cu nanospheres. As more Cu(100) is present on the surface in the former, the ethylene formation under alkaline condition is more pronounced (Jiang et al., 2018; Wang et al., 2019). Another example for the advanced C-C coupling on Cu(100) can be observed on CuCl-derived Cu electrodes as they show an increased selectivity for C2 products (Kibria et al., 2018). Compared to electropolished electrodes, those CuCl-derived ones show a change in preferential crystal orientation from Cu(111) to Cu(100). Upon transition from Cu(111) to Cu(100), FEs for C2+ products increased from 30% to 73%, that of propanol from 0% to 5%.When comparing Cu cubes and Cu octahedrons, the formation of C2H4 was also highest at the cubes, whereas CH4 formation was more pronounced at the octahedrons (Gregorio et al., 2020). Furthermore, it was shown, using Cu-Zn catalysts as an example, that the roughness factor of the surface directly influences the product distribution. Higher roughness correlated with higher FEs for C2+ products (da Silva et al., 2020). Figure 5 shows the influence of surface morphology on CO2RR in terms of C2+ product distribution and the influence of Cu-Zn ratio on catalytic activity. While the Faraday efficiency for the formation of C2+ products increases with increasing surface roughness, it simultaneously decreases for CH4 and H2 (Jeong et al., 2020). The presence of corners and steps on the surface promotes the adsorption of C1 products and this, in turn, leads to an improvement in the dimerization to C2+ products (Hoang et al., 2018). The improvement in C2+ production due to both more sharply defined structures and more curved surfaces is expected to occur as a result of improved bubble nucleation, a concentration of stabilizing cations as well as high local fields and thus increased current density (Luna et al., 2018). Electro-redeposition is expected to lead to these electronic and morphological effects, which improves selectivity and activity of Cu in the production of C2+ during CO2RR. Furthermore, the yield of C2-C3 products could be significantly increased by in situ structural transformation of densely packed Cu-NPs by electrolysis to cube-shaped catalytically active structures (Kim et al., 2017).Besides the surface roughness, porosity also plays an important role in the electrochemical performance of CO2RR (Han et al., 2020a). For example, the transport of CO2 through the electrolyte-electrode interface at high current densities is facilitated when using GDEs with highly porous structures (Lv et al., 2018a). In addition, the micropores are also expected to play an important role in the adsorption capacity of CO2 by the catalysts (Han et al., 2020a).In addition to the catalysts themselves, the type of electrode and its manufacture also have a significant influence on the final performance in CO2RR (Tan et al., 2020). Catalyst ink-based preparation techniques, for example, offer the possibility to influence catalyst surfaces via multiple parameters. In addition to using different techniques such as dropcasting, airbrushing, or hand painting, the drying temperature can also be adjusted. Overall, thinner porous catalyst layers, e.g. obtained by dropcasting or hand painting, should result in fewer C2+ products being formed. If, on the other hand, the catalyst layer is enlarged, there is better CO2 mass transfer within the porous layer. Simulations suggest that the layer thickness is more important than the porosity for controlling the local concentration of CO2 (Tan et al., 2020). In addition to the layer thickness, the loading of catalyst also influences the results. For example, the study of Cu-NPs in combination with pyridinic N species in N-doped porous carbon showed that a copper loading of 10% was not sufficient, whereas 30% was too much and led to preferential ethylene formation instead of EtOH and PrOH. The highest yields for multicarbon alcohols were obtained at 20% Cu loading (Han et al., 2020a).Whether to use copper, copper oxide, or OD-copper electrodes is a frequently discussed topic. In comparison to pure copper electrodes, oxide-derived copper electrodes contain remaining oxides, which should simplify the adsorption of ∗CO and the C-C coupling (Ting et al., 2020). Thus, OD-Cu should increase the selectivity for C2+ products (Iijima et al., 2019). Furthermore, investigations have shown that a thin layer of metastable Cu2O on an electrode made of OD-Cu can result in an increase in selectivity in favor of C2 products due to an improved stabilization of intermediates of CO2RR (Shah et al., 2020). It was also shown that current densities are higher on plasma-activated Cu foil (CuO2) than on electropolished Cu (Singh et al., 2016; Gao et al., 2018). This is not due to structural changes but can rather be understood as a chemical effect of Cu+ species. An increase in the product ratio for CO2RR of C2+/C1 with FEs of up to 61% for C2+ products was also shown by using a GDE, which is carbon-based and contains Cu derived from Cu4O3. Partial current densities of 185 mA cm−2 were obtained and due to the same reaction paths and intermediates of EtOH and ethylene, an increase in both C2H4 and EtOH yield was obtained by improvement with OD-Cu compared to normal Cu (Martić et al., 2019). An improvement in C2+ selectivity was also recently achieved by using Cu catalysts with nanocavities in which carbonaceous intermediates are trapped (Yang et al., 2020). The intermediates would not only cover the surface of the catalyst but also stabilize the Cu+ present there, which is thus also retained during CO2RR and allows the selectivity to be increased (75.2% FE at 267 mA cm−2) (Yang et al., 2020).However, other investigations show that only metallic copper is active, while oxides remaining in OD electrodes are unstable and inactive under CO2RR conditions during catalysis (Ting et al., 2020). Spectroscopic investigations have shown that there is a low CO intermediate formation on Cu2O, resulting in a low activity toward CO2RR. According to Han and coworkers, CO2RR takes place on Cu0 and not Cu+ or Cu2+ and the oxides are not decisive for selectivity toward C2+ products. Instead, they examined the grain sizes and found a decrease of selectivity in the order Cu0 > Cu+ > Cu2+, and that the reduction of the oxides leads to fragmentation and thus to an increase in surface roughness (Lei et al., 2020). A direct comparison of electropolished Cu electrodes with those containing Cu oxide or Cu hydroxide showed that electrodes containing Cu oxides or hydroxide showed better selectivity for C2+ products while suppressing the formation of CH4 (Lei et al., 2020). The best results were obtained with Cu oxide electrodes with an FE of 68.2% and up to 64 times higher current densities than the pure Cu electrode. In the catalyst, three Cu species coexisted in different layers—Cu0, Cu+, and Cu2+. Within 1 h of CO2RR, all species were reduced to Cu0, but fragmentation to irregular nanoparticles also took place. The resulting network shows an enrichment of highly active sites, which facilitates CO adsorption. Furthermore, more high-index facets were exposed. These effects resulted in the improved selectivity (Lei et al., 2020). An investigation on Cu(100) surfaces using pulsed potential sequences (0.6 V and −1.0 V for 1 s each) also led to an increase in selectivity for C2+ products. While potentiometric measurements at −1 V on Cu single-crystal electrodes achieved FEs for EtOH of 8% and for ethylene of 45%, the overall value increased to 76% for the products, with ethanol FEs around 30%. The increased selectivity for ethanol is explained via a continuous in situ regeneration of Cu(I) and thus the co-existence of Cu(I) present as Cu2O and Cu(0) on the surface, the Cu(100) domain, and the defect sites (Arán-Ais et al., 2020).Sargent and coworkers showed with the help of XAS measurements of GDEs that a direct reduction to metallic copper in the catalyst layer was achieved within 16 s, which implies that Cu0 is responsible for the selectivity toward EtOH and not the presence of oxides (Wang et al., 2020b). In addition to the question to what extent oxides themselves have an influence on the selectivity of CO2RR at Cu electrodes, the influence of interparticle distances between CuOx nanoparticles was also investigated (Jeong et al., 2020). It was shown that increasing the distance between those NPs improves the C2+ selectivity, as long as it is still < 1nm. The C1 product formation was lowered and the obtained current densities were up to 12 times higher than with the unmodified catalyst. The reason for this was a higher surface roughness (increased ECSA) and a lowered energy barrier for CO2RR. Again, Cu+ was reduced to Cu0 during the reduction reaction (Jeong et al., 2020).Another strategy is the combination of copper oxides with copper in the catalyst via the formation of Cu@Cu2O core-shell catalysts (Shang et al., 2019). The synergy between Cu0 and Cu+ leads to an increase in selectivity and efficiency in the formation of C2+ products, whereby dimerization should be facilitated by promoting the formation of a positive- and a negative-charged carbon atom (Shang et al., 2019).Besides the influence of surface activation or, for example, the use of OD-Cu electrodes, the influence of halides in Cu-based electrodes was also of interest for the CO2RR. Therefore, Wang and coworkers produced halide-containing copper catalysts via a precipitation process and found during electrochemical measurements in a flow cell that the adsorption capacity increases in the following order: Cu < I-Cu < Br-Cu < Cl-Cu < F-Cu (Ma et al., 2020b). Overall, the C-C coupling works better the higher the coverage of the surface with ∗CO is. The authors suggested that the presence of Cu+ sites may increase CO adsorption. In connection with C2H4 formation, they found that in the Cu-halides catalysts with increasing electronegativity of the halide, only a slight decrease of the onset potential could be observed. This indicates that the copper catalysts modification with halides promotes the first step after the ∗CO intermediate formation. Furthermore, a dependence on the local pH value was observed. Thus, a significant increase in C2+ formation (with FEs of EtOH up to 15%) for F-Cu catalysts with increasing local pH was observed when using different 0.5 M electrolytes in the following order: K2HPO4 < K2CO3 < K2SO4 (Ma et al., 2020b). Figure 6 shows both the influence of the halide on the formation of C2+ products and the influence of the KOH concentration for the Cu-F catalyst.Another recently published study shows the production of a halide-containing copper catalyst by oxidative-reductive recycling of polycrystalline copper in KHCO3 solution with addition of the corresponding potassium salt (Han et al., 2020b). While Cl− and Br− stabilized Cu+ and thus tend to be promoters of Cu dissolution, I− inhibited it by forming an almost insoluble polyhedral CuI, along with the associated passivation of the surface. This cycling of copper in KHCO3 solution resulted in different structures on the surface of the Cu electrode depending on the halide. Although the reconstructed (re) Cu-I electrode had less Cu(100) on the surface compared to re-Cu-Br and re-Cu-Cl electrodes, the best selectivity for these electrodes was obtained for the copper electrodes modified with iodide with 80% FEC2. XAS measurements showed the same ratio of Cu0 to Cu+ for all three electrodes, rendering it not decisive for the selectivity. However, a correlation of the electrochemical performance during CO2RR was observed with the porous, in the case of re-Cu-I intertwined and spiderweb-like, hierarchical structure on the surface. The intermediately generated CO is supposed to be trapped inside the pores, providing an increased ∗CO-coverage, which leads to an increased dimerization (Han et al., 2020b). The question of how the addition of halides in the electrolytes affects the electrochemical performance of the catalysts and electrodes is discussed in the following chapter about process conditions.In addition to the previously discussed modifications of the catalysts with halides, there are also investigations on the influence of hydroxide. It was shown that the presence of OH groups near the catalyst surface improves the reaction kinetics and stabilizes the oxygen in CuxO catalysts during the reduction reaction (Xiang et al., 2019). Also, with increasing number of OH− bound to Cu, the adsorption of CO and thus also dimerization should be supported (Iijima et al., 2019). The adsorption energy of CO will be increased compared to pure Cu surfaces, because the OH layer will probably bring CO molecules closer together while a simultaneous reduction to C2+ products takes place.A further modification reported in the literature is the coating of a Cu foil with a 50 nm thick polyaniline film (PANI), whereby an improvement of the C2+ selectivity from an FE from 15% to 60%, for the coating of Cu nanoparticles even to 80%, was achieved (Wei et al., 2020). The PANI layer is intended to increase the coverage of the surface with CO and improves the interaction of these molecules. At the same time, HER is significantly reduced, probably due to the increased hydrophobicity. Moreover, Mougel and coworkers created a superhydrophobic surface on their applied electrode by treating Cu dendrites with 1-octadecanthiol, resulting in an FE of 56% for C2H4 and 17% for EtOH at neutral pH (Wakerley et al., 2019). The gas was captured at the electrode-electrolyte interface, which resulted in an increase in CO2RR and C2+ selectivity. Modification of surface hydrophobicity and adsorption energies is also possible by combining the use of halides and organic compounds (here dodecanethiol) (Wang et al., 2021). Dodecanethiol lowers the selectivity for H2 and CH4 by decreasing the amount of adsorbed H∗. The bromide introduced into the copper catalyst, on the other hand, shifts the selectivity to ethanol by stabilizing positive Cu valence sites, which are expected to have a significant effect on the product distribution in CO2RR (Wang et al., 2021).In addition to varying the oxidation states of copper or creating specific structures on the catalyst surface, bifunctional catalysts can be used to improve the selectivity for C2+ products. The potential for the formation of CO at the co-catalyst should correspond to the potential range for the formation of the target product at copper (Ren et al., 2019). As discussed before in multimetallic and bifunctional catalysts, the combination with ZnO can increase the C-C coupling kinetics by increasing the local concentration of the intermediate CO (Zhang et al., 2020b). In the case of Cu/ZnO tandem electrodes, additional CO was generated at the ZnO, and the resulting CO excess increased the C2+ selectivity by facilitating C-C coupling. The electrodes showed a stability of 10 h at 600 mA cm−2 (Zhang et al., 2020b). The use of ZnO for increased selectivity of C2+ products in a Cu/ZnO tandem catalyst as a bifunctional catalyst with different domains was also shown by other groups. Grätzel and coworkers modified CuO nanowires via atomic layer deposition with ZnO, thus shifting the selectivity of CO and HCOO− (selectively formed on Cu nanowires) toward EtOH (Ren et al., 2019). Herein, the additional active sites of zinc available for CO intermediate formation increase the amount of CO for C-C coupling and thus reduce HER at the same time. Figure 7 shows the proposed mechanism in more detail including the impact of varying the overpotential. Higher overpotentials lead to higher production of ∗CH3, which can then be coupled with CO to form ethanol. Another example for bifunctional catalysts in CO2RR is the combination of copper with silver for obtaining enhanced yields for C2+ product (Hoang et al., 2018). Sargent and coworkers made efforts in designing catalysts that favor the CO2RR pathway to ethanol. The diverse binding sites, existing in Ag-Cu bimetallic catalysts, led to a destabilization of the ethylene intermediates, probably due to a disruptive influence of Ag on ethylene-forming Cu sites. This resulted in an increased ethanol selectivity of 41% at −0.67 V vs. RHE, compared with an FE of 29% at best for the pure Cu catalyst (Li et al., 2019b). Cu-Pd foams also revealed good catalytic activity toward CO2RR to C2+ products. This catalyst shows phase segregation in the nm range, with Cu- and Pd-rich domains present. These lead to a 2 times higher selectivity toward PrOH instead of EtOH. The methane pathway (C1) is suppressed and a concerted spillover effect of ∗CO and ∗H adsorbed on Pd domains results in the preferential formation of C3 products (Rahaman et al., 2020). A catalyst for selective alcohol formation is an OD-Ag-Cu-foam of stoichiometry Ag15Cu85 (Dutta et al., 2020). CO is selectively formed in the silver domains and is transferred by surface diffusion to copper, where it is converted to alcohols by C-C coupling. The excess of CO at the catalysts surface leads to good selectivity with up to 34% FE for ethanol. Furthermore, a selective activation of the copper by oxide deposition and the subsequent reduction under CO2RR conditions takes place and enhances the selectivity as well. Doping biphasic (BP) copper(I) oxide with silver also yields significant improvements in EtOH yield, including a shift in product selectivity from ethylene to divalent alcohols (Lee et al., 2017). The FE for EtOH was raised from 11% to 35% for Ag-Cu2OBP compared to the undoped catalyst. Another option is the destabilization of the ethylene reaction path in favor of an increased EtOH production by using a Ag/Cu-alloy phase catalyst (Li et al., 2019b). Ethylene is preferentially formed at highly coordinated surfaces and the introduction of an element with a weaker bonding capacity to carbon than copper reduces the probability of the formation of ethylene intermediates by increasing the variety of available bonding sites. On Cu (111), there are four bonding sites available, on Ag-doped Cu (111), there are 16.A concept that has been applied several times is the use of core-shell catalysts, where CO is enriched inside the nanocaves by reduction of the core, and is then converted by the shell into the target product (Zhuang et al., 2018; Ren et al., 2019; Shang et al., 2019; Zhang et al., 2020a). An example for the production of ethanol is a catalyst consisting of Cu2O nanocavities with embedded gold nanoparticles, which shifts the selectivity for CO2RR from C1 to C2 products (Zhang et al., 2020a). The gold core reduces CO2 to CO in the nanocavities, resulting in a high local concentration of this intermediate. EtOH is then formed at the copper shell. Another core-shell catalyst developed by Sargent and coworkers consists of a Cu2S core and a Cu-V shell (Zhuang et al., 2018). This catalyst achieved an FE of 32% for alcohols, with 25% for EtOH and 7% for PrOH at a partial current density of 120 mA cm−2, resulting in a 6-fold improvement of the EtOH: ethylene ratio from 0.18 to 1.2 compared to pure Cu nanoparticles.Another possibility for increasing selectivity toward multicarbon alcohols is the use of metal organic frameworks. The use of Cu(II)- and Bi(III)-based MOFs resulted in a FEEtOH of 28.3% (Albo et al., 2019). However, these electrodes are only stable for 5 h. The increased EtOH formation can be explained by the reduction of CO2 at Bito HCOO−, which is then transferred to Cu and reduced to alcohol. Owing to longer diffusion paths within the MOFs compared to other catalysts, a longer contact of the products is guaranteed and a reduction of MeOH to EtOH under C-C coupling can take place. A longer stability with up to 140 h was achieved for electrodes by using Co-corrole carbon paper electrodes (Gonglach et al., 2019). The mechanism here is not based on CO as an intermediate but the formic acid pathway and Co-corroles stabilize various radical intermediates. EtOH could be obtained with an FE of 48%.Furthermore, metals were incorporated into various carbonaceous support materials, e.g. an N-doped porous carbon-supported copper catalyst was used for CO2RR to multicarbon alcohols (Han et al., 2020a). Pyridinic N-species were probably the CO-producing sites and copper the catalytic sites for the production of EtOH and PrOH. An increase in pyridinic nitrogen atoms improved both selectivity and activity toward multicarbon alcohols. The carbon support influenced the copper concerning structure and size, resulting in improved CO2 adsorption and CO production. Pyridine nitrogen was also used in a catalyst consisting of Ag nanoparticles in a 3D-graphene-wrapped nitrogen-doped carbon foam, as it can bind ∗CO intermediates better than other N-species (Lv et al., 2018b). EtOH is then gradually formed at the Ag-NPs. The catalyst is also characterized by high conductivity. The direct comparison of Cu nanorods with nitrogen-doped graphene quantumdots (NGQ) and Cu nanorods clearly shows higher EtOH and PrOH yields (Chen et al., 2020a). The reason for the increased formation of multicarbon alcohols is the better stabilization of the oxygen-containing intermediates. As mentioned and discussed before in multimetallic and bifunctional catalysts, there is also a synergistic effect, as C-C couplings occur at both the copper nanorods and the NGQ, and the formation of the desired C2+ products is greatly enhanced by these dual active sites.Besides the mentioned combination of metals and carbonaceous supports and the usage of bimetallic catalysts, a molecule-metal composite has been proposed. The porphyrin-based co-catalyst increased ∗CO coverage on the metal surface, promoting C-C coupling and favoring the ethanol pathway. The FE for ethanol was 41% at −0.82 V vs. RHE, higher than 29% FE at –0.84 V observed for pure Cu (Li et al., 2020). Molecular cobalt corrole complexes have been described, with the electron-donating ligands favoring a square-planar cobalt(I) complex as active species. It could reach an ethanol FE of 48% at −0.8 V vs. RHE (Gonglach et al., 2019). Also, acetate as potential C2 product could be obtained using a manganese corrole complex with 63% FE at −0.67 V vs. RHE (Schoefberger et al., 2020).Metal-free catalysts have also already been used for selective EtOH production, e.g. N-doped mesoporous carbon. High local electrical potentials within the mesoporous channel walls lead to an improved activation of CO2. In addition, this also facilitates C-C coupling through the pyridine and pyrollic nitrogen atoms. The micropores contained in the channel walls increase the selectivity of the catalyst for EtOH as well as the reactivity (Song et al., 2020).In recent years, there has been a steady stream of new investigations of CO2RR with constantly new catalysts (Table 1) and a wide variety of production methods (Chapter 3.2). Overall, although the catalyst has a great impact on the selectivity and efficiency of the electrosynthesis, it is very difficult to compare catalysts due to large differences in electrode preparation, the test setup itself, and different electrolyte solutions, pressures, temperatures, etc. Here, standardized cells and reaction conditions could help to classify the potential of the catalysts in a reasonable way. In this area, there are already initiatives such as NFDI4Cat, which deals with the sharing of metadata in the entire field of catalysis and thus aims to create a research data infrastructure (Wulf et al., 2021). Regarding a potential industrial application, an additional focus should be on simplicity, scalability, and the lowest possible cost of production, as well as on long-term stability as numerous catalysts have been tested only for their capability in reducing CO2 for few minutes. In addition, more emphasis should be given to the use of flow cells or MEAs for testing the catalysts to achieve higher current densities. Finally, as already discussed mechanistic studies, for example, by using in situ methods and carrying out of operando studies, should be given greater emphasis.Important for the successful electrolysis of CO2 to valuable products is not only the choice of the appropriate catalyst but also suitable process conditions.One key parameter with a strong influence on catalyst/electrode performance is the electrolyte. For example, compared to KHCO3, a higher selectivity to carbonaceous products using KOH was shown. High local pH values, which can be favored by an electrolyte with low buffer capacity, have been shown to improve the product distribution toward higher hydrocarbons (Hori et al., 1989, 1997; Schouten et al., 2014; Varela et al., 2016; Xiao et al., 2016; Wang et al., 2018). Thus, alkaline electrolytes have been used in flow cells with promising results (Ma et al., 2016; Dinh et al., 2018a; García de Arquer et al., 2020).Owing to the competition between CO2 reduction and hydrogen evolution, alkaline conditions are required for an efficient performance of CO2 electrolysis (Pătru et al., 2019). In addition, the electrolyte used should be as conductive as possible in order to achieve higher energy efficiencies for the CO2RR (Dinh et al., 2018a). How much this affects the overall cell performance is shown by a comparison between 10 M KOH and 0.1 M KHCO3, according to which the ohmic losses in the formation of C2H4 were reduced by a factor of 47 under the highly alkaline conditions (Dinh et al., 2018a). Even when comparing 1 M KOH with 1 M or 0.1 M KHCO3, clear differences can already be seen. Although the same current densities can be achieved in principle with both electrolytes, the same current densities can be reached with 1 M KOH at considerably lower voltages, because the CO2RR activity is significantly higher there (Dinh et al., 2018b) — in a catholyte with a higher basicity, less energy is therefore required for the CO2RR (Xiang et al., 2019). In addition, the use of 1 M KOH also shifts the selectivity toward carbonaceous products (Dinh et al., 2018b; Lv et al., 2018a; Xiang et al., 2019). Thus, by changing from 1 M KHCO3 to 1 M KOH at an Ag/PTFE-GDE, instead of 80%, an FE of 90% for CO could be achieved (Dinh et al., 2018b). Furthermore, C2 products should be obtained mainly at KOH concentrations above 0.5 M (Xiang et al., 2019). An increase in FE for these was observed with a) more negative potentials and b) higher KOH concentrations. The current density was also significantly increased by a higher KOH concentration. Furthermore, OH groups in the vicinity of the catalyst surface should improve the reaction kinetics and, in the case of CuxO catalysts, stabilize the oxygen of the catalyst during the reduction reaction (Xiang et al., 2019). However, a recent study by Zhang and coworkers showed the opposite trend with a decrease in overall C2+ product formation (from 76.1%, 1 M KOH) and ethanol with increasing KOH concentration 7 M (60.4%, 7 M KOH) (Figure 8 (right)) (Duan et al., 2021). The authors explained this deviation from previous publications with the high carbonate formation due to the high current densities of 400 mA cm−2 used. The described dependence on KOH concentration was performed on a poly(ionic liquid)-based Cu0-CuI tandem catalyst and also shows a significant increase of C2+ products for using 1 M KOH instead of 1 M KHCO3 or 1 M KCl. While the formation of hydrogen decreases from 22.7% (KHCO3) to 6.6% (KOH), the FE for ethanol increases significantly (Figure 8 (left) (Duan et al., 2021)). However, there are also studies which do not only deal with the basicity and thus the OH− concentration, but focus on the cation of the electrolyte solution. Thus, there are also results that indicate that OH− is not the promoter of CO2 reduction. In this study, the concentration of Na+ and OH− was varied while keeping the other ionic content constant and the result was that the main supporting effect in the formation of C2+ products is caused by the sodium cation (Li et al., 2019a).However, there are also disadvantages of using basic electrolytes, such as the already mentioned instability of imidazolium-based ionomers in alkaline environments (Kutz et al., 2017), but also the required stability of the catalysts and GDE. For example, C-based GDEs degrade after about 2 h when using a basic electrolyte (Dinh et al., 2018b). Furthermore, the dilution of CO2 in basic electrolyte leads first to the creation of HCO3 −, followed by the conversion into CO3 2− (Leonard et al., 2020; Yang et al., 2020). This results in an indirect slowing down of the kinetics by initiating a shift of the pH value toward more neutral values and to the formation of barriers within the gas diffusion electrodes due to salinization, which in turn hinders the CO2 flow, promotes hydrogen formation and decreases the current density continuously (Endrődi et al., 2019; Yang et al., 2020). In addition, this storage of CO2 in the electrolyte can lead to an overestimation of the products FEs (Ma et al., 2020a), to conductivity losses within the system, as well as to energy efficiency losses in the overall electrolytic cell (Gabardo et al., 2019).Finally, it should be noted that the amount of electrolyte used also influences the CO2RR performance. If there are larger amounts of electrolyte between the membrane and cathode, kinetics of HER is suppressed and separation of the liquid products is simplified, but larger ohmic losses occur within the cell, leading to higher cell voltages at moderate current densities (Chen et al., 2020b).The local pH value has a strong impact on the product distribution, as a more alkaline environment promotes CO and multicarbon product formation and suppresses HER and CH4 formation (Burdyny and Smith, 2019). Thus, the local pH has an influence on the energetics of the different products of the CO2RR. It has been observed that the pH in weak buffering solutions, such as KHCO3 or KCl, at the electrode can be shifted up to 6 units in the beginning of the electrolysis. The large pH difference can also cause difficulties in determining the equilibrium potential between the working and reference electrode correctly, which in turn influences the onset potentials. Within the catalyst layer, pH values above 12 may occur at current densities >200 mA cm−2. The use of acidic electrolytes in CO2RR is often regarded to be no alternative as hydrogen formation would become too dominant (Burdyny and Smith, 2019). However, recently Sargent and coworkers have shown CO2RR at 1.2 A cm−2 in 1 M H3PO4 yielding 50% FE of C2+ products, which is possible due to the drop of local pH during operation (Huang et al., 2021). Regarding the F-Cu catalyst investigated by Wang and coworkers, a correlation between the local pH value and the catalyst could be found (Ma et al., 2020b). It was shown that the local pH value at the electrode increases significantly in the order K2HPO4 < K2CO3 < K2SO4 due to the high concentration of OH− produced during CO2RR, which cannot be buffered by electrolytes like K2SO4; however, the buffering of the pH value is better with, e. g. K2HPO4 (Lv et al., 2018a). At the same time, there is also an increase in C2+ formation in this order, which is more pronounced compared to the pure copper catalyst. In conclusion, it is however difficult to precisely estimate the extent of the pH influence on the catalyst (Ma et al., 2020b). A recent study by Jung and coworkers on Cu/Cu2O aerogel catalysts also shows that the use of electrolytes with lower buffer effect leads to higher FEs of ethanol at simultaneously lower FEs for HER (Kim et al., 2021). Thus, solvents with a higher buffering capacity should neutralize the OH− generated during CO2RR and thus oppose the local pH effects. Figure 9 shows the FEs of EtOH and H2 as a function of the selected electrolyte, with an increase observed for ethanol in the order K2HPO4 < KHCO3 < KClO4 < KCl. Studies on an electrode with electrodeposited copper showed that the local pH at the oxidized copper electrode decreases from 10.4 to 9.3 with increasing negative applied potential ranging between −0.4 and −1.2 V and using 1 M KOH (Henckel et al., 2021). The decrease in pH is due to the formation of HCO3 −, while at the same time, malachite is formed at the electrode at the beginning of the reduction of the copper oxide. Malachite shows highest thermodynamic stability between pH 8.0 to 10.5 and precipitates at the Cu surface of the electrode due to the carbonate-rich environment. These processes could influence the CO2RR product distribution. Thus, it should lead to higher Faraday efficiencies for the formation of ethylene than pure Cu foil. The subsequent further reduction of copper oxide and malachite finally leads again to a pH value of >11 (Henckel et al., 2021).Recently, the local pH for pulsed electrolysis at CO2RR in CsHCO3 and LiHCO3, respectively, was determined by simulations (Kim et al., 2020a). These show deviating values depending on the applied potential. In the period of the pulse of −0.8 V, a high concentration of CO2 is present at the cathode surface; the local pH value of nine is low. When the potential is raised to −1.15 V, the pH also increases to 11, and the CO2 concentration decreases from the previous 31 mM to13 mM. It is likely that this results in additionally increased adsorption of ∗CO compared to ∗H, which is accompanied by increased C2+ selectivity with a concomitant decrease in HER (Kim et al., 2020a).Studies on the most appropriate electrolyte also considered the optimal choice of cations. With respect to alkali metals, the following trend was found for the selective formation of CO and EtOH (Karapinar et al., 2019) and for the current densities (Gao et al., 2018): Li+ < Na+ < K+ < Cs+. The difference is also very clear when comparing the FEs for the formation of ethanol using different electrolyte solutions. The use of a Cu-N-C catalyst achieved an FE for EtOH of 2% in CO2RR at −1.2 V in LiHCO3 solution, but 42% in CsHCO3 (Karapinar et al., 2019). If larger cations are used, these are less strongly hydrated, facilitating the adsorption on the surface of the catalyst (Karapinar et al., 2019; Kibria et al., 2019). The adsorption of those cations leads to a more positive potential of the outer Helmholtz layer (OHP), which in turn reduces the H+ concentration at the electrode, consequently lowering the extent of HER (Karapinar et al., 2019; Kibria et al., 2019; Lamaison et al., 2020). In addition, hydrated Cs+ ions near the catalyst surface can buffer pH changes and increase the amount of locally dissolved CO2 (Kibria et al., 2019; Jeong et al., 2020). Furthermore, the hydrated Cs+ ions would impose an electric field on the external OHP and thus promote C-C bond formation by coupling adsorbed ∗CO with ∗HCO (Jeong et al., 2020). With regard to the anions used, the best selectivity has so far been shown for OH− in both CO and C2+ formation (Lv et al., 2018a; Kibria et al., 2019). Jiao and coworkers investigated the influence of the anions using KOH, KHCO3, KCl, and K2SO4 while keeping the K+ concentration constant (Lv et al., 2018a). The pH value in bulk was determined before and after electrolysis. It was found that KOH and KHCO3 showed hardly any changes in pH value in contrast to an increase of up to 4 pH units in the other two non-buffering electrolytes. The clearly best current densities for C2+ formation were obtained in KOH; in KCl and K2SO4, a high resistance was measured at higher overpotentials and a rapid overloading of the system occurred. The reason for the high resistance is probably the poor ionic conductivity of the membranes in these two electrolyte solutions (Lv et al., 2018a).As already described with respect to the modification of catalysts with halides, the use of chloride, bromide, and iodide exerts a significant influence on the cell performance. Not only modifying catalysts but also adding halides to the electrolyte leads to considerable changes. For example, the addition of KX salts (X = Cl−, Br− and I−) led to significantly increased current densities for the reduction of CO2 within an H-cell of plasma-activated copper catalysts in the order Cl− < Br− < I− (Ma et al., 2020b). The FE of the C2+ products remained unchanged; current densities and formation rate for the products increased with increasing electronegativity of the halides. Investigations with KI addition showed that the increased activity for CO2RR takes place by accelerated hydrogenation of adsorbed CO intermediates (Dinh et al., 2018a). A significant rise in current density was also observed when CsI was added to a CsHCO3 electrolyte solution (Gao et al., 2018). It is assumed that the iodide adsorbs and thereby increases the roughness of the catalyst by quasi I−-induced nanostructuring. Thus, Cu+ is also stabilized by the iodide. According to the previous discussion concerning the cations' choice on the CO2RR, it can be observed that CsI, due to its larger cation, supports the CO2RR more than the addition of KI (Gao et al., 2018).Because most studies currently focus on the formation of CO or C2+ products in general, it is necessary that further research on the influence of electrolyte solutions on the CO2RR to higher alcohols should be performed. Herein, preferential formation of C2+ products was reported frequently when using basic electrolyte solutions such as 1 M KOH and high local pH values. A disadvantage here, however, is the formation of carbonates in the GDE, which block diffusion pathways and facilitate HER. Contrary, recent results show that acidic electrolytes also have great potential for the formation of multicarbon products. A greater focus should therefore also be given to these systems, as it might be feasible to reduce carbonization effects in the GDE and enable a long-time stable system. In the context of zero-gap reactors, investigations should also be carried out using different solid electrolytes. In particular, materials should be found, which contribute to a low cell resistance, but at the same time are stable against the alcohols produced.Another impact which was investigated on CO2RR is that of temperature. When using H-type cells, where the availability of CO2 at the cathode depends on the solubility of this gas in the used electrolyte, lower temperatures have been shown to facilitate a higher ratio of CO2RR to HER due to the better availability of CO2 (Ahn et al., 2017). Palmore and coworkers investigated the influence of the temperature on the CO2 reduction on polycrystalline copper. They reported that the temperature affects various electrolyte parameters like CO2 solubility, pH, resistance of the solution, and diffusion rate of the reactants. While the FE for methane increased with lower temperatures and peaked at 2°C, ethylene FE increased with higher temperatures reaching its maximum at 22°C. Activity for HER rose with increasing temperatures (Ahn et al., 2017). The effect of increased ethylene FEs with simultaneously lower methane FEs at elevated temperatures has also been described by other authors (Hori et al., 1986; Cook, 1988; Kim et al., 1988).Besides H-type cells, investigation of temperature effects have been performed in flow cells. Klemm and coworkers researched the impact on the Sn catalyst-based CO2RR to formate in a liquid-phase flow cell. The optimum performance was observed at 50°C with over 80% formate FE at 1 A cm−2, while higher and lower temperatures led to increased HER (Figure 10 ). The increased HER at other temperatures than 50°C can be assigned due to the oppositional effects of reduced CO2 solubility and increased diffusion coefficients as well as faster reaction kinetics with increased temperature (Löwe et al., 2019). McIlwain and coworkers reported a reduction of the cell voltage by 1.57 V at 70 mA cm−2 during CO2RR to syngas, using a liquid-phase electrolyzer with an Ag-based GDE when the temperature is raised from room temperature to 70°C (Dufek et al., 2011). According to Sargent and coworkers, increasing the temperature to 60°C and the associated faster reduction kinetics and extended mass transport through the ionomer layer resulted in obtaining comparable Faraday efficiencies for C2+ products even at lower overpotentials (García de Arquer et al., 2020).In gas-phase electrolyzers, higher temperature can increase the electrochemical performance. According to Park and coworkers, reduction to formate on Sn nanoparticles increased more than 2-fold, when rising the temperature from 30°C to 90°C (Lee et al., 2018). Aricò and coworkers reported a significantly increased methanol production rate on a PtRu catalyst with higher temperatures, in a MEA-type setup, however, with overall low FEs (Sebastián et al., 2017). Sinton and coworkers also investigated the performance of their copper catalyst-based MEA electrolyzer at temperatures of 20°C, 40°C, and 60°C. An increase of temperature herein led to higher current densities for ethylene and hydrogen as well as higher FEs for the latter. Higher temperatures also increased the obtained ethanol output at the cathode side from 0.5 wt % at 20°C, peaking at 40°C with 2.3 wt %. This was attributed to an enhanced transport of water from anode to cathode side as well as increased vaporization of ethanol. The increased temperature was suggested to be a key factor in facilitating a highly concentrated output stream of liquid products (Gabardo et al., 2019). The influence of temperature on the MEA in terms of current density at different overpotentials as well as the described influence on ethanol yield is shown in Figure 11 . The maximum temperature in gas-phase electrolyzers is limited due to a high rate of water crossover and the performance of the membrane (Kibria et al., 2019).Initial studies have been carried out regarding temperature effects showing increased selectivity for ethanol at elevated temperatures. However, most studies addressing these reaction conditions deal with the formation of other products like formic acid or CO which means that there is still a lack on investigations showing the impact of various temperatures for a wider range of catalysts and electrodes for multicarbon alcohol production.Another important parameter is the pressure of CO2. For cell types using CO2 dissolved in the electrolyte, the solubility of CO2 is increased with rising pressure, which results in higher current densities. As early as 1995, Sakata and coworkers investigated various metals at an elevated pressure of 30.4 bar in an autoclave H-type cell. The use of Ag, Au, Zn, Pb, and In for the CO2RR results in the preferential formation of CO and formic acid at standard conditions, and an increase in pressure also provided higher resulting current densities due to reduced overpotentials. For metals in groups 8–10 such as Fe, Co, Rh, Ni, Pd, and Pt, applying a pressure of 30.4 bar resulted in a shift in selectivity from HER to the formation of CO and HCOOH (Hara et al., 1995). A possible explanation for the change in selectivity is the facilitated desorption of CO under higher CO2 pressure (Hori and Murata, 1990; Kudo et al., 1993; Kibria et al., 2019). Using a dendritic Ag-Zn catalyst in an H-type cell containing 0.1 M CsHCO3, Vlugt and coworkers achieved a stable FE of over 90% for 40 h at 10 mA cm−2 and about −1.0 V vs. RHE. Pressurized measurements were performed in a single-chamber cell at 200 mA cm−2. Raising the pressure from 1 to 3 bar resulted in an increase in the partial current density for CO formation from about 30 mA cm−2 to 131 mA cm−2 at −2.0 V vs. RHE. A further increase to 6 bar even provided partial current densities of 188 mA cm−2 at −1.2 V vs. RHE (Ramdin et al., 2019). Mul and coworkers used Cu-NPs and KHCO3 to evaluate the resulting FEs in CO2RR under variation of pressure between 1 and 9.1 bar. The pressure increase raised the FE of ethylene from 10.8% to 43.7% while lowering the FE(CH4) from 21.3% to 1.8% and decreasing the HER. Although a lower local pH was calculated for higher pressures, the increased ethylene selectivity is associated with increased surface coverage of CO, which also causes higher yields of CO under pressure (Kas et al., 2015). Sakata and coworkers varied the pressure for the study of Cu electrodes in an H-cell between 1 and 60.8 bar, with an initial shift in selectivity from HER to hydrocarbons formation. The maximum was obtained at 40.5 bar, and as the pressure was increased further, the product spectrum shifted further toward CO and HCOOH (Hara, 1994).Variation of pressure was also investigated in flow cells. So far, studies have mainly been performed on Ag-GDEs. Using these GDEs, a significant reduction in cell voltages was achieved by increasing the pressure (Hara, 1997; Dufek et al., 2012). McIlwain and coworkers combined the simultaneous increase of temperature and pressure (Dufek et al., 2012). Raising the temperature from 60°C to 90°C and increasing the pressure to 18.7 bar reduced the cell voltage from 4.01 V to below 3 V in the CO2RR to CO, with an FE(CO) of 82% (Dufek et al., 2012). Sinton and coworkers studied pressures from 1 to 7.1 bar using KOH electrolyte. The high pressure combined with 7 M KOH resulted in a low overpotential for reduction to CO of 300 mV at 300 mA cm−2 and an FE of almost 100%. Furthermore, the highest half-cell energy efficiency (EE) of 81.5% was achieved here compared with lower pressures (Gabardo et al., 2018). Recent studies on sputtered Ag-GDEs also show the effect of improved overall energy efficiency in CO2RR to CO of up to 67% at 202 mA cm−2, where the pressure was 50 bar and 5 M KOH was used (Edwards et al., 2020). In contrast, experiments by Schmid and coworkers on a silver-based GDE in a liquid flow cell showed no dependence of the FEs for CO when increasing the CO2 pressure from 0 to 25 bar. The CO2 pressure increase only caused a cell potential rise from 6 to 7 V, resulting in overall poorer energy efficiencies while no changes in FE were observable for CO formation (Krause et al., 2020).In addition to CO2RR under pressure using aqueous electrolyte solutions, studies have also been conducted on the reduction of CO2 from supercritical CO2. CO2 behaves as a supercritical fluid meeting the critical pressure and temperature of 73.8 bar and 31.0°C (Span and Wagner, 1996). In this state, CO2 has the density of a liquid but the viscosity of a gas and is infinitely miscible with other gases (Abbott and Eardley, 2000; Melchaeva et al., 2017). Battistel and coworkers studied CO2 reduction on Cu electrodes in supercritical CO2 using acetonitrile as cosolvent and tetrabutyl-ammonium hexafluorophosphate to increase conductivity. In addition, protic solvents of different pH values were added to allow the formation of hydrocarbons and to influence the selectivity. The use of water and 1 M CsHCO3 resulted in FEs of 11.1% for ethanol and 7.5% for methanol. However, the overall FE was limited to 34%, possibly due to reoxidation at the anode and other sources of loss. In addition, corrosion of the Cu electrodes was observed (Melchaeva et al., 2017). Our group recently showed that under high pressure conditions in supercritical CO2, suppression of HER from an FE of 60% to below 8% is possible. The resulting shift in product distribution compared to measurements under ambient conditions led to current efficiencies of up to 66% for the formation of formic acid (Junge Puring et al., 2020).Overall, the adjustment of process parameters could allow further optimization of electrochemical CO2 reduction, but the influence of temperature and pressure, especially for the reduction of CO2 to alcohols, has only been researched to a limited extent. In addition, aspects relevant for industrial implementation, such as long-term stability and integration into upstream and downstream processes, need to be evaluated.Finally, it can be stated that within the last few years, enormous efforts have been made regarding the design of catalysts for the electrochemical CO2RR to multicarbon products especially alcohols. The basis for these catalysts was almost exclusively copper-based systems. Overall, increasingly better Faraday efficiencies are being achieved for the formation of higher alcohols, with some exceeding 50% for C2+ alcohols (Karapinar et al., 2019; Chen et al., 2020a; Han et al., 2020a; Song et al., 2020; Wang et al., 2020b; Zhang et al., 2020a). Promising results were obtained by using alloys or bimetallic catalysts of Cu and, for example, Ag (Hoang et al., 2018; Li et al., 2019b; Dutta et al., 2020; Kim et al., 2020b; Martić et al., 2020; Zhang et al., 2020a), Zn (Ren et al., 2019; da Silva et al., 2020), or Pd (Rahaman et al., 2020). Likewise, catalysts with Cu and N-doped carbon showed encouraging results (Karapinar et al., 2019; Chen et al., 2020a; Han et al., 2020a; Wang et al., 2020b). These catalysts were prepared using different methods, of which sputtering (Li et al., 2019b, 2020; Wang et al., 2020b), electrodeposition (Dutta et al., 2020; Kim et al., 2020b; Rahaman et al., 2020; Kong et al., 2021), and precipitation conceivably followed by calcination, should be highlighted (Lv et al., 2018a; Zhou et al., 2018; Martić et al., 2019; Wei et al., 2020).However, the selectivity does not solely depend on the catalyst, but also on the overall system. To achieve industrially relevant current densities, it is necessary to use flow cells or cells utilizing membrane electrode assemblies. H-cells, in which the CO2 transport to the catalyst is largely determined by its solubility in the electrolyte, should therefore be avoided in the future, especially because some studies show large differences in the product distribution for the same catalyst occur between H-cell and flow cell (Kibria et al., 2018; Gregorio et al., 2020; Wang et al., 2020a). In addition to the cell design, the design of the electrodes themselves is also of great importance. For a more detailed consideration of these two points, reference is made to the second part of the review “Electrochemical CO 2 reduction - The macroscopic world of electrode design, reactor concepts & economic aspects”. There are few studies so far on the influence of temperature and pressure on the formation of multicarbon alcohols in CO2RR, but it was shown, for example by Sinton and coworkers, that a higher yield for EtOH could be obtained at a temperature of 40°C than at RT or at 60°C (Gabardo et al., 2019). Initial results are also available on the influence of pressure, even for using supercritical CO2 (Melchaeva et al., 2017).Owing to the many influences, for example, from the process conditions, but also from the design of the electrodes and cells, further development of catalysts in terms of their selectivity is indeed sensible, but the following points in particular should be given more attention: • Catalyst synthesis routes that are as simple as possible and do not involve particularly cost-intensive process parameters (such as high pressure) for insignificantly better selectivities (Siegmund et al., 2021) — better catalysts capable to reverse ethylene:ethanol selectivity are required • A particular focus should be placed on the development of further tandem catalysts, as these materials have already shown promising results in terms of selectivity to multicarbon alcohols • The targeted investigation and use of confinement effects, as already used for thermal catalysis (Mouarrawis et al., 2018) • Research on ethanol and propanol selectivity in context of temperature increase should be investigated in detail • Increasing the long-term testing and stability of the catalysts (for industrial implementation more than 1000 h tests are required (Masel et al., 2021)) • Reducing the cost of carbon capture by developing catalysts, electrodes, and cells that show good selectivities in terms of CO2RR even with lower CO2 concentrations • Developing new electrode and cell designs (including membrane development) that allow for more selective and energy-efficient CO2RR (further discussion see “Electrochemical CO 2 reduction - The macroscopic world of electrode design, reactor concepts & economic aspects”) • Less catalyst testing in H-cells, because achievable current densities below 100 mA cm−2 are not industrially relevant • Despite better CO2RR product distribution with regard to multicarbon products, turning away from KOH, because of the formation of carbonates and the oxidation of copper without applied potential—here more research regarding electrolyte influence on ethanol/propanol formation is essential • Moving away from the use of liquid electrolytes through the application of MEAs to realize lower cell voltages and counteract flooding of electrodes, thus enabling higher long-term stability and continuous CO2RR Catalyst synthesis routes that are as simple as possible and do not involve particularly cost-intensive process parameters (such as high pressure) for insignificantly better selectivities (Siegmund et al., 2021) — better catalysts capable to reverse ethylene:ethanol selectivity are requiredA particular focus should be placed on the development of further tandem catalysts, as these materials have already shown promising results in terms of selectivity to multicarbon alcoholsThe targeted investigation and use of confinement effects, as already used for thermal catalysis (Mouarrawis et al., 2018)Research on ethanol and propanol selectivity in context of temperature increase should be investigated in detailIncreasing the long-term testing and stability of the catalysts (for industrial implementation more than 1000 h tests are required (Masel et al., 2021))Reducing the cost of carbon capture by developing catalysts, electrodes, and cells that show good selectivities in terms of CO2RR even with lower CO2 concentrationsDeveloping new electrode and cell designs (including membrane development) that allow for more selective and energy-efficient CO2RR (further discussion see “Electrochemical CO 2 reduction - The macroscopic world of electrode design, reactor concepts & economic aspects”)Less catalyst testing in H-cells, because achievable current densities below 100 mA cm−2 are not industrially relevantDespite better CO2RR product distribution with regard to multicarbon products, turning away from KOH, because of the formation of carbonates and the oxidation of copper without applied potential—here more research regarding electrolyte influence on ethanol/propanol formation is essentialMoving away from the use of liquid electrolytes through the application of MEAs to realize lower cell voltages and counteract flooding of electrodes, thus enabling higher long-term stability and continuous CO2RRInvestigation of mechanistic understanding, i. e. use of in situ technologies and operando methods like Raman or IR spectroscopy to realize better catalyst design resulting in higher selectivity toward multicarbon alcohols as products in CO2RROwing to the enormous number of studies in the research field of CO2 reduction and the steadily increasing number of reports, it is not possible to know and cite every publication. It is pointed out that no author was specifically excluded. In order to give a comprehensive overview despite the high number of publications on the topic, two review parts have been written. This part deals especially with the catalysts/mechanisms/influences of the formation of higher alcohols during CO2RR; for other products please check other review papers.The authors are grateful for financial support from the German Federal Ministry for Economic Affairs and Energy (projects “ElkaSyn – Steigerung der Energieeffizienz der elektrokatalytischen Alkoholsynthese”, grant 03ET1642C, and “E4MeWi – Energie-Effiziente Erneuerbare-Energien basierte Methanol-Wirtschaft”, grant 03EI3035A-D). U.-P. A. is grateful for the financial support by the Deutsche Forschungsgemeinschaft (under Germany’s Excellence Strategy – EXC-2033 – Project number 390677874) and the Fraunhofer Internal Programs under Grant No. Attract 097-602175.Conceptualization: T.J., A.G., and D.S..; Investigation: T.J., A..G., and J.H.; Writing (Original Draft), T.J., A.G., D.S., and H.L., Writing (Review & Editing): T.J., A.G., D.S., H.L., U-P.A., and E.K.; Supervision: U-P.A. and E.K.The authors declare no competing interests.

Tackling climate change is one of the undoubtedly most important challenges at the present time. This review deals mainly with the chemical aspects of the current status for converting the greenhouse gas CO2 via electrochemical CO2 reduction reaction (CO2RR) to multicarbon alcohols as valuable products. Feasible reaction routes are presented, as well as catalyst synthesis methods such as electrodeposition, precipitation, or sputtering. In addition, a comprehensive overview of the currently achievable selectivities for multicarbon alcohols in CO2RR is given. It is also outlined to what extent, for example, modifications of the catalyst surfaces or the use of bifunctional compounds the product distribution is shifted. In addition, the influence of varying electrolyte, temperature, and pressure is described and discussed.

At present, the demand in olefin hydrocarbons steadily grows due to the rising production of polymers and chemical compounds on the basis thereof. Thermal and catalytic cracking of heavy hydrocarbons (mainly oil) and dehydrogenation of paraffins are the major industrial methods to manufacture the unsaturated hydrocarbons. The processes of catalytic dehydrogenation of hydrocarbons (STAR, Oleflex, BASF-Linde process, Catofin, FDB by Snamprogetti and Yarsintez) are among the priority processes in the petrochemical industry (Sattler et al., 2014), with the dehydrogenation of light paraffins with the oxidants (oxygen or carbon dioxide) over various catalysts being widely discussed in the literature (Fattahi et al., 2013, 2011; Darvishi et al., 2016; Bugrova et al., 2019).Various catalysts have been proposed for the dehydrogenation of light paraffins, including conventional catalysts based on deposited precious metals, primarily platinum (Long et al., 2014; Li et al., 2017; Zhou et al., 2017), chromium (Słoczyński et al., 2011; Li et al., 2016; Cheng et al., 2015), and vanadium (Rodemerck et al., 2016, 2017; Tian et al., 2016) oxides, and also the catalysts based on ordered mesoporous materials (Xu et al., 2013; Shee and Sayari, 2010), metal–organic frameworks (Zhao et al., 2013), activated carbon (Xu et al., 2014; Li et al., 2015), zirconia (Otroshchenko et al., 2015, 2016) and even bare alumina (Rodemerck et al., 2016). The γ-Al2O3-supported chromium oxide catalysts are widely used in industry (Ruettinger et al., 2010; Busygin et al., 2013) in the processes with fluidized bed of microspherical catalysts (Kataev et al., 2015; Gilmanov et al., 2015) and with fixed bed of palletized catalysts (Catofin). According to the patent (Fridman, 2012), the isobutane conversion of ~56% with the selectivity towards isobutylene of 92% may by achieved at ~540 °C when fresh CrOx/Al2O3 catalyst is used. The aged catalyst is characterized by the isobutane conversion of 48% and selectivity of 87.6% at ~565 °C.There are a number of drawbacks of fluidized-bed paraffin dehydrogenation process such as high toxicity of microspherical Cr-containing catalysts, equipment deterioration, high catalyst consumptionn, and relatively low alkene yields. Thus, the global tendency is connected with the dehydrogenation of paraffins in the fixed-bed reactor or reactors with the moving bed of catalyst as more effective and eco-friendly processes. The synthesis of such catalysts is complicated by the need to prepare the catalyst granules with high porosity, strength and homogeneous distribution of supported active components.Impregnation technique is mostly used to synthesize the supported catalysts due to simplicity and effectiveness. The γ-Al2O3 support granules with diameter of ~3 mm (1/8″) are impregnated with an aqueous solution containing the precursors of the active component and modifiers to prepare the CrOx/Al2O3 catalyst (Fridman, 2012; Ruettinger and Jacubinas, 2016; Salaeva et al., 2020). The CrO3 is used as a precursor of the active component in industry because of its high solubility (it allows introducing up to ~28 %wt. of Cr2O3 into the alumina support), the minimal amount of aggressive gases released during the catalyst calcination as well as the adsorption of chromate ions on the surface of γ-Al2O3 leading to the uniform distribution of the active component and its stabilization in a highly dispersed state (Spanos et al., 1994). The main disadvantage of CrO3 application is low pH of the impregnation solution (pH ≈ 0) because of high concentration of H2CrO4 therein that requires the use of special acid-resistant equipment. Aluminum oxide is also known to be dissolved in acids that implies additional requirements to the impregnation conditions.One of the approaches that makes it possible to shorten the contact time of the impregnation solution with the alumina support is the vacuum impregnation-drying method used in particular to produce microspherical CrOx/Al2O3 catalysts (Bekmukhamedov et al., 2016). The reduced pressure inside the pores of alumina support provides penetration of the impregnating solution in small pores that leads to more homogeneous distribution of the active components in catalyst granules. The impregnation under moderate vacuum may be taken into account for catalyst preparation, since it can be implemented in industry.The aim of this work was to reveal the influence of impregnation conditions (pressure during impregnation) of a molded alumina support on the properties of the produced alumina catalysts and their activity in the dehydrogenation of paraffin hydrocarbons with a fixed bed catalyst. A series of catalysts were synthesized with different pressures during the impregnation, studied by a complex of physical–chemical methods, and tested in the fixed-bed isobutane dehydrogenation.The alumina support was prepared using a thermochemically activated aluminum trihydroxide (TCA THA) that was extruded with a small amount of nitric acid and 2 wt% of wood flour as a porogen (Bugrova et al., 2016) into cylindrical granules with a diameter of ~3 mm and a length of 4–6 mm. The support granules were dried at 120 °C and calcined at 700 °C for 4 h. The temperatures of drying and calcination were previously optimized taking into account the phase transformation of alumina hydroxides and boehmite into γ-Al2O3 without the transformation into δ- or α-Al2O3 with lower specific surface area (Zykova et al., 2015). The conditions were close to those used to prepare the industrial CrOx/Al2O3 catalysts (Fridman, 2012).To study the stability of the alumina support towards impregnating solution, a model impregnating solution with pH≈0 containing 2.29 %wt. of chromium and 0.35 %wt. of potassium were prepared by dissolving the corresponding amounts of CrO3 (chemically pure, Vekton, Russia) and KNO3 (chemically pure, Vekton, Russia). The support granules were placed into the model impregnation solution, then the probe of solution (~100 μl) was taken every 5–7 min to determine the content of Cr, K, and Al by atomic emission spectroscopy (“Agilent 4100″ spectrometer with microwave plasma).Chromia-containing catalysts were prepared by impregnation using an aqueous solution of precursors of the active component (CrO3, chemically pure) and alkaline modifier (KNO3, chemically pure). The excess of impregnating solution (15.2 ml) was prepared by dissolution of CrO3 (10.3 g) and KNO3 (1.7 g) in distilled water. The loadings of the components in the catalyst were 20 %wt. Cr2O3 and 2 %wt. K2O, which was close to those in industrial catalysts (Li et al., 2015; Otroshchenko et al., 2016; Fridman and Urbancic, 2015). Thealumina support granules were dried at 150 °C overnight to remove the moisture. Then the granules (10 g) were put into the three-neck flask equipped with the vacuum pump and dropping funnel with a pressure compensator. The support granules were degassed at 1.0, 0.85 and 0.70 atm for 20 min, then the excess of the impregnating solution was added from the dropping funnel at pressures of 1.0, 0.85, and 0.7 atm. After that, the flask was opened accurately, and the pressure was normalized to the atmospheric one. The excess of impregnating solution was drained. The impregnated granules were immediately dried at 95 °C and calcined at 400 °C for 2 h.Chemical analysis of the samples was carried out by dissolving them in a mixture of sulfuric and nitric acids and analyzing the resulting solution by atomic emission spectroscopy (AES) using the “Agilent 4100″ spectrometer with microwave plasma. The porous structure of the samples was carried out by low-temperature (77 K) nitrogen adsorption using the ”Tristar 3020″ analyzer (Micromeritics, USA). The determination of the specific surface area (SBET) was carried out using the multipoint BET method to straighten the nitrogen adsorption isotherm in the range of relative pressures p/po from 0.05 to 0.30. The pore size distribution was obtained using the BJH-desorption method analyzing the desorption branch of nitrogen adsorption–desorption isotherm. The Hg intrusion porosimetry measurements were carried out using the Poremaster-33 (Quantachrome, USA).The phase composition of the synthesized catalysts was studied by powder X-ray diffraction (XRD) using the Shimadzu XRD 6000 diffractometer with CuKα radiation and a Ni-filter. The diffraction peaks of the crystalline phases were processed using the POWDER CELL 2.4 software and compared with those peaks of standard compounds from the PCPDFWIN database. The sizes of the crystallites of the metal oxides were calculated using the Scherrer equation. The scanning electron microscopy (SEM) was used to characterize the broken granules. The SEM 515 (Philips) microscope with the accelerating voltage of 30 kV was used. The features of the sample reduction were studied by the temperature-programmed reduction (TPR-H2). The experiments were carried out on the chemisorption analyzer ChemiSorb 2750 (Micromeritics, USA) for as-prepared samples using a 10% H2/Ar gas mixture at a flow rate of 20 ml/min and a heating rate of 10 deg/min.Catalytic properties of the samples were studied through the isobutane dehydrogenation. The experiments were carried out on the “Katakon” flow catalytic unit (Katakon, Russia) in a tubular metal reactor with a stationary catalyst bed at temperatures of 570, 590, 610 °C. A reaction mixture of 15% i-C4H10 and the balance of Ar was fed through a catalyst bed (10 cm3 of catalyst pellets, 8.5–9.0 g) at a rate of 25.2 l/h. The experiment was carried out in a cyclic mode: reduction (H2/Ar mixture, 3 min), dehydrogenation (isobutene, 9 min), regeneration (atmosphere-air, 9 min). The Ar flow (3 min) was fed through the catalysts between each step. The gas probe for analysis was taken at 7th min of the dehydrogenation step. Analysis of the reaction mixture and reaction products was carried out using the gas chromatograph “Khromos GKh-1000” (Khromos, Russia) with a flame ionization detector and two microcatharometers. The products were separated at 50 °C using the quartz capillary column with poly(trimethylsilyl)propene (PTMSP), a packed column with Chromosorb 106 (60/80 mesh) and a packed column with NaX molecular sieves (45/60 mesh). The quantitative calculation of the volume fraction of the components of the gas mixture was determined using the Khromos software 2.16.43. Before the catalytic test, the experiment with quartz balls was carried out. The isobutane conversion was 3–5% at 570–610 °C that indicates the negligible insignificant influence of the inert balls, reactor walls and homogeneous reaction on the catalytic results.To study the stability of the alumina support in contact with the impregnating solution, the support granules (particles with sizes of 0.5–1 mm) were placed in a model impregnating solution containing 2.35 %wt. of chromium (as CrO4 2− ions) and 0.35 %wt. of potassium (as K+ ions) with pH ≈ 0. Fig. 1 a shows the concentration dependences for chromium, potassium and aluminum in the solution as a function of contact time of alumina granules with the solution. The Cr concentration in the solution drops sharply during the first five minutes of impregnation from 2.29 %wt. to ~1.7 %wt. indicating the sorption of negatively charged chromate ions (CrO4 2−) on the alumina support surface positively charged at pH ~ 0. During the next 40 min, the Cr concentration in the solution changes insignificantly indicating the achievement of the adsorption–desorption equilibrium. The K concentration in the impregnating solution does not change indicating the absence of adsorption of positively charged potassium ions on the positively charged alumina surface.It is noteworthy that the Al concentration in the model impregnating solution grows rapidly enough indicating the dissolution of alumina support in acidic conditions under the action of a chemically aggressive impregnating solution. At 50 min of granule contact with the solution, the Al concentration in the solution increases up to 0.274 %wt., which corresponds to ~0.8 % support dissolution. The support dissolution in the impregnating solution is a negative effect leading to the formation of mixed chromium-aluminum oxides inactive in the dehydrogenation reaction (Nemykina et al., 2010), decreased specific surface area (Bugrova and Mamontov, 2016), etc. The aluminum presence in excess of the impregnating solution limits the reuse of the latter for impregnating the next support batch under industrial conditions (Mamontov et al., 2017).Thus, the contact of the impregnating solution containing chromic acid with the alumina support granules was shown to lead to rather rapid alumina dissolution. Reducing of impregnation time is essential to minimize this negative effect. On the other hand, time of alumina granule impregnation should be enough for the impregnating solution to penetrate inside the granules providing homogeneous distribution of active component and alkali modifier. Fig. 1b shows the intact and broken granules of alumina-chromia catalysts prepared by impregnation at atmospheric pressure during 5 and 15 min. The gradient of color from dark-green to yellow and white across the breaks of granules (“egg-shell” structure) is observed after 5 min of alumina support contact with the impregnating solution. This means that 5 min is not enough for homogeneous distribution of active component in the catalyst granules. Impregnation of alumina support during 15 min leads to more homogeneous distribution of active component across the granule diameter (absence of color gradient at the granule break, “whole-egg” structure). Thus, the impregnation time should be optimized to achieve both introduction of active component and minimize the support dissolving. The granule impregnation at reduced pressure may be used to increase the solution penetration rate inside the support pores and minimize the contact time with the solution. This technique may be implemented in industry only in the case of insignificant pressure reduction to 0.5–0.7 atm (Ertl et al., 2008).Three alumina-chromia catalysts were prepared by varying the pressure during the impregnation (1, 0.85 and 0.7 atm). Table 1 shows the chemical analysis data and textural characteristics of the prepared catalysts. The contents of potassium and chromium oxides in all catalysts are close to the given values ​​(2 %wt. and 20 %wt., respectively). The observed overestimation of the chromium oxide content by about 1%wt. (the nominal loading of 20 wt%) may be a consequence of the Cr precursor sorption on the support during the impregnation from the excess of the impregnating solution. The Cr sorption is observed during the alumina support contact with the model impregnating solution (Fig. 1a). Thus, from the data obtained, it can be concluded that the pressure change during the impregnation does not significantly influence on the catalyst chemical composition. Fig. 2 shows the SEM images of the catalyst granules (broken immediately before the SEM studies). It can be seen from the SEM images that catalyst structure feature roughness that results from the packing of alumina powder particles during the extrusion. Wide pores with sizes of 10–50 μm and pores with size of from few hundreds nm to few micrometers are observed.The presence of these pores is important for both the penetration of the impregnating solution during the catalyst preparation and for effective reagent transport during the high temperature catalytic process. The size of primary alumina particles constituting the catalyst structure ranges from several micrometers to ~20 μm (Fig. 2b). The space between these primary particles provides high volume of macropores, and some of them cannot be detected by the SEM. Fig. 3 shows the N2 adsorption–desorption isotherms and pore size distributions for alumina support and the catalysts on the basis thereof. The support is characterized by a mesoporous structure that is evidenced by the presence of a hysteresis loop in the region of relative pressures of 0.5–1.0 on the N2 adsorption–desorption isotherms. The pore size distribution resides in the region from 2 to 20 nm with the distribution maximum at ~5 nm. The support specific surface area and the pore volume are 139 m2/g and 0.350 cm3/g, respectively (Table 1). The strength of alumina support granules is calculated as P = F/D*h, where F is a force of granule breaking, D is a granule diameter, and h is a granule length. The measurement for 30 granules show the strength of 8.7 ± 0.5 MPa that is sufficient for industrial catalysts. Thus, the presence of macropores shown by SEM does not decrease the granule strength.The catalysts are characterized by the decreased specific surface area (56–91 m2/g) and pore volume (0.166–0.225 cm3/g). It is clearly seen from Table 1 and Fig. 3a that the pressure decreasing during the impregnation leads to consistently reduced pore volume and specific surface area without significant changes of chemical composition (Table 1). A decreased pore volume throughout the pore size range for the catalysts (Fig. 3b) indicates the active component distribution in the support pores. The shifting of the pore size distribution maximum for Cr/Al2O3-0.7 catalyst towards smaller sizes can be caused by coarsening of the CrOx particles in mesopores with sizes of 5–20 nm. The pore size distribution in the range of 2–3.5 nm for all samples is practically the same, which may indicate the difficulties in the active component penetration into the fine support pores, even if vacuum impregnation is used.To study the porosity, the Hg intrusion porosimetry was applied for Cr/Al2O3-1.0 catalyst. The differential pore size distribution (Fig. 3c) features two types of pores, namely, mesopores with diameter from few nm to ~40 nm and wide macropores with sizes of from 50 nm to ~8 μm with a maximum at ~200 nm. The presence of macropores may be caused by using the wood flour as a porogen during the granule extrusion. These results are consistent with the SEM data. Thus, the catalyst granules are characterized by the hierarchical porous structure (meso- and macropores).The XRD patterns for the catalysts and the results of qualitative and quantitative analysis of XRD are shown in Fig. 4 and Table 2 , respectively. All catalysts feature high content of amorphous phase (68.1–71.5%). The support constitutes the γ-Al2O3 and amorphous phases. The active component (CrOx) exists in the amorphous state, with a small part comprising α-Cr2O3 phase. The content of the α-Cr2O3 phase increases from 3.4 %wt. to 5.3 %wt. with a pressure decrease during the impregnation from 1.0 to 0.7 atm. The α-Cr2O3 particle size (coherent scattering region) increases from 7.0 to 15.3 and 20.4 nm for pressure of impregnation of 1.0, 0.85 and 0.7 atm, respectively. This may explain the decreased specific surface area of the catalysts obtained under reduced pressure due to the partial blocking of the alumina support pores by chromia particles. Fig. 5 shows the TPR results for the obtained samples. All catalysts are characterized by the hydrogen consumption in the temperature range from 200 °C to 450 °C associated with the Cr(VI) reduction to Cr(III) (Bugrova et al., 2019). Two peaks of hydrogen consumption are observed for Cr/Al2O3-1 catalyst indicating coexistence of two Cr(VI) states. The first peak with a maximum at 320 °C can be attributed to the reduction of monomeric and/or oligomeric forms of Cr (VI) (Fridman et al., 2016). In our previous work (Salaeva et al., 2020), using Raman spectroscopy, we showed the key roles of monomeric and dimeric chromia species in isobutane dehydrogenation. A high-temperature peak at 420 °C may be attributed to reduction of small particles of Cr (VI) oxide or potassium chromates (Neri et al., 2004; Rombi et al., 2003). The shifting of TPR peak from 320 °C to 360 °C may be caused by an increase in the particle size of the chromium (VI) compounds or increased interaction of chromia species with alumina, which both lead to the decreasing of the catalytic activity (Salaeva et al., 2019).Thus, from the TPR results it can be concluded that aside from α-Cr2O3 detected by XRD, the catalysts contain the Cr(VI) species that are reduced at 200–450 °C. The reductive pretreatment of the catalysts before the catalytic experiments leads to Cr(VI) reduction into Cr(III) sites (Bugrova and Mamontov, 2018). The real state of the active catalyst surfaces during the dehydrogenation is presented by two types of Cr(III) species: the first one is Cr(III) found in the as-prepared catalysts (including α-Cr2O3 detected by XRD) and Cr(III) formed due to the Cr(VI) species reduction. The activity of these species depends on both their dispersion and distribution on the catalysts surface. According to N2 physisorption and TPR results, the chromia species distribution is more homogeneous for the catalyst prepared at atmospheric pressure. Thus, the use of vacuum impregnation allows obtaining the catalysts with high content of Cr(VI) species, but with the decreased dispersion of these species. Consequently, the decreased Cr(VI) dispersion may be the reason for reduced catalytic activity.The catalytic properties of the prepared catalysts were studied in isobutane dehydrogenation in a fixed-bed reactor (Fig. 6 ). The real catalyst granules were tested under conditions close to the industrial ones to show the real opportunity for catalyst application in the process. The isobutane conversion growth is observed for all catalyst as the temperature increases from 570 °C to 610 °C with the corresponding decreasing of isobutylene selectivity. The conversion growth in this temperature region indicates the minimal diffusion limitations that are attributed to the presence of macropores in the catalyst granules shown by the SEM. The catalyst synthesized by the impregnation at atmospheric pressure is characterized by higher catalytic activity. Thus, the isobutylene yield is ~70% (71.4% conversion and 95.6% selectivity) at 610 °C under the process conditions (15% i-C4H10 in Ar). These results are close to the performance of industrial CrOx/Al2O3 catalysts tested under similar conditions (Xing and Fridman, 2019).The conversion for the Cr/Al2O3–0.7 catalyst synthesized by vacuum impregnation at 0.7 atm. is lower by 11–21 %mol. compared to the sample impregnated at 1.0 atm. The isobutylene selectivity for Cr/Al2O3–0.7 is also slightly lower by 0.2–1.3 %. Thus, a regular decreasing in conversion and selectivity is observed for a number of catalysts prepared by impregnation at 1.0, 0.85 and 0.7 atm.To study the stability, the Cr/Al2O3-1.0 catalyst was tested during 25 catalytic cycles including oxidative regeneration and reductive activation between the dehydrogenation. It can be seen from Fig. 6b that relatively high stability is observed. The isobutane conversion is kept at 52–62% at 590 °C at a selectivity of 96–98%. This indicates the stability of active catalyst surface during the high-temperature oxidative-reductive treatments and relatively low amount of coke formed. The amount of coke was measured by TGA-DSC for Cr/Al2O3-1.0 catalysts after 3 catalytic cycles and cooling in inert atmosphere. The amount of coke was 1.26 %wt. after the cycle at 610 °C. This value is not high and the coke formation is an important process in the isobutane dehydrogenation in a fixed-bed reactor. The coke burning during the oxidative treatment leads to the catalyst bed overheating and this heat is used in the endothermic dehydrogenation process.Thus, the analysis of results of catalytic and physical–chemical studies shows that for the catalysts prepared at 1.0, 0.85, and 0.7 atm, the following regularities are observed: – a decreasing of specific surface area and pore volume due to partial blocking of pores by chromia, – an increased amount of α-Cr2O3 with the growth of the particle size for this phase that is undesirable because the α-Cr2O3 phase is characterized by very low dehydrogenation activity, – an increased temperature of Cr(VI) reduction indicating the Cr(VI) species agglomeration or enhanced chromia-support interaction that is also undesirable and leads to the decreased activity of chromia species, – the decreased activity in a row Cr/Al2O3-1.0 > Cr/Al2O3-0.85 > Cr/Al2O3-0.7. a decreasing of specific surface area and pore volume due to partial blocking of pores by chromia,an increased amount of α-Cr2O3 with the growth of the particle size for this phase that is undesirable because the α-Cr2O3 phase is characterized by very low dehydrogenation activity,an increased temperature of Cr(VI) reduction indicating the Cr(VI) species agglomeration or enhanced chromia-support interaction that is also undesirable and leads to the decreased activity of chromia species,the decreased activity in a row Cr/Al2O3-1.0 > Cr/Al2O3-0.85 > Cr/Al2O3-0.7.Thus, the preparation of CrOx/Al2O3 catalysts is accompanied by many challenges: dissolving of alumina support in acidic impregnating solution, limited penetration of the impregnating solution inside the alumina support granules, etc. The decrease in the pressure during the impregnation of the granules of alumina support is a promising way to decrease the time of contact of support granules with the impregnating solution and enhance the penetration of impregnating solution inside the granules because of pressure change. However, the impregnation under vacuum leads to blocking of the support pores by chromia that is confirmed by the decreased SBET and pore volume. A reduced catalytic activity may be a result of both reduction of active catalyst surface and increased amount of inactive α-Cr2O3 phase.It is known that the dehydrogenation of light paraffins at 570–610 °C may be hindered by the internal diffusion of reagent to the active site and product elimination from the catalyst pores (Barghi et al., 2012,2014). A porous structure of the catalyst plays an important role in high-temperature processes (Lee and Kim, 2013). Wide mesopores and macropores provide isobutane diffusion towards active sites on the catalysts surface. Therewith the reaction products (isobutene and hydrogen) should be released from the active surface to prevent the hydrogenation reaction. Wide mesopores provide product transport from the catalyst granules. Blocking of the mesopores by chromia nanoparticles may be a reason for additional diffusion limitation and decreased activity of the catalysts prepared by vacuum impregnation. Besides, the Cr/Al2O3-1.0 catalyst prepared at atmospheric pressure is characterized by both relatively high surface area and the presence of macropores that was demonstrated by SEM and Hg intrusion porosimetry. Thus, the combination of macropores and relatively high surface area for this catalyst is favorable for relatively homogeneous distribution of the active component inside the alumina granules (“whole-egg” structure, Fig. 1b), high activity and stability in the isobutane dehydrogenation.Thus, it was shown that the production of alumina-chromia catalysts is significantly limited by dissolution of the alumina support by impregnating solution containing chromic acid. Optimization of the impregnation conditions is required to synthesize highly active CrOx/Al2O3 catalysts for dehydrogenation of light paraffins in a fixed-bed reactor. On the one hand, the time of impregnation should be minimized to prevent the alumina support dissolution by the impregnating solution containing chromic acid. On the other hand, the time and pressure of impregnation should provide penetration of the impregnating solution inside the alumina granules with a diameter of ~3 mm for the homogeneous distribution of active component and modifiers on the catalysts surface. The vacuum impregnation allowed us to introduce the impregnating solution rapidly, but it led to the decreased catalyst porosity as well as decreased activity in isobutane dehydrogenation. The role of macropores in the active component distribution during the impregnation and the role of macropores in minimization of diffusion limitations was demonstrated.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 carried out within the framework of the state assignment of the Ministry of Science and Higher Education of the Russian Federation (project No. 0721-2020-0037). Author thanks Mrs. Elena Blokhina (Tomsk State University) for AES studies.

The development of catalysts for dehydrogenation of light paraffin hydrocarbons in a fixed bed reactor is of great importance for the world petrochemical industry. The preparation of granules (~3 mm in diameter) of CrOx/Al2O3 catalysts is hindered by such problems as homogeneous distribution of active component and modifiers, high strength of granules, etc. In this paper, the alumina support dissolution in the impregnating solution containing chromic acid and the opportunity to apply vacuum impregnation to minimize this effect in the preparation of CrOx/Al2O3 catalysts are discussed. A series of catalysts is synthesized at different impregnation pressures (1, 0.85, and 0.7 atm), characterized by a complex of physical–chemical methods (low-temperature N2 adsorption, SEM, XRD, TPR-H2), and tested in isobutane dehydrogenation. The use of vacuum impregnation is shown to lead to the reduction of the specific surface area of the catalysts from to 91 to 56 m2/g and the growth of content of CrOx phases that decreases the catalytic activity in dehydrogenation. The isobutylene yield at 610 °C decreases from 68% to 54% for the catalyst prepared at P = 0.7 atm as compared with the one prepared at atmospheric pressure. The high activity and stability are connected with the hierarchical structure of the alumina support and homogeneous chromia distribution on its surface.

The progress of modern industrialization is accompanied by the massive consumption of fossil fuels, particularly oil and coal, leading to the shortage of these unrenewable resources and emission of large amounts of carbon dioxide (CO2) [1–3]. Recently, CO2 concentrations in the atmosphere have exceeded 415 ppm, which is about 50% higher than pre-industrial levels. Induced by massive CO2 emissions, global warming has seriously threatened the balance of natural ecosystems [4]. Similar to CO2, global emissions of methane (CH4), which is another major greenhouse gas [5], have increased by nearly 10% over the past two decades, and its atmospheric concentration has set a new record of 1.875 ppm [6,7]. Given the warming power of CH4 is 80 times as high as that of CO2, CH4 is believed as the second most prevalent greenhouse gas from human activities. Subsequently, how to deal with these greenhouse gasses is an urgent problem to be solved [6,8,9]. Conversion of two major greenhouse gasses into value-added syngas ( C H 4 + C O 2 → 2 C O + 2 H 2 Δ H 298 K = 247 k J / m o l ) , which is also called CO2 reduction by methane (CRM), is considered to be one of the most promising approaches to achieve sustainable development [10].Different metal catalysts have been employed as active components for CRM reaction, such as Fe, Co, Ni, Ru, Rh and so on. Noble metals have high catalytic activity, but their high costs and limited availability prevent their practical large-scale applications. Ni has become the most widely used catalyst for its comparably high activity and low cost, but it suffers from deactivation due to carbon deposition and sintering of Ni nanoparticles (NPs). Carbon deposition mainly comes from the two side reactions, i.e., of methane dissociation ( C H 4 → C + 2 H 2 Δ H 298 K = 75 k J / m o l ) and carbon monoxide disproportionation ( 2 C O → C + C O 2 Δ H = − 172 k J / m o l ) [11,12]. Reducing the size of NPs has been demonstrated to limit carbon nucleation and growth, but nanoparticles tend to aggregate into large particles at high reaction temperature, leading to poor stability. Confining metal NPs inside mesoporous materials, such as porous shells and matrixes, has been reported to be effective in mitigating sintering [13,14]. Nevertheless, some active sites are inevitably covered and inhibited to interact with reaction gasses, leading to decreasing CRM activities. Therefore, enabling metal NPs to possess small sizes, high activity and high stability simultaneously is still a daunting challenge.Another concern of CRM is its highly endothermic nature, so that massive thermal energy is needed to drive reactions. Emerging solar-driven CRM not only supplies thermal energy required in a low carbon way, but also can store solar energy in the form of important feedstocks or fuels [15–18], thus serving as a promising candidate to tackle global energy and climate change problems simultaneously. The key parameter determining whether solar-driven CRM can be widely deployed is high solar-to-fuel efficiency. People have devoted extensive efforts to enhancing solar-to-fuel efficiency. For example, many different catalysts have been developed to improve the catalytic activity and efficiency [19–26], and the highest solar-to-fuel efficiency reported under mild conditions is only 33.8%. Further enhancing solar-to-fuel efficiency over 35% is still a desire.Here, we proposed interconnected Ni/MgAlO x nanoflakes grown on SiO2 particles to achieve highly efficient solar-driven CO2-to-fuel conversion (shown in Scheme 1 ). An extremely large light-to-fuel efficiency of 35.7% and very high fuel production rates of H2 and CO (136.6 and 148.2 mmol min−1 g − 1) are achieved under focused illumination. Excellent spatial confinement of active sites, strong metal-support interactions, improved CO2 absorption and activation, and decreased apparent activation energy of C* and CH* species under direct light illumination are considered the main mechanisms, as confirmed by both experimental measurements and DFT calculations. In addition, the lattice oxygen of MgAlO x in the nanocomposite takes part in the reaction which helps to decrease carbon species as will be discussed later.Mg(NO3)2·6H2O (0.164 g, 0.64 mmol), Al(NO3)3·9H2O (0.135 g, 0.36 mmol), SiO2 (0.12 g) and CO(NH2)2 (2.7 g) were dispersed in 8 ml deionized water. 9 ml C2H5OH and 8 ml Ni(NO3)2·6H2O (0.1 M) were then added. After stirring for five hours, the solution was dropped in a Teflon-lined stainless steel autoclave heating for 36 h at 190 °C. The resulting suspension was centrifuged and washed three times with ethanol, and then the product was dried overnight. Finally, the powders were reduced under 10% H2/Ar at 700 °C for 3 h. The reduced sample was labeled as Ni/Mg1.78AlOx@SiO2. The samples with the Mg/Al molar ratio of 0.67, 1.22 and 2.03 were prepared after following similar procedures. Correspondingly, 0.26 g Mg(NO3)2·6H2O (or 0.37 g Al(NO3)3·9H2O), 0.12 g SiO2 and 2.7 g CO(NH2)2 were employed to obtain Ni/MgO@SiO2 or Ni/Al2O3@SiO2. The procedures were the same as Ni/Mg1.78AlOx@SiO2. Ni@SiO2 sample was synthesized via the same procedures except that Mg(NO3)2·6H2O was not added.The CO2-to-fuel conversion was conducted in a homemade reactor with a quartz window. We put 0.019 g of catalysts in the reactor for every test. A stream of CH4/CO2/N2 (43.2%/43.2%/13.6%) was continuously fed to the reactor at 104.2 ml min−1. A 300 W Xe lamp was used as the light source without using any other heating devices. The irradiation power focused on samples was measured by a laser power meter, which was calibrated by AM 1.5 global solar light with a standard Si solar cell. The power of the focused UV–Vis-IR illumination is measured to be 12.0 W. Since the spot diameter is 6 mm, the irradiation density reaches 424.6 kW m − 2.The light-to-fuel efficiency(η) is defined as follows η = ( r H 2 × Δ c H H 2 0 + r C O × Δ c H C O 0 − r C H 4 × Δ c H C H 4 0 ) P irradiation where r H2 and r CO are the molar production rate of H2 and CO, respectively, and r CH4 is the reaction rate of CH4. Δ c H CO 2 0 , Δ c H CO 0 and Δ c H CH 4 0 are the standard heat of combustion ( Δ c H 0 , 298.15 K) of H2, CO and CH4 fuel, respectively (note: CO2 is not a fuel, so Δ c H CO 2 0 of CO2 is 0), and P irradiation is the irradiation power focused on the reactor.The structure and components of prepared catalysts were investigated by X-ray diffraction (XRD). MgO and Al2O3 are in the form of Mg-phyllosilicate Mg3Si4O10(OH)2 (PDF 29–1493) and Al-phyllosilicate Al2Si2O5(OH)4 (PDF 14–0164), respectively, both of which are at the same peak (2θ =42.6°) in Fig. 1 a. When MgO is combined with Al2O3, it exists in the form of MgAl2Si2O6(OH)4 (PDF 35–0489). This fully shows that there is a strong interaction between magnesium oxide, aluminum oxide and silicon oxide [27]. All the samples show three diffraction peaks at 2θ of 44.5°, 51.8°and 76.3°, which correspond to the (111), (200) and (220) crystal plane of Ni [28,29], respectively. As shown in Fig. 1b, there is one more peak at 26.619° belonging to MgAl2Si2O6(OH)4 in Ni/Mg1.78AlO x @SiO2, suggesting that different Mg/Al molar ratios affect the combination of MgO, Al2O3 and SiO2. XRD results confirm that Mg2+ and Al3+ exist in the form of phyllosilicates instead of bulk MgO and Al2O3, illustrating strong interactions between Mg-Al phyllosilicates and the support SiO2.The surface morphology and element mapping analysis of the catalysts were investigated by a transmission electron microscope (TEM) and scanning electron microscopy (SEM). TEM images (Fig. 1c) show that the structure of the Ni/Mg1.78AlO x @SiO2 catalyst is similar to a sphere. Internal pores are beneficial to enhance the surface area and limit the agglomeration of metal particles. According to N2 adsorption/desorption measurements (Fig. S1), the surface area of Ni/Mg1.78AlO x @SiO2 is 112.5 cm2 (Table 1 ). Although this value is not the largest among different samples, the average particle size of Ni nanoparticles before the reaction is the smallest with a value of only 8.65 nm (Fig. S3). Small Ni nanoparticles can suppress coking since carbon nanofibers are more difficult to nucleate, not to mention subsequent growth [28,30]. Ni nanoparticles are shown as bright dots in Fig. 1d, and their interplanar distance is 0.208 nm as shown in Fig. 1e [31]. Due to the addition of excessive urea, the solution is alkaline, so SiO2 particles partially dissolve and form Mg-Al phyllosilicate with Mg2+ and Al3+ (Fig. 1f). MgAl-phyllosilicate crystal lattice effectively suppressed the growth and migration of Ni nanoparticles. Through the corresponding element mapping of Ni, Mg, Al, Si, O (Fig. 1g), all the elements are dispersed uniformly on the surface of the carrier. Good dispersion and small size of Ni nanoparticles help to achieve both good catalytic activity and durability.It is well known that the metal-support interaction is an important factor affecting CRM performances [32]. Here, H2-temperature programmed reduction (H2-TPR) experiments of the samples were conducted to check their metal-support interactions (Fig. 2 a). There is a broad peak between 400 °C and 600 °C for each catalyst, which is corresponding to the medium interaction between Ni species and the support [29,33]. It is worth noting that catalysts containing Mg-Al supports have a narrow reduction peak over 600 °C, which indicates that a small part of NiO has strong interactions with the support. An additional peak centered on 732.4 °C is found for Ni/Mg1.78AlO x @SiO2, which suggests even stronger metal-support interaction [13]. A higher reduction temperature means the sintering resistance of metal nanoparticles is better. That also indicates the combination of magnesium and aluminum effectively increases metal-support interactions, which contributes to the small sizes of Ni nanoparticles (Fig. S3).The basic sites of catalyst surfaces have a great influence on the adsorption and dissociation of CO2 [34,35]. Strong alkalinity can effectively promote CO2 adsorption and dissociation [36]. As shown in Fig. 2b, samples exhibited a low temperature desorption peak between 100 and 200 °C which is attributed to weak basic sites in catalysts. The desorption peak between 400 and 500 °C is attributed to strong basic sites. Interestingly, Ni/MgO@SiO2 has a strong desorption peak around 860 °C, suggesting it has strong alkalinity at high temperature. Therefore, Ni/Mg1.78AlO x @SiO2 displays the most basic sites that contributed to strong carbon dioxide adsorption capacity [37]. CO2 is the only oxygen-containing reactant in CRM, and generated active oxygen from CO2 dissociation can interact with carbon species to avoid continuous deposition of carbon on the catalyst surface ( C O 2 + C → 2 C O ). The adsorbed CO2 reacts with newly formed carbon species, which plays a significant role in eliminating carbon deposition. Thus, it is expected to achieve low carbon deposition and good stability for those catalysts possessing more basic sites, such as Ni/Mg1.78AlO x @SiO2.The photothermocatalytic activity of CRM was conducted in a homemade reactor with a quartz window (Fig. S5). Upon the focused UV–vis-IR irradiation, the surface temperature of the samples reached the equilibrium temperature quickly (Fig. S6). Gas chromatography was employed to detect both reactants and products. As shown in Fig. 3 a, b, for Ni@SiO2, the production rates of H2 (r H2) and CO (r CO) are 45.1 mmol min−1 g−1 and 62.4 mmol min−1 g−1, respectively. The r H2 and r CO of Ni/Al2O3@SiO2 increase to 92.8 mmol min−1 g−1 and 115.4 mmol min−1 g−1, and the r H2 and r CO of Ni/MgO@SiO2 become 89.7 mmol min−1 g−1 and 109.6 mmol min−1 g−1 respectively. It is obvious that magnesium aluminum silicate as the support increases both activity and stability significantly. For Ni/Mg0.67AlO x @SiO2, r H2 and r CO are 120.8 mmol min−1 g−1 and 138.4 mmol min−1 g−1, respectively. When the Mg/Al molar ratio is 1.22, its r H2 and r CO increase further to 115.6 mmol min−1 g−1 and 135.5 mmol min−1 g−1, respectively. Especially, Ni/Mg1.78AlO x @SiO2 exhibits the best catalytic performance. Its r H2 and r CO are 136.6 mmol min−1 g−1 and 148.2 mmol min−1 g − 1, respectively. Reaction rates of CH4(r CH4) and CO2 (r CO2) are 75.0 mmol min−1 g−1 and 81.2 mmol min−1 g−1, respectively. In addition, r H2 is slightly lower than r CO in all experiments, so that H2/CO ratio is less than 1 (Table 2 ), which is attributed to the existence of the reverse water-gas shift reaction (CO2+H2 CO+H2O, RWGS). Catalysts with magnesium aluminum silicate as the support have a higher molar ratio of H2/CO relatively, demonstrating their capabilities of inhibiting RWGS reaction.The change of light-to-fuel efficiency η over time is shown in Fig. 3c, the average light-to-fuel efficiency η of Ni@SiO2 is 11.6%. The reason is that when metal particles are directly exposed to the outer surface of SiO2 without any restriction, metal particles tend to grow or aggregate, so that catalytic activities will be inhibited. The average light-to-fuel efficiency η of Ni/Al2O3@SiO2 is 26.7%, and is lower than that of Ni/MgO@SiO2 (η is 27.4%). The main problem lies in relatively poor stability of Ni/Al2O3@SiO2, whose performance has an obvious decline with time compared with Ni/MgO@SiO2. The average light-to-fuel efficiency η of Ni/Mg0.67AlO x @SiO2 reaches 29.6%. As the molar ratio of Mg/Al increases, the light-to-fuel efficiency η further rises. The average light-to-fuel efficiency of Ni/Mg1.22AlO x @SiO2 is 34.5%. Ni/Mg1.78AlO x @SiO2 has the highest average light-to-fuel efficiency η of 35.7%. However, as the proportion of Mg2+ continues to increase, the performance begins to decline. The average light-to-fuel efficiency η of Ni/Mg2.03AlO x @SiO2 is only 31.7%. In contrast to other strategies for solar thermochemical CO2 reduction by CH4 below 800 °C (Fig. 3d), Ni/Mg1.78AlO x @SiO2 possesses a record-high light-to-fuel efficiency and its conversion of CH4 is as high as 70.9% (Table 2), which is a significant advantage. Excellent photothermocatalytic durability is another advantage. After 24 h of reaction (Fig. 3e), it's r H2 and r CO values slightly decrease, and the η value remains as high as 34.6%. And the average size of Ni nanoparticles of used catalysts is 13.7 nm (Fig. 3f), which maintains a small size. Furthermore, there are no obvious carbon species on the catalyst. Other catalysts with molar ratios of Mg/Al of 1.22, 0.67 and 2.03 also show relatively better stability than those catalysts containing single metal support during experiment tests (Fig. S7). Another important reason why Ni/MgAlO x @SiO2 catalysts have different catalytic performances lies in different Ni nanoparticle sizes (Fig. S3). Basically, the smaller the metal size, the better activity and stability the catalyst exhibits.It is well known that carbon deposition properties have a heavy influence on catalyst activity [40,41], thus thermal gravimetric (TG) was employed (Figs. 4 a and S7) to quantify the amount of carbon deposited during reactions. All spent catalysts show a significant weight loss in the range between 500 and 700 °C, due to the oxidation of these carbon species. These carbon species largely covering active sites severely damage catalysis performance. The amount of carbon deposited on the Ni/MgO@SiO2 and Ni/Al2O3@SiO2 are 45.57% and 42.05% respectively. In contrast, the Ni/Mg1.78AlO x @SiO2 catalyst displays the lowest weight loss of 12.15%, suggesting that the quantity of active and graphitic carbon formed is minimal. Ni/Mg1.78AlO x @SiO2 also has the lowest carbon deposition rate of 0.005 gcg−1 cat h − 1 (Fig. 4b). XRD analysis shows that the deposited carbon exists in the form of graphite 2H (PDF 75–1621) in spent samples of Ni/Mg1.78AlO x @SiO2 (Fig. S8). The carbon deposition rates of Ni/Al2O3@SiO2 (r c=0.056 gcg−1 cat h − 1) and Ni/MgO@SiO2 (r c=0.070 gcg−1 cat h − 1) are 11.2 times and 14 times as high as that of Ni/Mg1.78AlO x @SiO2, respectively. The results indicate Ni/MgAlO x @SiO2catalysts are good at inhibiting carbon deposition. According to Fig. 4c, carbon species type, active or graphitic, can be determined with the help of the Raman spectrum [42]. One peak at 1342 cm−1 could be assigned to D band, coming from active carbon, and the other at 1578cm−1 could be assigned to G band, coming from graphitic carbon [43,44]. The D band is deemed to be the vibration of carbon atoms with dangling bonds in an amorphous carbon network while the G band is contributed by the CC stretching vibrations of graphite layers [29,45]. The relative intensity of IG/ID could reflect the ratio of active carbon to graphite carbon [13,46] (Fig. 4d). The high IG/ID ratio of Ni @SiO2, Ni/MgO@SiO2 and Ni/Al2O3@SiO2 confirmed that main carbon species are active carbon [13]. Activated carbon can cover more catalytic active sites, which is more harmful to the stability of the reaction compared with graphitic carbon. The IG/ID of Ni/MgAlO x @SiO2 samples is lower, demonstrating the advantage of combining MgO and Al2O3 in enhancing stability. The IG/ID of Ni/Mg1.78AlO x @SiO2 is 0.92, which is the lowest among all samples, agreeing with its excellent stability shown in Fig. 3c.In most cases, CH4 dissociation and CO disproportionation are generally sources of carbon deposition [21,38]. To investigate the main source, temperature-programmed CH4 decomposition (TPMD) and CO disproportionation (TPCD) were conducted. As shown in Fig. S12, Ni/MgAlO x @SiO2 catalysts display a relatively stronger TCD signal with CH4 decomposition rate rising, implying the formation of carbon deposition. CO disproportionation begins to occur above 300 °C (Fig. 4e). Notably, Ni/MgO@SiO2 and Ni/Al2O3@SiO2 show the strongest and weakest CO consumption peaks, respectively. This suggests that Al2O3 helps to inhibit the side reaction of CO disproportionation. Although peaking around 395 °C, TPCD signals of Ni/MgAlO x @SiO2 catalysts also have two relatively weak peaks around 450 °C and 550 °C. Around practical operation temperature of 700 to 800 °C, TPCD signals become weak since CO disproportionation itself is an exothermic reaction and will be inhibited at high temperature conditions. The carbon deposition amounts of TPMD and TPCD are determined by TG. Carbon deposition rates of TPMD are higher than that of TPCD for all samples (Fig. 4f), illustrating that methane cracking is the main source of carbon deposition. Note that the carbon deposition of Ni/Mg1.78AlO x @SiO2 during TPCD or TPMD is not the lowest, but these carbon species can be quickly oxidized by CO2. This is because Ni/Mg1.78AlO x @SiO2 has a strong adsorption of CO2 as confirmed by previous CO2-TPD measurements (Fig. 2b). That explains why Ni/Mg1.78AlO x @SiO2 has the lowest carbon deposition rate during practical operation tests (Fig. 4b).To identify whether the lattice oxygen of MgAlO x @SiO2 contributes to inhibiting carbon deposition, an isotope labeling experiment using 12C18O2 and 12CH4 was performed (Supplementary Information). The gas in the reactor cavity was injected into GC–MS for detection before turning on the lamp. Only three peaks can be observed corresponding to carrier gas (Ar), and reaction gas 12CH4 and C18O2 (Fig. 5 a). The retention time is located at 8.4–8.6 min corresponding to 12CH4, and m/z = 16.1,15.1 and 14.1 belong to 12CH4 and its fragments. The retention time of C18O2 is located at 11.4–12.0 min, m/z = 48.1, 46.1, and 44 belong to 12C18O2,12C18O16O and 12C16O2, respectively (Fig. 5b), indicating that the air in the reactor had been cleaned in advance. Then we turned on the Xe lamp for 2 h to introduce concentrated light irradiation. After that, the reacted gas was injected into GC–MS for measurement. The intensity of m/z = 16.1,15.1,14.1(CH4) and m/z = 48.0 (C18O2) is weakened (Fig. 5b), and the additional crack peak at 7.6–7.8 min was attributed to CO (Fig. 5a). Corresponding intensities of fragments of m/z = 28.1 (12C16O) and 30.1 (12C18O) significantly increased (Fig. 5b), illustrating that CRM reaction occurred. The fragment strength corresponding to 12C18O2 decreased, while the fragment strength corresponding to 12C16O2 and 12C16O18O significantly increased. During the reaction, 12C18O2 is the only oxygen-containing reactant. As a result, the only source of 16O is the catalyst, which comes from the lattice oxygen in MgAlO x @SiO2. This is beneficial to reducing carbon deposition and promoting a highly active and stable photothermocatalytic reaction.In order to understand the improvement of Ni nanocluster catalysts with different substrates, we modeled pyramidal NPs loaded on slabs of MgAl2Si2O10H4, Mg3Si4O12H2 and Al2Si2O9H4. For comparison, Ni (111) surface models were also established. All substrates are constructed with 4 × 2 × 1 supercell with the (001) facet cleaved. Ni NPs are built as a pyramid of exposed (111) surfaces with 30 atoms. Calculations are carried out in the framework of DFT using the generalized gradient approximation (GGA) of Perdew–Burke-Ernzerhof (PBE) [47]. The VASP (Vienna ab initio simulation package) package is employed with the projected augmented-wave method [48,49]. The kinetic cutoff energy for the plane-wave basis is set to 400 eV. The Brillouin zone integration was performed on a Gamma-centered 1 × 1 × 1 K mesh. All the atoms are fully relaxed until the force on each atom is less than 0.05 eV /Å. To analyze the performance of catalysts, we used the periodic slab models with a vacuum layer of 15 Å. Our unit cell contained four layers with two bottom layers fixed to relax the module of the slab. Transition state searches were conducted using the climbing image nudged elastic band (CI-NEB) [50,51].CH4 and CO2 are activated to form active species, which are the premise of CRM. The elementary steps of CRM reaction mainly include three parts: CH4 activation dehydrogenation, CO2 activation and oxidation of CH* and C* species. It is crucial for CRM reaction to oxidize CH* or C* species to remove the carbon deposition and suppress the deactivation of the catalyst. The reaction energy (ΔE) and activation energy (Eact) of the elementary steps of the DRM reaction are shown in Table S1. For simplifying the calculation, the Eact calculates several key steps in the reaction. The reaction energy diagram for CRM on Ni (111) surfaces is depicted in Fig. 5c. It can be seen that the Eact for CH4 to remove an H atom to become CH3* on the surface of Ni/Mg1.78AlO x @SiO2 is lower than that of other supports, which only needs 0.85 eV. This ensures a high reaction rate (entry 1 of Table S1) of CH4 activation and is also consistent with our experimental results. Besides, from the perspective of reducing carbon deposition, the Eact value of CH* oxidation to CHO* (1.02 eV) is less than the Eact value of CH* dissociation to C* (1.22 eV) for Ni/Mg1.78AlO x @SiO2 (Fig. S13). Subsequently, C* species formation is suppressed. On the contrary, Ni/Al2O3@SiO2 needs the highest energy to oxidize CH*(1.44 eV), while the Eact value of CH* dissociation is only 1.24 eV, which causes more carbon species deposition (Table S1).It has been reported that Ni nanoparticles can simultaneously act as active sites and plasmonic promoters under light illumination [52–54]. Hence, distinct catalysis performances may be observed between light illumination and dark conditions. To reveal how the light affects the photothermocatalytic CRM reaction, optical absorption properties of Ni@SiO2, Ni/MgO@SiO2, Ni/Al2O3@SiO2 and Ni/Mg1.78AlO x @SiO2 were measured firstly (Fig. 6 a). Since light excites surface plasmon resonances of Ni nanoparticles [55], all samples show good solar absorption properties. Although Ni/Mg1.78AlO x @SiO2 has an intermediate absorptance compared with Ni/MgO@SiO2 and Ni/Al2O3@SiO2, its value is still over 80% across the entire solar spectra. To check whether high temperature plays a vital role in the solar-driven CRM, the experiment was conducted at near room temperature for Ni/Mg1.78AlO x @SiO2. No H2 and CO were detected (Fig. S14). This demonstrates that the high photothermocatalytic activity of Ni/MgAlO x @SiO2 is derived from light-driven thermocatalytic CRM.To directly compare differences between photothermocatalysis and thermocatalysis, CRM reactions over Ni/Mg1.78AlO x @SiO2 were performed under light irradiation and dark conditions. It can be seen from Fig. 6b that whether driven by light or heat, the reaction rate of the reactants increases with temperature, indicating that high temperature is conducive to the catalytic reaction. At any temperature between 660 °C and 860 °C, the reaction activity of Ni/Mg1.78AlO x @SiO2 under light irradiation is better than that in dark conditions. The ratio of H2/CO under dark conditions is always lower than that of photothermocatalysis at the same temperature (Fig. 6c), although it increases with temperature for both cases since RWGS is inhibited by high-temperature conditions.Kinetic studies are conducted to further explore how light irradiation affects solar CRM performances. Arrhenius plots using the conversion rate of CH4 under both UV–vis-IR illumination and dark conditions are presented in Fig. 6d. These curves demonstrate a good linear relationship, and have a good agreement with the Arrhenius equation [26,56] ( k = A e − E a / R T ). Accordingly, the apparent activation energy for CH4 of Ni/Mg1.78AlO x @SiO2 with focused UV–vis-IR irradiation is 31.2 kJ mol−1, which is much less than that under dark conditions (70.3 kJ mol−1). The decrease in apparent activation energy can be ascribed to the excitation of hot electrons in metallic Ni. It has been demonstrated by several references that excited Ni can dramatically decrease the activation energy of CO2 dissociation and CH* oxidation compared with ground states (dark conditions) [16,19,20,52]. This explains the reduced apparent activation energy and promoted activity of Ni/Mg1.78AlO x @SiO2 under direct light illumination.In summary, highly efficient solar-driven CO2 conversion with CH4 is achieved via interconnected Ni/MgAlO x nanoflakes grown on SiO2 particles with an ultrahigh light-to-fuel efficiency of 35.7% below 800 °C. The excellent performance can be ascribed to the following three aspects. First of all, highly dispersed nickel nanoparticles with small sizes and strong metal-support interactions are realized on Ni/MgAlO x @SiO2. And the formation of Mg-Al phyllosilicate provides many basic sites, promoting the absorption and activation of CO2 molecules. Secondly, the active oxygen in the carrier participates in the solar-driven CRM reaction, which is beneficial to suppressing the formation of carbon species produced by CH4 dissociation and CO disproportionation. DFT calculations also demonstrate that the reaction on MgAlO x @SiO2 has a lower activation energy of CH* oxidation to CHO* and improves the dissociation of CH4 to CH3*. At last, full-spectrum solar energy can be efficiently captured and the light-driven CRM greatly reduces the apparent activation energy, thereby significantly improving catalytic activities under direct light illumination. Our work demonstrates that Ni/MgAlO x @SiO2 can realize solar-driven CO2 conversion with ultrahigh light-to-fuel efficiency and superior stability, thus is promising to provide new opportunities for tackling global climate change and energy shortage problems.The authors declare no conflicts of interest in this work.This work was financially supported by the Basic Science Center Program for Ordered Energy Conversion of the National Natural Science Foundation of China (Grant No. 51888103), the National Key R&D Program of China (Grant No. 2021YFF0500700) and the Basic Research Program of Frontier Leading Technologies in Jiangsu Province (Grant No. BK20202008).Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.fmre.2022.04.011. Image, application 1

Solar-driven CO2-to-fuel conversion assisted by another major greenhouse gas CH4 is promising to concurrently tackle energy shortage and global warming problems. However, current techniques still suffer from drawbacks of low efficiency, poor stability, and low selectivity. Here, a novel nanocomposite composed of interconnected Ni/MgAlO x nanoflakes grown on SiO2 particles with excellent spatial confinement of active sites is proposed for direct solar-driven CO2-to-fuel conversion. An ultrahigh light-to-fuel efficiency up to 35.7 %, high production rates of H2 (136.6 mmol min−1 g  − 1) and CO (148.2 mmol min−1 g−1), excellent selectivity (H2/CO ratio of 0.92), and good stability are reported simultaneously. These outstanding performances are attributed to strong metal-support interactions, improved CO2 absorption and activation, and decreased apparent activation energy under direct light illumination. MgAlO x @SiO2 support helps to lower the activation energy of CH* oxidation to CHO* and improve the dissociation of CH4 to CH3* as confirmed by DFT calculations. Moreover, the lattice oxygen of MgAlO x participates in the reaction and contributes to the removal of carbon deposition. This work provides promising routes for the conversion of greenhouse gasses into industrially valuable syngas with high efficiency, high selectivity, and benign sustainability.

The high demand for energy in modern society has resulted in anthropogenic activities of humans leading to a sharp increase in the atmospheric CO2 concentration (>400 ppm) causing major concern for attempts to manage global warming and the greenhouse effect [1–3]. In this regard, CO2 valorization is attracting interest among researchers around the globe. Carbon dioxide is considered an alternative carbon source from which to prepare a variety of C1 feedstock chemicals and fuels (including such as formaldehyde, formic acid, methanol, and methane) through the hydrogenation reaction [4–6]. Among these hydrogenation products, formic acid is an attractive commodity chemical with numerous applications involving several industries such as food and agriculture, leather, pharmaceuticals, and textiles [7–9]. Moreover, it is a promising source for use in storing carbon–neutral hydrogen [10,11]. However, the H2 gas required for CO2 hydrogenation is currently produced from fossil fuels through methane steam reforming, which produces a large amount of CO2 and signifies the need for alternative renewable and environmentally benign H2 sources [12,13].Numerous alcohols have been utilized as liquid hydrogen sources in several catalytic transfer hydrogenation (CTH) reactions [14–18]. However, most of the alcohols were obtained from non-renewable sources and their dehydrogenated products are of low value. From this view, glycerol could emerge as a viable H2 source for two basic reasons. First, glycerol is renewable and its overproduction in the biodiesel industry makes it cheap and abundant. Second, the dehydrogenation of glycerol offers value-added chemicals such as aldehydes, ketones, diols, or carboxylic acids, which improves the economics of the biodiesel industry [19]. Along with several other important uses of glycerol [20,21], it has also been successfully utilized as a hydrogen source for the reduction of CC and CO compounds with high product yields under strongly basic aqueous conditions [22,23]. Nevertheless, the reduction of thermodynamically stable C–O bonds in a CO2 molecule is more challenging than the reduction of the C–O bond in carbonyl groups of organic compounds [24]. To overcome the high stability of gas-phase CO2 (ΔG°298 K = 33 kJ·mol−1), hydrogenation reactions were performed in aqueous media, where CO2 was in equilibrium with HCO3 − (pKa1 = 6.35) and the reaction became slightly exergonic (ΔG°298 K = −4 kJ·mol−1) [25–27].Interestingly, the simultaneous conversion of one industrial waste and one side product (CO2 and glycerol) to create value-added commodity chemicals (formic acid and lactic acid) can be regarded as a “two birds, one stone” strategy. Several researchers studied the reduction of CO2 in the form of bicarbonate or carbonate with glycerol and alcohol in high-temperature water (HTW), with or without transition metals (Fe, Zn, Ni, etc.) as catalysts [28–31]. Jin et al. reported the hydrogen-transfer reduction of NaHCO3 with glycerol and isopropanol in alkaline HTW without catalyst; however, the formic acid yield was less than 10% relative to the initial NaHCO3 concentration, indicating the low bicarbonate reduction efficiency of this reaction system [28,29]. Furthermore, the same group studied the reduction of bicarbonate in the presence of transition metals in alkaline HTW. Transition metals acted as reducing agents to reduce CO2 to formate using water as a hydrogen source. The metals were oxidized simultaneously. In a second step, glycerol might reduce metal oxides back to metals by H2 donation [30,31]. Nevertheless, the use of an equimolar (or higher) amount of metal relative to bicarbonate, higher reaction temperature (300 °C), and higher leaching probability of the catalyst due to continuous phase-change, limit the process for large scale applications.Recently, homogeneous Ir-carbene catalysts were utilized for CO2 and carbonate transfer hydrogenation (TH) using glycerol [32–34]. The catalysts selectively afford FA and LA with high turnover numbers (TON) within a moderate range of reaction temperature (150–180 °C). Although these catalysts are active at low reaction temperatures, the development of new and more efficient heterogeneous catalysts are indispensable considering the catalyst-separation issue in homogeneous catalysis. From this perspective, Lin et al. screened several commercial, supported, noble-metal heterogeneous catalysts in one-pot aqueous-phase TH involving glycerol and bicarbonates to increase the LA and FA yield at milder reaction temperatures [35]. Carbon supported Pd showed the best results among the catalysts tested: 55% LA and 30% FA yield at 240 °C. Very recently, our research group also demonstrated that meticulously prepared, ZIF-11 derived, graphitic nanoporous carbon (Ru/NCT) catalysts could efficiently produce LA and FA from simultaneous conversion of glycerol and carbonates [36]. Moreover, this was achieved with high TON and space–time yield (STY). These studies point out the enormous scope for further development of new and efficient heterogeneous catalysts.Encouraged by these previous studies, we decided to develop laboratory synthesized, supported noble-metal catalysts for the one-pot conversion of glycerol and carbonates. Among the noble metals, Pt-based catalysts are most effective for alcohol and glycerol dehydrogenation reactions [37–41]. Additionally, supported Pt catalysts have also been successfully utilized in several transfer hydrogenation reactions [22,42–44]. In this context, we planned to investigate the performance of supported Pt catalysts for the simultaneous conversion of glycerol and carbonate involving both dehydrogenation and hydrogenation steps. The reaction parameters, including temperature, time, the concentration of glycerol and carbonate, catalyst loading, water amount, and effect of the CO2 substrate, were studied systematically. The catalyst was recycled in four consecutive cycles with little change in its activity and product selectivity. Several methods were adopted for the characterization of fresh and used catalysts to better understand the factors governing the catalytic activity and the changes that occurred in the recycled catalyst.Glycerol (99%), formic acid (99%), potassium carbonate anhydrous (99.55%), potassium bicarbonate (99%), sodium carbonate anhydrous (99%), and sodium bicarbonate (99%) were purchased from Samchun Chemicals, Korea. Lactic acid solution (50% in water), glyceraldehyde (90%), 1,2 propanediol (99.5%), potassium hydroxide (90%), chloroplatinic acid hexahydrate, ruthenium(III) chloride hydrate, potassium tetrachloropalladate(II) (98%), and zirconium oxide ceramic grade (99%), were procured from Sigma-Aldrich. Aluminum oxide activated acidic gamma (96%) and activated carbon were obtained from Alfa Aesar and Strem chemicals, respectively. All the chemicals were of commercial grade and were used without further purification.Oxide-supported metal catalysts were prepared by the wet-impregnation method as follows. A portion (2 g) of the solid support and an aqueous solution (100 mL) of the metal precursor was stirred for 12 h at ambient temperature. Water was evaporated by a rotary evaporator at 50 °C under reduced pressure. After drying in an oven, the solid was calcined at 400 °C for 4 h with a heating ramp of 2 °C/min in the muffle furnace under constant airflow (150 cc/min). 1 g of calcined powder pressed into the pellet, crushed and sieved (No. 20–40 mesh) to make granules, and then reduced at 300 °C for 2 h (2 °C/min heating ramp) with 10% H2 in N2 flow (50 mL/min) in a conventional stainless-steel fix-bed reactor (i.d. 5 mm, length 300 mm). For the platinum-on-carbon catalyst, the calcination temperature was set to 300 °C for 3 h with a heating ramp of 2 °C/min and reduction was done in a tube furnace at 300 °C for 2 h (2 °C/min heating ramp) with 5% H2 in Ar flow (300 mL min−1).Powder X-ray diffraction patterns (PXRD) of all the catalysts were obtained using a Rigaku D/Max-2200 V X-ray diffractometer (Cu Kα-radiation, λ = 1.5406 Å) at 40 kV and 40 mA. The N2 adsorption–desorption isotherms were measured at 77 K using a Micromeritics Tristar 3000 system. The samples were dehydrated under vacuum at 423 K for 12 h before analysis. The specific surface areas were evaluated using the Brunauer-Emmett-Teller (BET) method. Scanning transmission electron microscopy (STEM) analysis of the reduced catalysts was performed to analyze the distribution of metal particles, and a minimum of two hundred metal particles were considered for calculation of average metal particle size. TEM-Talos (F200X system) operating at an accelerating voltage of 200 kV was used. X-ray photoelectron spectra (XPS) were measured on a Kratos AXIS SUPRA instrument (UK) with a monochromatic Al Kα X-ray source operated at 20 eV pass energy. Inductively coupled plasma-atomic emission spectrometry (ICP-AES) was conducted on an iCAP 6500 duo series (ICP-AES) analyzer (Thermo, USA). Temperature programmed desorption (TPD) profiles of the catalysts were measured on a Micromeritics AutoChem II 2920 V5.02 apparatus equipped with a thermal conductivity detector. About 100 mg of samples were pretreated at 400 °C for 1 h in a helium flow (50 mL/min), before the adsorption step. Subsequently, pretreated samples were exposed to NH3 or CO2 gas at 40 °C for 15 min with a flow rate of 50 mL/min. Physically adsorbed NH3 or CO2 gases were removed by purging with helium gas for 30 min at the same temperature and flow rate. TPD data were recorded between 40 and 400 °C at a heating rate of 10 °C/min. Temperature programmed reduction (TPR) experiments were carried out using an AutoChem II-2920 V4.06 (Micromeritics) analyzer. About 100 mg of sample was used for each experiment, preconditioned at 400 °C/h in He gas flow at 50 mL/min. This was followed by sample analysis from room temperature to 600 °C at a heating rate of 5 °C/min under 10% Ar-H2 gas at a constant flow rate of 50 mL/min. FTIR spectra were recorded on a Nicolet FTIR spectrometer (MAGNA-IR 560) using KBr pellets.All experiments were carried out in a locally made high-temperature, high-pressure, 100 mL capacity batch reactor fitted with a standard mechanical stirrer and thermocouple. The required amount of glycerol, carbonate/bicarbonate, water, and catalyst was placed in the reactor. The reactor was sealed, purged three times with nitrogen and pressurized to 400 psi. The reaction mixture was stirred at 500 rpm and the reaction time was counted once the desired reaction temperature was reached. After reaction, the reactor was cooled to room temperature, the pressure was released by the vent valve, and the off-gas was collected in a gasbag (only for a couple of reactions). The reaction mixture was centrifuged to separate the catalyst from the liquid solution. The recovered catalyst was washed with water and methanol and dried at 100 °C in an oven before being subjected to the next cycle. Glycerol conversion and product yield were calculated based on the following formulas. C o n v e r s i o n % = I n i t i a l m o l e s o f g l y c e r o l - F i n a l m o l e s o f g l y c e r o l I n i t i a l m o l e s o f g l y c e r o l × 100 Y i e l d % = M o l e s o f p r o d u c t × C a t o m s i n p r o d u c t M o l e s o f r e a c t a n t × C a t o m s i n r e a c t a n t × 100 Aqueous samples were collected, diluted with water, and passed through a 0.22 μm pore-size filter. The aqueous portion was then subjected to high-performance liquid chromatography (HPLC) analysis. HPLC analysis was performed using a YL9100 HPLC system equipped with a UV–VIS detector (YL9120) at 218 nm and a refractive index detector (YL9170). The samples were separated in an Aminex 87-H column from Bio-Rad, using 5 mM H2SO4 as the mobile phase at 0.5 mL/min flow and a column temperature of 45 °C. The measure of LA, FA, glyceraldehyde (GLA), and 1,2 propanediol (1,2-PDO) was calibrated using multiple-point external standard methods to calculate their yield via HPLC. Typical HPLC chromatograms with UV and RI detectors were depicted in Fig. S1. With our HPLC operational conditions, glycerol, LA, and FA peaks overlapped in the RID detector, which makes quantification of glycerol difficult. For this reason, its conversion was calculated from proton nuclear magnetic resonance ( 1H NMR) spectroscopy using a Bruker 500 MHz NMR spectrometer. A typical 1H NMR spectrum of a liquid sample is shown in Fig. S2. The analysis sample was prepared by dissolving a known amount (~5 mg) of isonicotinic acid as an external standard in 2.5 mL of product solution, followed by dilution with D2O. Collected gas samples were analyzed by a gas chromatograph (GC) equipped with a Carboxen 1010 PLOT column and a thermal conductivity detector (TCD). A typical chromatogram is depicted in Fig. S3.In general, bicarbonates/carbonates are utilized as CO2 sources considering their high solubility in water, operational simplicity, and accuracy in the quantification of CO2 [28,29,32–36,45]. Moreover, because it was decided to perform the reaction using a more environmentally friendly method, highly corrosive strong bases were avoided. Therefore, carbonate was chosen over bicarbonate to increase the initial pH of the reaction that required to promote glycerol dehydrogenation.As shown in Table 1 , a catalyst screening study was performed for simultaneous conversion of potassium carbonate and glycerol at 180 °C. The four major reaction products (LA, FA, GLA, and 1,2-PDO) were observed for almost all the catalysts tested. GLA is the most common product of glycerol dehydrogenation (the first and key step) and subsequently undergoes several consecutive reactions that include dehydration, keto-enol tautomerization, and Cannizzaro reaction to produce lactic acid (Scheme 1 ) [46]. On the other hand, formation of 1, 2-PDO and formic acid was attributed to the hydrogenation of intermediate pyruvaldehyde and a CO2 source, respectively. Importantly, the hydrogen eliminated from glycerol in the dehydrogenation step was utilized for these hydrogenation reactions.The glycerol conversion and product yields were negligible in the absence of the catalyst (Table 1, entry 1); however, when Pt/γ-Al2O3 was used for the reaction (Table 1, entry 5) 25% glycerol conversion and ~13% yields of LA and FA were achieved. This highlighted the role of the catalyst in the simultaneous conversion of glycerol and carbonate. Pt/γ-Al2O3 catalyst lowers the reaction temperature required for this reaction, which in the absence of a catalyst, is reported in previous studies to require ~300 °C [28,29]. Moreover, PtOx/Al2O3 (before reduction) was found less effective than Pt/γ-Al2O3 (after reduction) (Table 1, entry 2 and 5). This suggests that metallic Pt species on the surface of the catalyst are more active than PtOx species, confirmed through XPS analysis (Table S1 and Fig. S4). It is also possible that some of the PtOx gets reduced during the reaction by the hydrogen released from glycerol.In an aqueous-phase glycerol conversion reaction, formic acid formation through the decomposition of glycerol derived products such as lactic acid cannot be ignored [46,47]. Hence, to determine whether FA is generated through the hydrogenation of carbonate or from the degradation of lactic acid, we performed two sets of controlled experiments. In the first set (Table 1, entry 4), the reaction was performed in the absence of K2CO3; however, KOH was added to adjust the pH of the reaction mixture to 11.7 (initial pH with K2CO3). In the second set (Table 1, entry 3), the reaction was carried out without the addition of glycerol. In both cases, no FA was detected, which confirmed that FA was formed through the reduction of K2CO3 by hydrogen released from the glycerol.We screened a variety of supported noble-metal catalysts for the simultaneous conversion of glycerol and carbonate under the same reaction conditions. First, γ-Al2O3-supported Pt, Pd, and Ru catalysts were tested and it was found that the Pt catalyst (Pt/γ-Al2O3) showed the highest activity, followed by Pd and Ru (Table 1, entry 5–7). Pt/γ-Al2O3 catalyst showed potential for both glycerol dehydrogenation and hydrogenation of carbonate, whereas Pd/γ-Al2O3 was just a little behind the gamma-alumina supported Pt catalyst in terms of glycerol conversion and LA yield however, it produced an equal amount of FA, indicating its potential in the hydrogenation reaction. Glycerol conversion and acid yields were significantly lower in the case of Ru/γ-Al2O3 catalyst as compared to Pt and Pd catalysts supported on gamma-alumina. In addition, when we compared these catalysts in terms of TON, once again Pt/γ-Al2O3 was significantly ahead of other catalysts (the highest TON: 459 and 116 for LA and FA, respectively). To investigate the difference in catalytic activity, catalysts were subjected to various characterization techniques, and their textural and chemical properties are reported in Table S1 and Table 2 . As shown in Table 2 the actual metal content of the catalysts measured by ICP-AES analysis is close to their theoretical content (3%). Despite having a higher metallic content (M0), alumina-supported Ru showed lower activity than Pt and Pd catalysts did. This was primarily attributed to its poor distribution on the alumina support, and significantly bigger particle size discussed thoroughly in the subsequent section.As shown in Fig. 1 , Pt and Pd particles are finely distributed on the alumina support, with a narrow size distribution in the ranges 0.8–3.2 nm and 1.6–4.8 nm, and average particle size of 1.7 and 3.0 nm, respectively (Fig. 1a and b). In contrast, Ru particles are poorly distributed on the alumina support, with broader particle-size distribution in the range 10–50 nm and an average particle size of 20.6 nm (Fig. 1c). Despite having similar metal content, a significant difference between the dispersion of Ru particles and Pt and Pd particles on gamma-alumina support is most probably associated with the location of metal in the precursor. Pt/γ-Al2O3 and Pd/γ-Al2O3 prepared from H2PtCl6 and K2PdCl4 precursors, respectively, where metals are in anionic position, whereas Ru/γ-Al2O3 prepared from RuCl3 precursor with Ru in cationic part. This may further suggest that precursor with metal in the anionic group is favorable for preparing catalyst with better dispersion and smaller particle size [48]. However, some other effects such as valance state of metal in precursor, the content of chloride ions in precursor, and pH of metal precursor solution in water on decomposition procedure of precursor and dispersion of metal particles on support cannot be ruled out completely. Therefore, a further detailed investigation is required in this matter. The results from the STEM analysis, which are in good agreement with the PXRD analysis results, reveal large, visible diffraction peaks of Ru0. The peaks for Pd0 are small, yet visible, and there is no observable diffraction peak for Pt0. This confirms its high dispersibility on the alumina support (Fig. 2 ).The reducibility of the gamma-alumina supported catalysts was examined by H2-TPR analysis, and the results are shown in Fig. 3 . All three catalyst profiles feature essentially one main reduction peak related to the reduction of respective metal oxides to metal, and are consistent with previous reports [49–51]. However, such events are observed at different reduction temperatures, probably due to differences in strength of the metal-support interaction. A sharp, intense, reduction peak at 85 °C was observed for Ru/γ-Al2O3, whereas less intense and broader peaks were detected for alumina-supported Pd and Pt catalysts in the temperature range 40–250 °C. This indicates that Pt strongly interacts with the gamma-alumina support (followed by Pd and Ru, respectively). This strong metal-support interaction leads to fine dispersibility of Pt on gamma-alumina as we have already confirmed through PXRD and STEM analysis.To check the role of the support in simultaneous conversion of glycerol and carbonate, Pt was further impregnated on ZrO2 and carbon supports and then their catalytic activities were compared with Pt/γ-Al2O3 (Table 1 entry 8 and 9). Using Pt/ZrO2 catalyst, yields of LA and FA are much lower than with Pt/γ-Al2O3 catalyst, even after achieving a similar level of glycerol conversion. On the other hand, Pt/C showed much higher glycerol conversion (41.7%) and similar LA yield (11.7%). However, the FA yield (4.8%) was significantly lower than with Pt/γ-Al2O3. Both Pt/ZrO2 and Pt/C catalyst seem less active in transferring hydrogen to carbonate and end-up with very low FA yields. Moreover, it is noteworthy that selectivity for LA from glycerol is higher on all the gamma-alumina-supported catalysts, in comparison with the carbon- and zirconium oxide-supported catalysts.To rationalize the differences in catalytic activities, catalysts were characterized using various techniques. Although the surface area of Pt/ZrO2 is ~20 times smaller than that of Pt/γ-Al2O3, we found very little difference in the glycerol conversions. In contrast, Pt/C has around 10 times larger surface area than Pt/γ-Al2O3 (Table 2, Fig. S5). However, that perspective enhancement in glycerol conversion was not significant, indicating that the surface area of the catalyst alone does not have a substantial effect on its catalytic activity. STEM analysis (Fig. 1a, d, and e) showed that the Pt metal particles are well dispersed on all three of the supports. The average metal particle size of Pt on the alumina or carbon support is the same (1.7 nm), while it is slightly higher on the zirconium oxide support (2.2 nm). This high dispersibility of Pt on different supports was further confirmed with the results from PXRD analysis (Fig. 2). There, no metal diffraction peak was observed for Pt/C or Pt/γ-Al2O3, and a very small peak was observed for Pt/ZrO2. Because all three catalysts showed fine metal dispersion on their respective support and did not have a significant difference in metal particle size, we deepened our study by accessing the chemical properties of the catalyst using TPD analysis.The acidic properties of the catalysts were studied using NH3-TPD and the results presented in Fig. 4 , Fig. S6, and Table 2. All three gamma-alumina-supported catalysts (Ru, Pd, and Pt) possess a high concentration of weak to moderate acid sites (0.6–0.74 mmol/g) in the temperature range of 40–400 °C. However, when we compare acidity of Pt supported on three different supports (Fig. 4), it can be seen that Pt/γ-Al2O3 show weak and moderate acid sites in the temperature ranges 40–215 °C and 215–400 °C, respectively. In contrast, the results for the other two catalysts indicate the presence of only weak acid sites in the temperature range 40–215 °C. In Table 2, the total amount of acid sites from the results of the integration of the TPD profile was reported as moles of NH3 per unit weight. The concentration of acid sites is higher in the case of the Pt/γ-Al2O3 catalyst than with the other two catalysts (Fig. 4). The Pt/C catalyst showed little acidity, probably due to the presence of surface functional groups such as OH and COOH, as shown in Fig. S7. When we correlated the acid properties of the catalysts with their catalytic reactivity, it was found that the Pt/γ-Al2O3 catalyst, which has a large number of acid sites with weak to moderate strength, might be favorable for the adsorption of glycerol and carbonate on its surface. In this process, it may engage with carbonate/bicarbonate anions in the aqueous medium and decrease OH− ions availability. However, the concentration of carbonates/bicarbonates is still higher than the concentration of acid sites of the catalyst utilized in the reaction, which provides room to proceed. In compensation, Pt/γ-Al2O3 would have higher carbonate molecules adsorbed on its surface than Pt/C and Pt/ZrO2 surfaces, resulting in nearly three times higher formate yield for earlier compared to later catalysts (Table 1, entry 5, 8, and 9). Furthermore, the high dispersibility and small particle size of Pt on its surface facilitates the reaction to produce LA and FA. Sievers et al. thoroughly studied the surface interaction of glycerol with various metal oxides having acid and base properties using FTIR spectroscopy. The study results reveal that even in the presence of water, the primary alcohol groups of glycerol strongly bonded to the Al sites of gamma-alumina to form bridging alkoxy bonds. The secondary alcohol groups of glycerol exhibited an additional hydrogen-bonding interaction with surface oxygen atoms of the gamma-alumina [52]. This implies that the γ-Al2O3 support has a clear advantage over ZrO2 and carbon support for the adsorption of glycerol molecules to its surface.Based on preliminary catalyst screening tests and a characterization study, higher activity was observed with Pt/γ-Al2O3. This is attributed to the high dispersion of Pt nanoparticles on the alumina support and its strong interactions with glycerol due to the presence of a large number of acidic sites with weak to moderate strength, compared to the other tested supports. The reaction temperature used for catalyst screening in this work was much lower (180 °C) than that used with the previously studied heterogeneous, carbon-supported noble-metal catalysts at 240 °C, for the same reaction [35]. This resulted in lower catalytic activity. Glycerol-to-lactic acid occurs via an endothermic reaction, it is favored at high temperatures. This is because, under mild conditions, it is difficult to remove OH groups from glycerol and then oxidize it to glyceraldehyde [53]. Considering this point, we examined the effect of the reaction temperature on catalyst activity and product yield, as shown in Fig. 5 a. Glycerol conversion was low at lower reaction temperatures (from 160 to 200 °C); however, thereafter, rapid improvement was noticed. Lactate and formate yields also increased with reaction temperatures up to 220 °C; however increasing the temperature further to 240 °C, resulted in decease in the formate yield. This was probably due to the decomposition of potassium formate in the presence of catalyst at a higher temperature. This kind of decomposition pattern for LA and FA salts was also reported over Pd/C catalysts, although at higher temperature 270 °C [35]. Fig. 5b and c show the effect of glycerol and the K2CO3 concentration on product yields, respectively. As the concentration of glycerol declined from 2 M to 0.5 M without changing the K2CO3 concentration, the glycerol conversion and lactate yield increased steadily from ~49 to 97% and from ~22 to ~46%, respectively. In contrast, the formate yield progressively decreased from ~24 to ~12%. In hydrothermal media, OH− ions generated from equilibrium reactions between carbonates and water [54] might also act as a catalyst to promote the glycerol-to-lactate reaction [46,55]. At low glycerol concentration, more CO3 2− ions generate more OH− ions, which promote the dehydrogenation of glycerol and increase the lactate yield. On the other side, because the lower quantity of glycerol hydrogen produced from the dehydrogenation step is insufficient to reduce the carbonate, the result is a decrease in the formate yield.To achieve higher formate yield, we reduced the concentration of carbonate from 0.5 M to 0.125 M while the glycerol concentration was kept constant at 0.5 M, as shown in Fig. 5c. As we expected, the FA yield increased from ~12% at 0.5 M to 26% at 0.25 M. However, it dropped to ~17% with further decline in the carbonate concentration to 0.125 M. In contrast, a decrease in the carbonate concentration ultimately reduced the reaction pH and the concentration of OH− ions in the aqueous media. Because fewer OH− ions were available to promote the reaction, glycerol conversion dropped from 97 to ~67%, and the corresponding lactate yield from ~46 to 36%. This observation indicates that an optimal ratio of glycerol to carbonate is required to maximize the lactate and formate yields. The hydrogenation of bicarbonate/carbonate, and the decomposition of formate, are both reversible reactions. Thus, it is very likely that at a certain temperature and pressure, the formate yield is limited by the reaction equilibrium.Owing to the importance of subcritical water in an aqueous-phase glycerol transformation reaction, the effect of the amount of water on the product formation was examined (Fig. 5d). With increasing water amount from ~38 to ~58 mL, conversion dropped slightly from 84 to 78%. This could be related to a slight drop in the reaction pH. The formate yield remains stable at ~25 to 26% while a moderate improvement from ~41 to 50% in the lactate yield was observed. This is probably due to the higher distribution of glycerol molecules on the catalyst surface, or to lower decomposition of lactate under the higher-pressure conditions generated by increased volume of the reaction solution.The effect of catalyst amount is shown in Fig. 6 a. It can be seen that, by increasing the amount of the catalyst from 0.1 to 0.2 g, glycerol conversion increases (54–78%), along with the yield of lactate (22–50%) and formate (10–26%). Catalyst amounts higher than 0.2 g inversely affect the glycerol conversion and lactate yield (drop to 61 and 28%, respectively, using 0.4 g of the catalyst). This might be associated with the presence of an excess amount of acidity from the gamma-alumina support, which could engage with carbonate anions in the aqueous medium and thereby decrease OH− ions responsible for promoting lactate formation through the glycerol dehydrogenation step. Another reason for a drop in lactate yield evident at higher catalyst loading hinted at the possibility of lactate decomposition in the presence of excess acidity and metallic sites. On the other hand, an increase in formate yields at higher catalyst loading is probably associated with an increase in metallic sites and the decomposition of lactate. In brief, a catalyst amount beyond 0.2 g inhibits the formation of lactate and promotes the formation of formate. Therefore, 0.2 g of catalyst amount was chosen for further experiments.The influence of reaction time on the conversion and product yields was studied at 220 °C (Fig. 6b). The yield of lactate and formate increased rapidly with extension of the reaction time. Lactate reaches a plateau at 12 h with 50% yield, whereas formate yield starts declining after reaching the maximum yield of 29% at 8 h. The yield of 1, 2-PDO was initially stable (near 6% up to 12 h), but then climbed to 10% at 16 h. In similar results, Huo et al. reported that longer reaction resulted in the decomposition of formic acid and a drop in its yield during hydrogenation of bicarbonate over skeletal CuAlZn catalyst [45]. Therefore, 8 h of reaction time was chosen for further experiments.As shown in Fig. 6c, the various CO2 sources (K2CO3, Na2CO3, KHCO3, NaHCO3, and CO2) were tested as hydrogen acceptors against glycerol. It has already been reported that only bicarbonate is known to undergo hydrogenation to produce FA [25]. Among the CO2 sources used herein, carbonates were favored over bicarbonates in terms of LA and FA yield. Moreover, the best result with K2CO3 (compared to those with the other CO2 sources) is attributed to the initial high pH of the solution, which is essential for the dehydrogenation of glycerol to produce LA. Furthermore, alkali cations on carbonates can also affect the reaction. This is probably due to their different solubility and basicity, as well as to the formation of alkali salt products. The latter could be realized due to the dissimilar results when using K2CO3 and Na2CO3 as CO2 sources. Direct use of gaseous CO2 after adjusting the initial pH with KOH (same value as with K2CO3) produces 1,2-PDO as a predominant product with very little lactate and no formate. This indicates that the addition of CO2 makes the reaction solution acidic. This causes utilization of a different reaction path from pyruvaldehyde to 1,2-PDO, instead of to LA. It has already been reported that a basic condition is required to facilitate the production of LA from glycerol and that 1,2-PDO is favored by lowering the reaction pH [56]. These results are similar to those in previous studies using homogeneous and heterogeneous catalysts for the reduction of direct CO2 using glycerol as hydrogen source: very low yield of lactate and formate [32,35].The recycle tests of the Pt/γ-Al2O3 catalysts were performed under optimized reaction conditions; however, they were done at lower conversions in order to clearly observe the changes in catalytic activity. After each cycle, the catalyst was simply washed with water and methanol to remove surface substances, and then reused in the next run after drying at 100 °C in an oven. As shown in Fig. 6d, the Pt/γ-Al2O3 catalyst can be reused four times, albeit with a little change in catalytic activity and product distribution. Surprisingly, after the first cycle, both glycerol conversion and lactate yield increased marginally from 70-85% and 34–41%, respectively. The catalyst remained stable thereafter until the fourth cycle. On the opposite side, formate yield dropped moderately from 22% to 16%. The substance 1, 2-PDO, a hydrogenation product of glycerol and a competitor with formate for accepting the H2 released from glycerol was increased slightly. In addition, no Pt species was detected by ICP analysis of the solution after the reaction. However, 2.5% and 3.1% Al leaching from the support was noticed after the first and fourth cycles, respectively. Dissolution of the catalyst support in a high-temperature aqueous environment was reported earlier, where up to 2% of the available aluminum was dissolved from Pt/γ-Al2O3 catalyst after 4 h in the water at 200 °C [57]. It was also reported that dissolved aluminum cations can have a catalytic role [57]; therefore, we performed a hot filtration leaching test (see Table S2) to check the effect of dissolved Al species. In this experiment, after 2 h the reaction was stopped, the catalyst was filtered in a hot condition, and then the reaction continued for the next 6 h without catalyst. The lactate yield was almost unchanged while the formate yield decreased from 20% to 13%. This indicates the decomposition of formate in the presence of leached Al species.To rationalize the changes in catalytic activity and product yields, the spent catalyst was characterized using PXRD, STEM, XPS, and NH3-TPD analyses, as shown in Figs. 7–9 , Fig. S6, and Table S3. Fig. 7 shows the PXRD patterns of the fresh and used Pt/γ- Al2O3 catalyst after the first and fourth cycles. It can be seen clearly that amorphous gamma-alumina converted to crystalline boehmite after the first use, and that the crystallinity further increased after consecutive reuses. Due to the high dispersion of Pt on γ-Al2O3, the diffraction peak of Pt was absent; however, it was clearly evident on the boehmite phase at 40 degree 2θ angle. The crystallite size of the metallic Pt calculated from the Scherrer equation was 3.3 and 4.2 nm after the first and fourth cycles, respectively, as denoted in the inset of Fig. 7. This kind of phase change of the Pt/γ-Al2O3 support from γ-Al2O3 to boehmite was also reported previously for aqueous-phase reforming of glycerol, and liquid-phase reforming of lignin at 220 and 225 °C, respectively [58,59].STEM analysis of fresh and used Pt/γ-Al2O3 catalysts suggested the aggregation of Pt nanoparticles after the first recycle test (Fig. 8). The size of the Pt nanoparticles increased from 1.7 nm to 5.2 and 5.6 nm after the first and fourth cycles, respectively.The surface chemical states of platinum in the fresh and used catalysts were further investigated by XPS (Fig. 9). Although the Pt 4f levels produce the most intense XPS line, this energy region became overshadowed by the presence of a very strong Al 2p peak from the support. Consequently, the energy region of the less intense Pt 4d peak was recorded. The binding energies are summarized in Table S3. All three samples show a broad and asymmetric Pt 4d5/2 peak that could be resolved (after curve fitting) into two components with binding energies of 314.2–314.4 and 315.7 eV, corresponding to the Pt0 and Pt2+ species, respectively [60–62]. The fresh catalysts show high content of oxidized Pt species on the surface alumina, probably due to strong Pt-support interaction that hinders reduction [63]. The used catalysts contained a high content of Pt0 species, ascribed to the potential for in situ reduction of Pt2+ in the presence of hydrogen released from glycerol during the reaction. H2 formation during the reaction was confirmed through GC-TCD analysis of the vent gas (Fig. S3). The changes in acidic properties of fresh and spent catalysts were probed by NH3-TPD (Fig. S6). It can be seen that spent catalyst shows weak to moderate acid sites in the temperature region of 40–400 °C as like fresh catalyst, however, in a lower concentration. The concentration of total acid sites decreased from 0.64 to 0.36 mmol/g after the first cycle.Therefore, the results obtained from the characterization of used catalysts demonstrated that change in the catalytic activity and product distribution mainly arose from phase-change on the support, which was responsible for three other consequences (i.e., leaching of Al species into the reaction solution, sintering of Pt nanoparticles and decrease in acidity of the catalyst). Glycerol conversion and lactate yield increases with the enlargement of the Pt particles. This could be correlated with the general fact that the larger the metal-particle size, the lower the undercoordinated sites and the higher the fraction of plane sites. A similar phenomenon was reported by Shimizu et al., where bigger Pt atoms showed higher activity than smaller undercoordinated Pt atoms for oxidant-free dehydrogenation of glycerol to form lactic acid [32]. Moreover, the metallic platinum content is higher in used catalyst than in fresh catalyst, enhancing the glycerol conversion and lactate yield. This is because Pt0 showed higher activity than PtOx did during the catalyst screening test (Table 1, entries 3 and 5). In addition, a decrease in acid sites in the spent catalyst might have also been helped to promote the glycerol conversion and LA yield (since the availability of OH− ions would be higher). On the other hand, the phase change of the support, sintering of the metallic Pt nanoparticles and decrease in acid sites (adsorption of carbonates would be lower) had adverse effects on the formation of formate from carbonate. Moreover, a higher percentage of dissolved Al species in the reaction solution after the first cycle could also be responsible for the reduction in the formate yield as per the results from the hot filtration leach test.In summary, we have extended the work on a very new and challenging topic comprising one-pot conversion of glycerol and a CO2 source into lactic and formic acid salts. Among the several supported noble-metal catalysts tested, Pt/γ-Al2O3 showed higher catalytic activity and product selectivity. This was due to the high dispersibility of Pt nanoparticles and to the higher acidity of gamma-alumina support. This probably helps to form strong interactions between glycerol and the carbonate molecules on its surface. Several reaction parameters were optimized and up to 50% lactate and 26% formate yields were achieved, most importantly without using any additional strong base and external H2, which otherwise a mandatory requirement to obtain high lactic and formic acid salts from glycerol and CO2 source, respectively. The catalyst was reused for four consecutive cycles with little variation in catalytic activity and product distribution. Phase transfer of support from amorphous gamma-alumina into crystalline boehmite causes aggregation of Pt nanoparticles and leaching of ‘Al’ species in reaction media which subsequently responsible for the change in the catalytic activity of the used catalyst. Our future research work on this topic will be focusing on the development of a robust bimetallic catalyst system that can overcome the issue of catalyst stability and low product yields. By utilizing this novel approach, waste biomass products from bio-refineries can be integrated with CO2 (industrial waste) to produce value-added chemicals which will ultimately help in the improvement of overall bio-refinery economics with considerable beneficial impact on the environment.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 Next Generation Carbon Upcycling Project (2017M1A2A2043143) by the National Research Foundation of Korea, and was partially supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry, and Energy (MOTIE) of the republic of Korea (No. 20202020800330). The authors would like to thank Dr. D. Y. Hong and Dr. D. W. Hwang for helpful discussions.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jiec.2021.06.023.The following are the Supplementary data to this article: Supplementary data 1

Glycerol and carbonates (used as CO2 sources) were simultaneously converted to carboxylic acid salts under mild hydrothermal media over supported Pt catalysts. The dehydrogenation of glycerol produced lactate (LA); at the same time, hydrogen molecules released from glycerol were effectively transferred to reduce carbonate or bicarbonate ions to formate (FA). Several reaction parameters, including temperature, time, glycerol and carbonate concentration, water amount, catalyst loading, and CO2 source were evaluated. Under the optimized reaction conditions, ~50% yield of LA from glycerol and ~26% yield of FA from potassium carbonate were achieved concomitantly over Pt/γ-Al2O3 catalyst. Importantly, this was done without using external H2 or additional strong base. The textural, structural, and chemical properties of the catalysts were evaluated using N2 adsorption–desorption, powder X-ray diffractometry (PXRD), inductively coupled plasma-atomic emission spectrometry, scanning transmission electron microscopy (STEM), temperature programmed reduction, and temperature programmed desorption analysis. The catalyst was reused for four consecutive cycles with little variation in catalytic activity and product distribution. Used catalysts were further characterized using, PXRD, STEM, and X-ray photoelectron spectral analysis to better understand the structural and chemical changes that occurred in the recycled catalysts, and the factors governing change in the catalytic activity. A plausible reaction pathway was proposed based on the catalytic results and the product distribution data obtained.

Data will be made available on request.Urban centers around the planet face problems generated by the high level of polluting gases and environmental deterioration which in conjunction with the depletion of fossil fuels and the growing energy demand, have led to the search for alternative energy sources through more efficient systems and sustainable processes [1–2]. In this context, the use of biomass has begun to spread since it turns out to be an available and sustainable alternative. Biomass comprises materials derived from plants or animals and is considered a renewable energy source, some examples of biomass are wood, crops, animal manure, among others. In particular, lignocellulosic biomass represents the most renewable and abundant carbon resources and is recognized as the most sustainable alternative to fossil resources [3], it can be divided into three categories, softwood, hardwood and grass, and is made up of cellulose (40–60 %), hemicellulose (10–40 %) and lignin (15–30 %) [4]. Especially, lignin is an important component of lignocellulosic biomass and due to its aromatic distinctiveness, this biopolymer becomes the perfect raw material to obtain low molecular weight aromatic substances. Lignins are phenylpropanoid polymers made up of three main units: syringyl, guaiacyl and hydroxyphenyl, which are recognized as the most abundant renewable aromatic carbon source on earth and their effective utilization is critical for the accelerated development of lignocellulosic biorefinery [5].Although biomass seems unlikely to replace oil, it has great potential as raw material for the generation of high-value chemical products in industry. The ideal depolymerization process involves the selective degradation of lignin into monomeric products [6].A great variety of techniques can be found for the depolymerization of lignin, such as pyrolysis, hydrocracking, hydrogenolysis, hydrolysis and oxidation. To know, the most studied strategies have been the oxidative approach, the solvolytic (basic or acid) and the pyrolytic process. Unfortunately, many of these methods remove too much oxygen and/or disrupt the aromatic ring to produce low-value chemicals.Hydrogenolysis is considered a promising method for the efficient depolymerization of lignin and lignin fragments since it requires less severe reaction conditions, and a moderate yield of monomers can be obtained. Heterogeneous hydrogenolysis catalysts based on transition metals (for example, Fe [7], Co [8] and Ni [9–10]) which are abundant on earth are actively explored with the aim to replace precious metal catalysts. Recently, Liu et al. reported the use of Ni@ZIF-8 catalyst in the reductive catalytic fractionation (RCF) of eucalyptus sawdust to produce phenolic compounds and β-O-4 structures using hydrogen (3 MPa) at 220–260 °C [9]. On the other hand, Li et al. found that nickel single atom catalysts exhibited twice higher activity in lignin depolymerization compared to nickel cluster catalysts, Scheme 1 [11]. However, the preparation of noble metal-free single-atom catalysts continues to be a challenge in metal anchorage, while for their part, avoiding the single-sites agglomeration that favors the nanoparticle formation encompasses another unmet need. Additionally, atomically dispersed supported metal catalysts maximize the efficiency of metal utilization and act as a robust support for metal nanoparticles [12]. MNC (Fe, Co, Ni) single-atom catalysts (SACs) with metals strongly linked to surrounding N atoms have shown promising performances in electrochemical [13] and thermochemical reactions [14].On the other hand, metal organic frameworks have emerged as a class of promising materials given their structure and adsorption capabilities [15] and their use as gas adsorbent materials, sensors [16] and catalysts [17–18], as previously described.In this study, we prepared and characterized catalysts based on NiNC material to evaluate their activity in transfer hydrogenolysis reactions using formic acid as hydrogen source finding excellent activity and selectivity in lignin-derived aryl ethers and kraft lignin. Our results in the catalytic hydrogenolysis reactions of guaiacylglycerol-β-guaiacyl and kraft lignin using nickel catalysts evidenced that nanometric and sub nanometric species are coexisting and having a synergic effect on the reaction. Additionally, zinc chloride and scandium triflate Lewis acids successfully catalyzed the hydrogenolysis of 1-(o-tolyloxy)propan-2-ol, a compound obtained from the nickel-catalyzed reaction of guaiacol and propylene carbonate. Overall, these results provide a synergic catalytic strategy to achieve selective hydrogenolysis of CO bonds in guaiacylglycerol-β-guaiacyl, 1-(o-tolyloxy)propan-2-ol, and kraft lignin. Ni-1 catalyst was prepared from Ni(OAc)2·4H2O and 1,10-phenanthroline ligand, followed by pyrolysis at 600 °C for 2 h and subsequent acid leaching as schematically detailed in Fig. 1 a, similar to the procedure reported by Wang et al., who described acid-resistant NiNC single-atom catalyst (SAC) to hydrogenation of some unsaturated cellulose-derived substrates [19].However, we found nickel nanoparticles formation of diameters larger than 5 nm which suggests the existence of sub nanometric species, Fig. 1b. As previously described, the presence of nanoparticles may not exclude the presence of single atoms, where they in fact coexist, it results difficult to obtain atomically dispersed species since they hold a high surface energy thus tending to agglomerate as clusters or nanoparticles in order to decrease such energy [20–21].In PXRD patterns, carbon-supported Ni materials have a broad shoulder diffraction peak around 25° due to the graphitized carbon support.In PXRD patterns, carbon-supported Ni materials have a broad shoulder diffraction peak around 25° due to the graphitized carbon support. In the case of Ni-1, other XRD signals are not detected due to the low metal loadings in single atoms or small clusters [22], whereas Ni/C material has a diffraction peak attributed to NiO (220) and metallic Ni [(111), (200), (220), (220); JCPDF No. 65–0380] [23–24], Fig. 2 a. On the other hand, EPR spectra of Ni-1 exhibited a signal at g = 2.003 due to carbon radicals coming from the material support at 77 K, while the Ni/C spectral profile is indicative of nanoparticles formation, Fig. 2b. To explore the chemical environment and bonding to Ni-1, which has the highest activity in the hydrogenolysis reaction, we carried out an X-ray photoelectron spectroscopy (XPS) analysis. The fitted Ni2p and N1s spectra are shown in Fig. 2c.Moreover, N1s spectrum can be fitted by four peaks indicating the presence of pyridinic N (398.89 eV), pyrrolic N (400.24 eV), graphitic N (402.14 eV) and, oxidized N (405.5 eV) species [25–26]. The wt.% of nickel in Ni-1 was found to be 4.14 wt%, calculated by XPS.We then proceeded to assess the evaluation of Ni-1 and other nickel sources toward guaiacylglycerol-β-guaiacyl ether (L1), an abundant linkage in lignin. Table 1 shows the L1 transfer hydrogenolysis with formic acid as a hydrogen donor at 150 °C for 2 h. Ni-1 showed the highest activity in CO cleavage compared to Ni/C and Ni/Zn-2-Meth. Ni-1 catalyzed the CO bond cleavage selectively to guaiacol and isoeugenol (2-metoxy-4-[(E)-1-prophenyl] phenol) using an ethanol/H2O solvent system. When the reaction time surpassed 24 h, formation of 2-methoxy-4-propylphenol product was observed, entry 6.[a] All yields were determined by GC–MS with dibenzothiophene as standard. [b] Ni:MgO ratio 1:160. [c] for 24 h.Based on these results and previously described works [27], we propose a reaction pathway for L1 hydrogenolysis to produce guaiacol and isoeugenol. First, a dehydration reaction of L1 occurs which is promoted by medium acidity. Then, compound A hydrogenation is reached by formic acid as hydrogen source, this reaction is catalyzed by Ni-1 to generate the alcohol function (compound D) and guaiacol, while subsequent dehydration and hydrogenation steps lead to the isoeugenol synthesis, Scheme 2 . Subsequently, the reaction between guaiacol and the product of propylene carbonate decarboxylation catalyzed by Zn/Ni-2-Meth and ZIF-8 to form 1-(o-tolyloxy)propan-2-ol was performed, thus investigating the reaction scheme with different phenol-derived substrates. Fortunately, we obtained excellent conversion values, above 90 % as shown in Table 2 . As mentioned, to compare with the guaiacol and propylene carbonate reaction scope, the phenolic substituents here explored were 2-metoxy-5-methylphenol, 2,6-dimethoxyphenol, 3-(dimethylamino)phenol, 2-metoxy-4-(2-propyl)phenol and 2-amino-4-tert-butylphenol, where the Ni/Zn-2-Meth catalyst successfully reproduced the reaction conversions above 90 % regardless of the substituent nature. Notably, the synthesis of these compounds was carried out under neat conditions since the reagents also acted as reaction solvents. As can be seen, those reaction products in the phenol reaction scope study were analogous to product C.It is proposed that in the first step of the reaction mechanism a phenol deprotonation occurs, whose species undergo a nucleophilic attack on propylene carbonate and promotes the ring-opening, Scheme 3 .In addition, the hydrogenolysis of lignin-related molecules in the presence of Lewis acids (LAs) has been described, exerting a synergic effect with palladium compounds in the monomer formation from lignin [28–29] Motivated by these results and the ability of LAs to effectively polarize CO bonds making them more susceptible to hydrogenolysis, we then turned to explore the effect of different LAs in the CO bond cleavage of lignin and lignin model molecules, Scheme 4 . Previously, Abu-Omar et al. described the synergic effect between Pd/C and Zn2+ in lignin hydrodeoxygenation [30]. Benzodioxanes can be obtained from 1-(o-tolyloxy)propan-2-ol with substoichiometric amounts of scandium triflate, the formation of guaiacol from the breaking of the CO bond requires the presence of formic acid, the disadvantage of using Lewis acids in the hydrogenolysis of L1 molecule relates to the inherent generation of multiple compounds without any selectivity. Tabanelli et al. prepared 2-hydroxymethyl-1,4-benzodioxane through the decarboxylation of cyclic carbonates and catechol as nucleophile in basic medium [31]. Additionally, Cui et al. described the use of NiAlOx catalyst in the presence of La(OTf)3 to produce aryl–alkyl derivatives through the reductive hydrogenolysis and hydrogenation under hydrogen pressure (4 bar) [32].Results of CO bond cleavage for 1-(o-tolyloxy)propan-2-ol with ZnCl2 are shown in Table 3 . However, only the esterification product, 7 % conversion of guaiacol and traces of benzodioxane were obtained, entry 4. On the other hand, when we changed the Lewis acid to Sc(OTf)3 it was possible to obtain the CO bond cleavage product above 50 % conversion in addition to increasing the yield of benzodioxanes formation.On the other hand, if comparing entries 1 and 3 (Table 4 ), it is observed that although they have the same reaction conditions, the solvent effect is notorious because PC (entry 1) tends to favor compound C2 contrasted to toluene (entry 3) which continues to favor the production of guaiacol. When examining entries 1, 4 and 5 we have the reaction times as the major difference resulting in different products. Further, as observed from entries 4 and 5, significant selectivity for guaiacol and compounds C1 and C2 (G:C1:C2) was obtained. In fact, at larger times (6 h) the guaiacol product was obtained in a higher conversion and no C1 esterification product was observed. To note from entry 1, having a time of 6 h translates into a higher product conversion, highlighting the formation of both, guaiacol in a greater extent, and compound C2. Here, in is worth mentioning the relevant role of formic acid as a hydrogen source, as concluded from entry 7.So far, unique results have been obtained for the hydrogenolysis of guaiacylglycerol-β-guaiacyl, particularly, hydrogenolysis of kraft lignin conducted with Ni-1 and formic acid in water/EtOH mainly produced guaiacol and various monomer products of lignin depolymerization (Fig. 3 ). Optimization for lignin and/or lignin kraft hydrogenolysis reactions has yet to be found, preliminary studies indicate that a change in lignin structure occurs, hydrogenolysis of different lignins are currently under way. Notably, the catalytic hydrogenolysis of pinus lignin was carried out with Ni-1 using formic acid as hydrogen source at 150 °C for 2 h. The gel permeation chromatography (GPC) analysis shows a Mw = 2930 corresponding to dioxasolv lignin while the Ni-1 catalyst utilized in the depolymerization reaction of lignin favored the formation of an oil product with a Mw = 1790, and significant amounts of lignin monomers, Fig. 4 .In summary, we prepared and evaluated different nickel catalysts, where Ni-1 showcased the higher activity and selectivity for the β-O-4 bond cleavage in Lignin-Derived Aryl Ethers such as guaiacylglycerol-β-guaiacyl ether to obtain guaiacol and isoeugenol selectively. Interestingly, the CO bond cleavage of alkyl-aryl ether compounds was demonstrated with scandium triflate, which in combination with the Ni-1 catalyst, a higher degree of kraft lignin depolymerization was achieved. Overall, the CO bond cleavage represents a crucial step in the valorization of lignin and lignin-related molecules, here, the development of more robust and versatile catalyst will impact the chemical space exploration through sustainable methodologies. The utility of the Ni-1 catalyst here described certainly provides new opportunities in the transfer hydrogenolysis reactions of lignin derivatives. Finally, the role of the Lewis acid in native lignin was crucial in the CO bond cleavage, assisting the lignin depolymerization into monomer products.All manipulations were conducted under an argon atmosphere unless otherwise specified.All chemicals were commercially obtained and used without additional purification. 4́hydroxy-3́-metoxyacetophenone, 2,6-dimethoxyphenol, 99 %; guaiacol; 1,10-phenanthroline ≥ 99 %; 3,4-dimetoxyacetophenone, 97 %; nickel(II) acetate tetrahydrate, Ni(OAc)2·4H2O 98 %, methanol, 98 %; anhydrous toluene, 99.8 %, anhydrous propylene carbonate, 99.9 %; formic acid ≥ 95 %, zinc chloride, ≥95 %, scandium (III) triflate, 99 %, nickel nitrate hexahydrate, puriss ≥ 98.5 %; (Ni(NO3)2·6H2O), magnesium oxide nanopowder, 2-methylimidazole, 99 % and zinc nitrate hexahydrate, purum ((ZnNO3)2·6H2O)), anhydrous ethanol, methylene chloride, anhydrous 1,4-dioxane were purchased from Sigma Aldrich. ZIF-8 [33], Zn/Ni-2-Meth [9]and Ni/C [34] were prepared according to reported procedures. L1 [35] were prepared according to previously reported procedures without further modification. Some experiments were carried out in an Anton Parr Monowave 50 + P apparatus. Ni-1 catalyst was prepared similarly to a procedure reported by Wang group [19] with some modifications. Briefly, for Ni-1 a 100 mL Schlenk flask was charged with Ni(OAc)2·4H2O (0.25 mmol) and 1,10-phenanthroline (0.75 mmol), then 25 mL of anhydrous ethanol was added and sonicated for 10 min at room temperature. To this solution, 1.6 g of MgO were added and sonicated for 10 min. The resulting suspension was stirred at 60 °C for 12 h. The solvent was removed under reduced pressure and dried for 24 h in vacuum. The resulting solid was ground in an Agate mortar, and the fine powder was transferred to a glass ampule and sealed under vacuum, then the ampule was pyrolyzed at 600 °C for 2 h. The obtained black solid was stirred with 100 mL of 0.5 M H2SO4 at 80 °C overnight to remove MgO support. The black solid was washed with deionized water and dried under vacuum for 24 h.In a typical experiment, a 25 mL Schlenk flask equipped with a Rotaflo valve, and a magnetic stirring bar was loaded with 0.038 g of L1 (0.12 mmol), 20 mg of Ni-1 and 2 mL of deionized water and 2 mL of ethanol. Then 14.5 μL of FA (3 equiv. 0.36 mmol) and finally 82.4 μL (5 equiv.0.6 mmol) of NEt3 were added. The flask was heated at 150 °C for 24 h and the solvent was removed with vacuum for 6 h, then the mixture was dissolved with methylene chloride for GC–MS quantification.A 25 mL Schlenk flask equipped with a Rotaflo valve and a stirring bar, was charged with guaiacol (0.58 mL), propylene carbonate solvent (0.5 mL) and Ni/Zn-2-Meth catalyst (12 mg, 1 mol%) under argon atmosphere and placed in an oil bath for 24 h at 150 °C. At the end of the heating time a TLC was taken to qualitatively observe the products, its composition was analyzed by NMR and CG-MS. Column chromatography using hexanes: EtOAc, 7.5:2.5 v/v was subsequently carried out to isolate the 1-(2-methoxyphenoxy) propan-2-ol oil (C). Finally, the solvent was evaporated under vacuum until compound C was obtained. Compound C. Colorless oil. 1-(o-tolyloxy)propan-2-ol. 1H NMR (400 MHz, chloroform‑d1 ) δ 6.87 – 6.73 (m,4H), 4.20 (d, J = 3.0 Hz, 1H), 4.12 (qd, J = 6.5, 3.3 Hz, 1H), 3.82 (dd, J = 9.5, 3.6 Hz, 1H), 3.75 (dd, J = 9.6, 7.5 Hz, 1H), 3.70 (s, 3H), 1.19 (d, J = 6.4 Hz, 3H).13C NMR (101 MHz, CDCl3) δ 149.81, 148.19, 121.98, 121.48, 121.05, 115.01, 113.72, 112.11, 111.92, 77.48, 77.16, 76.84, 75.65, 65.93, 55.80, 18.53. IR (ATR, cm−1): 3476, 2970, 2930, 1795, 1593, 1503, 1454, 1250, 1222, 1178, 1122, 936, 740. Mass (EI) m/z:182.Reactions were carried out in 25 mL Schlenk flasks equipped with a magnetic stirring bar and a Teflon valve, which were charged with 1-(2-methoxyphenoxy) propan-2-ol (60 mg), propylene carbonate solvent (2 mL) and Ni/Zn-2-Meth catalyst (5.8 mg, 1 mol%), under argon atmosphere. The Schlenk was placed in an oil bath for 24 h and 150 °C, at the end of the heating time a TLC was taken to qualitatively observe the products, the solvent was evaporated under vacuum. The crude product composition was analyzed by GC–MS.Dioxasolv lignin was isolated based on previously reported methods [35]. Briefly, 30 g of softwood sawdust was extracted with 1,4-dioxane/water (8:2, 200 mL) containing 0.1 M HCl at reflux for 1 h. The mixture was then cooled, filtered, and concentrated in vacuum. The resulting oil was dissolved in acetone/water (8:2) and precipitated in water (10 vols). The resulting powder was collected by filtration and air dried. The crude lignin was then dissolved in acetone/methanol (9:1) and precipitated in diethyl ether (10 vols). The purified lignin was collected by filtration and air dried to give the softwood dioxasolv lignin (0.403 g).In a typical experiment, a 10 mL borosilicate vial equipped with a magnetic stirring bar was loaded with 0.050 g of dioxasolv lignin or kraft lignin, 20 mg of Ni-1 and 2 mL of deionized water and 2 mL of ethanol. Then, 14.5 μL of FA (3 equiv. 0.36 mmol) and finally 82.4 μL (5 equiv.0.6 mmol) of NEt3 were added. The flask was heated at 150 °C for 2 h, and the solvent was removed with vacuum for 6 h, then the mixture was dissolved with DMSO‑d 6 for NMR characterization.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 acknowledge the financial support from PAPIIT TA200121. We are grateful with Prof. Diego Solis-Ibarra and Prof. Edilso Reguera Ruiz for granting us access to several instruments. We are also grateful to Virginia Gómez-Vidales (EPR), M. León, and E. Tapia from Laboratorio Nacional de Ciencias Patrimonio Cultural LANCIC-IQ-UNAM, CONACYT (LN 232619, LN 260779, LN 279740, LN 293904, LN 271614 y LN 293904), Josue Romero (LUME), Elizabeth Huerta Salazar (NMR), Elizabeth Hernández Álvarez (ICP-MS, Instituto de Geofísica), Adriana Romo (ATR, Instituto de Química) and Salvador López (GPC analysis, IIM) for technical assistance.Supplementary data to this article can be found online at https://doi.org/10.1016/j.rechem.2022.100729.The following are the Supplementary data to this article: Supplementary data 1

Renewable aromatic carbon sources and their transformation into value-added chemicals represent one of the most promising approaches of sustainable development. Here, we prepared and characterized nickel catalysts and evaluated their activity in transfer hydrogenolysis reactions using formic acid as hydrogen donor, finding excellent activity and high selectivity to CO bond cleavage in lignin and lignin model compounds using water/ethanol or propylene carbonate as solvents. Particularly, a nickel catalyst named Ni-1, was used in the guaiacylglycerol-β-guaiacyl hydrogenolysis to selectively obtain guaiacol and isoeugenol. Thus, in the case of Ni-1 (NiIINx) we found that small clusters and nanoparticles of 3 –5 nm diameters coexist in the material. Further, the Sc(OTf)3 Lewis acid (LA) successfully promoted the CO bond cleavage of 1-(o-tolyloxy)propan-2-ol to obtain guaiacol. Additionally, we describe the synthesis of benzodioxanes catalyzed by scandium triflate from 1-(o-tolyloxy)propan-2-ol as product of the propylene carbonate decarboxylation catalyzed by ZIF-8.

Use of natural gas as an alternative vehicle fuel to gasoline and diesel is growing worldwide due to the fact that it produces significantly less emissions of greenhouse gases, while being cheaper and safer [1]. However, more technological efforts are required to increase the efficiency of catalytic converters to reduce the trace amounts of unburned methane in the exhaust gases due to its notable environmental impact. It is widely accepted that palladium catalysts are the most active candidates for methane oxidation [2]. Alternatively, the main cheaper noble metal-free substitutes are cobalt-based catalysts, namely those based on cobalt oxides such as the spinel-type Co3O4 due to the remarkable mobility of its oxygen species [3,4]. This material has already been extensively investigated for numerous applications such as CO oxidation [5,6], N2O abatement [7,8] or oxidation of VOCs [9] and soot [10].Most studies on the design of efficient catalysts are referred to powdered or pelleted systems although the real implementation in natural gas vehicles would need a more suitable catalyst geometry that minimises gas flow resistance and facilitates intensification of the catalytic process of lean methane oxidation. Monolith catalysts are usually the preferred option due to their good thermal and mechanical resistance [11,12]. However, an alternative solution has lately appeared in the form of open cell foams made of ceramic or metallic materials, which are characterised by a cellular structure with interconnected and often non-ordered pores with a large volume [13]. Typically, only 5–25% of the total volume of the foam is the base material. The alleged advantage of this type of structured substrates when compared with more conventional monoliths lies on their high surface/volume ratio and random disposition of the void volume, which can aid in the mass and heat transfer between the gas and the solid phase and allow rector operation at relatively high flow rates [14]. The use of foam-supported catalysts is currently focused on both pollution abatement processes (catalytic converters) [15–17] or conventional catalytic processes such as methane reforming or CO2 methanation [18–20].The incorporation of a powdered catalyst onto a structured support can be mainly carried out by two methodologies. The most commonly applied procedure on an industrial scale is to prepare a washcoating slurry with the powdered catalyst and a fluid phase such as water or a water/glycerine mixture. The structured support is then dipped into the slurry until it is thoroughly coated. Next, the samples is dried and calcined to stabilise the catalytic phase material [21]. A second approach involves applying impregnation-based routes to deposit the catalyst formulation directly onto the surface of the structured support. The most frequently used methods in this case are wet impregnation (often in the presence of some surfactants) and solution combustion synthesis (SCS) [22,23]. Essentially, SCS is understood as a self-sustained reaction of metal nitrates and an organic fuel with varying chemical nature, which induces a high-temperature reaction between fuel and oxygen-containing species derived from the decomposition of the nitrates. This methodology entails a series of advantages. For instance, in addition to avoiding the intermediate and time-consuming steps of washcoating, the SCS route usually leads to well crystallised nanosized clusters after thermal stabilisation [24,25]. In this sense, when evaluating cobalt catalysts supported on α-Al2O3 coated monoliths for N2O decomposition, Wójcik et al. [26] evidenced a better catalytic performance of deposited Co3O4 by SCS with respect to conventional impregnation. The key operational parameters of the SCS route are basically the selection of the fuel and the appropriate fuel-to-oxidiser (metallic nitrates) ratio, which is typically denoted as Φ. These two factors strongly influence the mechanism of the combustion process and, in turn, the morphological properties of the active phase.In this work attention was paid to analysing the use of two different fuels, namely urea and glycine, since these are cheap and readily available commercially, while the Φ ratio was varied from 0.25 to 1.0, which corresponded to 25–100% stoichiometric amount of fuel, respectively. An α-Al2O3 open cell foam was chosen due to its stability at relatively high temperatures and chemical inertness. Based on our previous study [27] dealing with the design of α-Al2O3 supported Co-Ce powdered catalysts for lean methane oxidation, the selected active phase was Co3O4 with a loading of 10%wt modified with controlled amounts of cerium as a promoter (Ce/Co molar ratio of 0.05). Both active phases were simultaneously incorporated in the same SCS step. The set of structured catalysts prepared by solid combustion synthesis was examined in the oxidation of lean methane under realistic conditions (relatively high space velocity and simultaneous presence of notable amounts of H2O and CO2 in the flue gas) for a prolonged reaction time interval (285 h at 550 ºC). Catalytic results were kinetically analysed in terms of the reaction rate normalised to the Co3O4 mass for a selected reaction temperature (400 ºC).An α-Al2O3-based open cell foam (Lanik, s.r.o., 45 ppi, length = 30 mm and diameter = 8 mm) was selected as the structured support. Table S1, Supplementary Material, summarises the main geometric properties of the foam substrate. The average strut thickness (0.42 mm) and pore size (1 mm) were estimated from various SEM micrographs similar to those shown in Fig. S1, Supplementary Material. The calculated porosity or voidage was 0.78. The followed procedure for estimating this physical parameter, which depends on the average strut thickness and the average pore diameter, is detailed in the Supplementary Material. On the other hand, it should be noted that, in addition to alumina, appreciable amounts of silica (18%wt.) and magnesia (1%wt.) were present as determined by WDXRF.The foam catalysts were synthesised by solid combustion synthesis using urea and glycine as fuels. Samples were prepared with varying fuel/oxidiser ratio (Φ), namely 0.25, 0.50, 0.75 and 1.00. The selected cobalt loading was 10%wt.Co3O4 with a Ce/Co molar ratio of 0.05 that was equivalent to a 1 wt.CeO2%. The SCS impregnation gel was an aqueous solution of cobalt nitrate hexahydrate (Co(NO3)2.6 H2O) 0.4 M and cerium nitrate hexahydrate (Ce(NO3)3.6 H2O) 0.02 M in which adjusted amounts of the used fuel for the various selected Φ ratios were dissolved. In all cases, the open cell foams were dipped vertically into 25 ml of the corresponding impregnation solution for 5 min. Then, the excess was removed with compressed air. The impregnated foam was subsequently placed in an oven at 250 °C for 20 min with the aim of inducing the SCS reaction. This coating procedure was repeated several times to reach the desired cobalt and cerium concentration. After the last impregnation step, the coated foams were calcined at 600 °C for 4 h to produce the final catalysts. The samples were labelled as F(U) and F(G) when using urea and glycine, respectively.The chemical reactions that ideally occur between the metal nitrates and the selected fuels during the combustion step as a function of the Φ ratio are the following: Co NO 3 2 + 14 9 Φ CO N H 2 2 + 7 3 Φ − 1 O 2 → 1 3 Co 3 O 4 + 14 9 Φ CO 2 + 28 9 Φ H 2 O + 14 9 Φ + 1 N 2 Ce NO 3 3 + 7 3 Φ CO N H 2 2 + 7 2 Φ − 1 O 2 → Ce O 2 + 7 3 Φ CO 2 + 14 3 Φ H 2 O + 14 6 Φ + 3 2 N 2 Co NO 3 2 + 28 27 Φ C 2 H 5 NO 2 + 7 3 Φ − 1 O 2 → 1 3 Co 3 O 4 + 56 27 Φ CO 2 + 70 27 Φ H 2 O + 14 27 Φ + 1 N 2 Ce NO 3 3 + 14 9 Φ C 2 H 5 NO 2 + 7 2 Φ − 1 O 2 → Ce O 2 + 28 9 Φ CO 2 + 35 9 Φ H 2 O + 7 9 Φ + 3 2 N 2 Thus, when Φ = 0 the reaction corresponds to the simple thermal decomposition of the metallic (cobalt or cerium) nitrate. If Φ = 1, the corresponding stoichiometric redox reaction is then made explicit. Taking into account that the required Co3O4 and CeO2 concentrations were 10%wt and 1%wt., respectively, the extent of the redox reactions involving cobalt nitrate was comparatively more noticeable since a larger amount of this salt was used in the synthesis.The foam catalysts were characterised by a wide number of analytical techniques including scanning electron microscopy (SEM) coupled to energy dispersive X-ray spectroscopy (EDX), scanning transmission electron microscopy - high angle annular dark field (STEM-HAADF) coupled to EDX mapping, inductively coupled plasma atomic emission spectroscopy (ICP-AES), wavelength dispersive X-ray fluorescence (WDXRF), N2 physisorption, X-Ray diffraction (XRD), high resolution transmission electron microscopy (HRTEM), Raman spectroscopy, X-Ray photoelectron spectroscopy (XPS), temperature-programmed reduction with hydrogen (H2-TPR) and temperature programmed reaction with methane (CH4-TPRe). Except for SEM-EDX, the structured catalysts were crushed and milled to a fine powder before analysis. Although the experimental details are included elsewhere [28,29], some relevant details on the characterisation details are given below.Scanning Electron Microscopy images were obtained in a JEOL JSM-7000 F Schottky-type field emission microscope operated at 10 kV. The electronic microscope was equipped with a INCA X-sight Si(Li) series pentaFET EDX detector to allow for elemental analysis of the observed surfaces. On the other hand, High-Resolution Transmission Electron Microscopy and Scanning Transmission Electron Microscopy images were obtained with a in a Cs-image-corrected Titan (Thermofisher Scientific) at a working voltage of 300 kV with a 2k x 2k Ultrascan CCD camera (Gatan) positioned before the filter for TEM imaging (energy resolution of 0.7 eV). The microscope was equipped with a CCD camera (Gatan), a HAADF detector (Fischione) and an Ultim Max detector (Oxford Instruments) that allowed for EDX elemental mapping.The elemental composition of the synthesised catalysts was determined by ICP-AES, using a Thermo Elemental Iris Intrepid apparatus, and WDXRF with a PANalytical AXIOS sequential spectrometer. The textural properties, in terms of specific surface area (BET method) and pore volume (BJH method), were determined by nitrogen physisorption at − 196 °C in a Micromeritics TriStar II apparatus. Before the analysis, outgassing of the samples was carried out on a Micromeritics SmartPrep apparatus at 300 ºC for 10 h with a N2 flow.XRD analysis were carried out using Cu Kα radiation (λ = 1.5406 Å) on a X′PERT-PRO X-Ray diffractometer equipped with a Ni filter and operated at 40 kV and 40 mA. The samples were scanned from an initial value of 2θ = 5° to a final value of 2θ = 80°, with a step size of 0.026° and a counting time of 26.8 s. From the obtained diffractograms, the cell parameter of the Co3O4 phase was obtained by full profile matching using FullProf.2k software. On the other hand, the Raman spectra of the samples were obtained with Renishaw InVia Raman spectrometer, coupled to a Leica DMLM microscope, with an ion-argon laser (Modu-Laser, 514 nm). For each sample, five scans in the spectral window of 150–900 cm−1 and a spatial resolution of 2 µm were accumulated. Finally, XPS measurements were performed in a Kratos AXIS Supra spectrometer using a 225 W Al Kα radiation source with a pass energy of 20 eV.The redox properties of the catalysts were investigated on a Micromeritics Autochem 2920 apparatus coupled to a TCD detector by means of Temperature-Programmed Reduction with hydrogen (H2-TPR) and Temperature-Programmed Reaction with methane (CH4-TPRe). In both cases an initial pre-treatment step with a 5%O2/He mixture at 300 °C for 30 min was performed with the aim of removing impurities form the surface of the samples while at the same time fully restoring the oxygen vacancies of the spinel lattice before the analysis of the reducibility. After cooling down to room temperature with flowing He, the experiments were conducted up to 600 °C, with a 5%H2/Ar mixture and a 5%CH4/He mixture, respectively. In the CH4-TPRe the composition of the gaseous stream was monitored with a MKS Cirrus Quadrupole Mass Spectrometer.The efficiency of the foam catalysts for the complete oxidation of dilute methane was examined in a fixed bed quartz tubular (10 mm ID) reactor in the 200–600 °C temperature range with a heating rate of 1 ºC min−1. The runs were carried out with a single piece of foam catalysts (with a mass of 650–700 mg) that were deposited on a glass frit located near the bottom of the reactor tube. The GHSV calculated on the basis of the total volume of the foam catalyst (1.5 ml) was around 4000 h−1. This corresponded to a WHSV of 85 l gCo3O4 −1 h−1. In order to avoid gas channelling each structured catalyst was wrapped with an aluminium foil. Light-off tests were repeated at least three times to assure reproducibility, with an average 12 h of use for each structured catalyst. The composition of the feed stream was 1%CH4/10%O2/89%N2 with a total flow of 100 ml min−1. Note that the typically encountered O2/CH4 molar ratio can vary between 2 and 6 for stoichiometric engines and between 10 and 70 for lean engines. The used ratio in this work falls within this last range. The composition of the reaction gases was continuously analysed by a SRS RGA200 quadrupole mass spectrometer following the m/z = 44 (CO2), 32 (O2), 28 (CO) and 16 (CH4) signals. The analysis of the product stream was carried out in steps of 25 ºC, typically after 15 min on stream. Each analysis was performed in triplicate in order to check reproducibility. A margin of error of less than 1% was found. Methane conversion was determined by the difference between inlet and outlet CH4 molar flows. Additionally, the effect of the presence of water (10–30%vol.) and carbon dioxide (10%vol.) on the catalyst stability with time on stream was investigated at constant temperature (550 ºC) for a total reaction interval of 285 h. The influence of GHSV in the 4000–60,000 h−1 range (85–850 l gCo3O4 −1 h−1) was also studied.The suitability of SCS as an attractive methodology for producing efficient cobalt catalysts for lean methane oxidation was initially addressed. Thus, a bulk Co3O4 oxide was obtained using cobalt nitrate as precursor and glycine as fuel (Φ = 1). As aforementioned, the reactive mixture was heated at 250 °C for 30 min in order to activate the SCS reaction. The resulting sample was then calcined at 600 °C for 4 h. For comparative purposes, a reference Co3O4 catalyst was synthesised by simple calcination of the same cobalt precursor under identical thermal conditions (600 °C/4 h). Both samples were prepared without cerium as promoter. Their performance in the oxidation of methane was examined at 30 l gCo3O4 −1 h−1 in the 200–600 °C temperature range. The composition of the feedstream was 1%CH4/10%O2/89 N2%. Three consecutive light-off runs were recorded. While a slight decrease in activity with temperature was observed in the second run with respect to the first run, the third light-off curve was virtually identical to the second run. Hence, the light-off curves corresponding to the third cycle of each catalyst are shown in Fig. 1. It was found that the sample synthesised with glycine showed a T50 (temperature at which 50% conversion was attained) of 455 °C, while its counterpart required 480 °C. 90% conversion was obtained at 525 and 575 °C, respectively. A significantly higher reaction rate under differential conditions (375 °C) was also noticed (1.2 vs 0.8 mmol CH4 gCo3O4 −1 h−1). It must be pointed out that despite the fact that the reaction rate, on a surface area basis, of the sample prepared by calcination is twice that of the one prepared by SCS, the calcination method cannot produce catalysts with high specific surface areas, and therefore with a large population of active sites.The superior oxidation ability of the sample prepared by SCS was connected with its appreciably better textural properties although no significant difference in the crystallite size were found (84–89 nm). Hence, the increased volume of gases produced during the fuel-assisted combustion process provoked a higher porosity as revealed by the larger surface area (14 vs 5 m2 g−1) and pore volume (0.04 vs 0.02 cm3 g−1) and the smaller mean pore size (170 vs 355 Å). Likewise, a favoured reducibility at low temperatures was observed over the oxide synthesised with glycine as revealed by H2-TPR (Fig. S2, Supplementary Material). It was found that the onset reduction temperature was 280 °C compared with 300 °C. Besides, the H2 uptake at low temperatures (250–325 °C) was markedly larger 3.9 vs 2.6 mmo g−1). In sum, this preliminary catalytic evaluation accompanied by the characterisation of the textural and redox properties evidenced the potential of the combustion route aided by a fuel (glycine) for preparing promising oxidation cobalt catalysts [30,31], and therefore justifies a deeper analysis for its optimisation in the removal of lean methane. As stated earlier, our interest will be now focused on the investigation of this methodology for intensifying the methane oxidation process with highly active Ce-promoted cobalt catalysts supported on open-cell α-Al2O3 foams. In this sense, it must highlighted that the use of low amounts of Ce as an additive has been observed to promote the performance of cobalt catalysts notably since it increases the mobility of active oxygen species [27,28].For defining the number of cycles required to achieve the desired amount of CeO2-modified Co3O4 (approximately 10%wt.Co3O4 and 1%wt.CeO2, which corresponded to a Ce/Co molar ratio of 0.05) loaded onto the foam substrate, the evolution of the Co3O4 oxide mass concentration as function of the number of cycles is shown in Fig. S3, Supplementary Material. This graph includes the mean oxide concentration for each cycle, which was estimated from gravimetric measurements by the difference of the coated and base foams prepared in duplicate with both fuels and the entire Φ range (0.25–1.0). Thus, 16 measurements were averaged for each cycle. Hence, the consecutive cycles led to a gradual increase in Co3O4 concentration from about 1% (1st cycle), 2% (2nd cycle), 6% (3rd cycle) to 10% (4th cycle). It is worth pointing out that irrespective of the synthesis conditions (type of fuel and Φ ratio) the amount of oxides coated in each cycle was quite reproducible. Hence, four cycles were tentatively required to attain the target concentration (10%wt.Co3O4). It must be pointed out that the metal concentration by chemical analysis was not determined after each coating step, since that would have destroyed the sample after the corresponding coating step. Therefore, the actual metallic loadings were determined by ICP-AES was only measured for the foam catalysts coated after four consecutive runs. As will be shown later on, a significantly lower metallic content was found with respect to that expected from gravimetric measurements although the Ce/Co molar ratio was always equal to nominal value (0.05).On the other hand, the adhesion of the catalytic coating onto the open cell foams was examined by ultrasonic treatment. Several samples were submerged in a 50% isopropanol/50% water solution and subjected to sonication at 40 kHz and 200 W for 1 h in a Selecta ULTRASONS-H ultrasonic cleaner. Before and after the test, the samples were dried at 110 °C for 1 h and weighted to measure the mass loss owing to the sonication treatment. These tests revealed a minimal mass loss (0.2–0.4%wt.) after the sonication treatment, thus suggesting that the utilised combustion route was adequate to obtain structured samples with a relatively high mechanical stability of the deposited Ce-Co oxide.SEM analysis of the foam catalysts was carried out to ascertain differences in the morphology and homogeneity of the Ce-Co coating as function of the used fuel. Thus, Fig. 2 and Fig. S4, Supplementary Material include representative SEM images of the two samples prepared with urea and glycine with Φ = 1, along with images of the pristine foam. Note that the surface morphology of the bare foam substrate was rough and consisted of an agglomeration of crystallites with various sizes and shapes, probably due to the foam being a mixture of several ceramic materials. Complementary EDX analysis was performed for semi-quantitatively estimating the elemental composition of the catalyst surface (a depth of 0.5–1 µm). Thus, around 65 spot analyses on selected regions of both samples were carried out. Attention was paid to estimating the Co/Al molar ratio as a criterion for comparing the dispersion of cobalt on the foam, and more importantly, the Ce/Co molar ratio for characterising the contact between these two metals. Fig S5 (Supplementary material) shows the relative distribution of this ratio at the surface of the foam catalysts.After depositing the Ce-Co catalyst with urea, the formation of a distinct catalytic layer could not be distinguished. In fact, the observed structural morphology (x1300 magnification) was rather similar to that of the bare foam. The averaged Co/Al molar ratio derived from EDX was very low, around 0.06, which evidenced a relatively poor accessibility of cobalt species located on the surface of the foam. Although the mean Ce/Co molar ratio (0.06) was close to that determined by ICP-AES (0.05), a great variability in the relative abundance was detected on various regions of the catalyst, which suggested a non-homogenous distribution of these two metals. Hence, Ce-rich areas were identified on some regions (25% and 12% of the spot analysis evidenced a Ce/Co molar ratio higher than 0.06 and 0.1, respectively), while other zones were characterised by a low concentration of cerium species (43% of the spot analysis revealed a Ce/Co molar ratio lower than 0.03). On the other hand, high-magnification SEM images (x25,000–150,000) shown in Fig. 1 revealed that the surface was covered by round patches with sizes ranging 450–500 nm, although some smaller clusters of around 50–200 nm were also visible. Likewise, uncovered areas of foam could be observed.Conversely, when the catalyst was prepared with glycine the surface of the foam substrate was not visible due to being fully covered with a clearly observable catalytic layer, which in addition presented a porous, foamy morphology with large voids in its microstructure. Judging from the images at medium magnification (x1300–1800), the oxides were homogeneously deposited and well anchored on the structured support although superficial debris were also found (Fig. 2). On average, the estimated Co/Al molar ratio derived from spot EDX measurements was around 1.3, substantially higher than that of the urea-based counterpart (0.06). This suggested a better distribution of cobalt on the surface of the foam. On the other hand, a transversal cut from a piece of this catalyst (Fig. S5, Supplementary Material) revealed that the thickness of the catalytic coating was around 7 µm, with a part of the deposited cobalt being able to filter through the pores among the ceramic particles of the foam substrate. By zooming in on the foamy microstructure of the layer (x22,000–150,000) it was observed that it was actually formed by the aggregation of crystallites around 25–30 nm in size. The spongy structure and the relatively small Co3O4 crystallite were assigned to the easier and more violent combustion of the glycine nitrate gel and to the large amount of gases released during the combustion process that simultaneously inhibited sintering and favoured the creation of a porous network [32–34]. As for the relative abundance of cerium and cobalt species, the measured mean Ce/Co molar ratio was 0.06, close to the nominal value (0.05), thereby revealing an intimate mixing of both elements. Interestingly, almost 95% of the spot analysis evidenced a Ce/Co molar ratio between 0.05 and 0.06, which suggested a homogeneous relative distribution of both metals on the surface of the foam.Complementary HAADF-STEM coupled to EDX mapping was useful to determine the differences in the spatial distribution of cobalt and cerium on the surface of these two foam catalysts ( Fig. 3). This analysis required a previous crushing of the foams until obtaining a fine powder. As already noted, the surface of the urea-prepared sample was sparsely covered with bulky patches of catalytic material. Furthermore, both cobalt and cerium species were generally present as isolated entities, with very low mixing between the two metals. Therefore, it was clearly evidenced that the dispersion of both cobalt and cerium was certainly poor. In contrast, in the case of catalyst F(G), the surface of the foam was completely covered with both metals. Moreover, the cerium species exhibited good dispersion and mixing with cobalt, with almost no segregated clusters of ceria. These results evidenced the appreciably better structural properties of the supported catalyst prepared with glycine with respect to the urea-based counterpart.The crushed samples were also investigated by ICP-AES, N2 physisorption, XRD, HRTEM, Raman spectroscopy, XPS, H2-TPR and CH4-TPRe, with the aim of studying the effect of the type of fuel and Φ ratio on the physico-chemical properties of the deposited metal oxides. Firstly, the composition of the samples after the fourth SCS cycle was determined by ICP-AES. The corresponding results are given in Table 1. A slightly lower oxide loading was detected (6.8–8.8%wt.Co3O4 and 0.80–1.01%wt.CeO2) when compared with the estimate given by thermogravimetric measurements. Nevertheless, the actual Ce/Co molar ratio of all foam catalysts determined by chemical analysis was virtually identical to that expected theoretically (0.05). On the other hand, by means of Raman spectroscopy, the possible presence of carbonaceous species derived from the thermal decomposition of organic fuels during the SCS process was examined. Hence, the absence of the signals at about 1340 and 1580 cm−1 assigned to the so-called D and G bands [35] suggested that the combustion reaction of the fuel was complete. Accordingly, the observed mass loss of the samples by dynamic thermogravimetry (10 °C min−1, Setaram Setsys Evolution) up to 900 °C under oxidative conditions was negligible.The textural properties of the ceramic substrate and the two foam catalysts prepared with glycine and urea (Φ = 1) were compared. BET measurements of the bare foam were expected to reveal its macroporous character with a very low surface area (about 0.2 m2 g−1). Interestingly, an appreciable increase in surface area up to 2 m2 g−1 was found for the F(G) catalyst. This finding was consistent with the porosity of the deposited oxide catalyst as observed by SEM analysis. An estimate of the intrinsic surface area of the metallic phase resulted in around 26 m2 g−1. By contrast, the surface area (0.6 m2 g−1) of the F(U) catalyst was close to that of the blank substrate. The X-Ray diffractograms (with a step size of 0.026° and a counting time of 2.0 s) of all structured catalysts are shown in Fig. 4. The diffraction pattern of the bare foam substrate was also included for the sake of comparison. Its pattern was characterised by the intense signals of the trigonal phase of the alpha-alumina support (2θ = 25.7, 37.8, 43.5, 52.6, 57.6, 61.4, 66.6, 68.4 and 77.0°) (ICDD 01–081–1667). However, it must be pointed out that additional signals were noticed, which were assigned to impurities such as mullite (ICDD 01–074–2419) at 2θ = 16.5, 23.7, 26.3, 31.0, 33.3, 35.3, 37.1, 39.3, 41.0, 42.7, 49.5, 54.2, 60.8, 64.7, 70.5 74.4 and 75.2°; cristobalite (ICDD 01–076–0935) at 2θ = 21.9, 46.4, 48.5 and 36.1° and cordierite (ICDD 01–084–1221) at 2θ = 10.5, 28.4 and 29.6°, as can be seen in Fig. S6, Supplementary Material. The set of signals corresponding to the presence of the cobalt spinel oxide (Co3O4) at 2θ = 19.0, 31.3, 36.8, 38.5, 44.8, 59.4 and 65.2° (ICDD 00–042–1467) were identified for all foam catalysts. The presence of CoO or Co that could be formed by reduction of Co3O4 in the presence of the organic fuel was ruled out. Also, a weak signal at 2θ = 28.6°, attributable to the cubic phase of segregated CeO2 (ICDD 00–004–0593), was visible. Accordingly, HRTEM images of the samples prepared with Φ = 1 (Fig. S7, Supplementary Material) allowed the resolution of lattice fringes of Co3O4 (0.29 and 0.24 nm, which corresponded to the {220} and the {311} planes, respectively) and CeO2 crystallites (0.31 and 0.27 nm, which corresponded to the {111} and {200} planes, respectively). The observation of the latter oxide phase suggested that a fraction of cerium was not incorporated into the framework of the cobalt spinel. The mean crystallite size of Co3O4, determined by the Scherrer equation, was around 50 nm for the catalysts prepared with urea and between 19 and 28 nm for the F(G) catalysts (Table 1). Apparently, these sizes were not greatly influenced by the used Φ ratio for each fuel, and were in the same range as those reported by Toniolo et al. [36] for Co3O4 oxides synthesised with glycine (23–37 nm) and urea (50–77 nm) with varying Φ (0.25–1). On the other hand, a comparison of this crystallite size with that estimated by SEM analysis evidenced that the Co3O4 particles observed in the sample prepared with urea, unlike those present in the glycine-based counterpart, were formed by the apparent agglomeration of smaller crystallites.Finally, as indicated above, the introduction of cerium into the lattice of the cobalt spinel was not complete since relatively small crystallites of cerium oxide were observed. The estimated size of CeO2 crystallites ranged between 10 and 12 and 7–8 nm for the F(U) and F(G) catalysts, respectively. The extent of insertion of cerium atoms into the lattice of Co3O4 could be qualitatively evaluated by analysing its cell parameter, as shown in Fig. 5. The cell parameter was calculated via a full profile fitting of the high-resolution diffractograms by using FullProf.2k software. In principle, a larger value could be associated with a greater abundance of cerium atoms given the larger ionic radii of Ce4+ (101 pm) and Ce3+ (115 pm) compared with Co3+ (69 pm) and Co2+ (79 pm). As a reference, the cell parameter of the bulk Co3O4 sample prepared by SCS (glycine) was estimated (8.0988 Å). Regarding the F(U) samples, it was significantly larger (8.1060–8.1070 Å) although no appreciable differences were noticed with varying Φ ratio. However, in the case of the F(G) samples the insertion of cerium atoms was substantially promoted, and highly dependent on the used amount of glycine. Thus, the cell parameter was 8.1210 Å for the sample with Φ = 1. The distortion of the spinel structure caused by cerium as a function of the synthesis conditions (type and amount of fuel) was also followed by Raman spectroscopy. In this way, a shift of the main Raman bands could be taken as an evidence of the extent of this structural change. The Raman spectra of the F(U) and F(G) catalysts are shown in Fig. S8, Supplementary Material. The spectra of the bulk Ce-free Co3O4 sample was also included for comparative purposes as it was taken as a reference to examine the eventual shift corresponding to the Ce-modified foam catalysts. This pure oxide exhibited the five expected vibrations of the Co3O4 lattice, namely three bands at 187, 506 and 602 cm−1 from the F2 g vibration modes, an Eg vibration mode at 462 cm−1 and finally a signal at 667 cm−1 attributed to the A1 g vibration mode [37]. Fig. 4 includes the observed shift of the latter band (A1 g) as a function of the Φ ratio for both fuels. As for the F(U) samples, the shift was similar for all samples (11–12 cm−1), thereby revealing that cerium insertion was not favoured with increasing amounts of urea. However, when using glycine the shift was more marked (13–21 cm− 1). The results evidenced a greater distortion of the lattice, particularly for Φ = 1. To sum up, both XRD and Raman results suggested that the SCS route was suitable for partially doping the lattice of cobalt oxide, and that the extent of cerium insertion and the subsequent lattice distortion was favoured when large amounts (Φ = 0.75–1) of glycine were used. Although the possible insertion of cobalt into the ceria lattice could not be ruled out, this could not be evidenced by XRD and Raman spectroscopy, probably due to the low cerium content of the samples.The surface chemical state of the foam catalysts was investigated by XPS. Hence, the near-surface composition and distribution of cobalt, cerium, and oxygen species was determined by deconvolution and integration of the Co2p3/2 (777–792 eV), Ce3d (881–917 eV) and O1s (526–536 eV) spectra, respectively, as shown in Fig. S9, Supplementary Material. Additionally, the Al2p and Si2p spectra were integrated to include these elements in the surface elemental composition. The surface charging effect in the spectra was compensated against the C-H states in the C 1 s spectra with the energy assumed to be 284.6 eV. As shown in Table 2, it was found that cerium tended to be located preferentially on the surface of the samples with a concentration that was approximately ten times higher (7.8–9.1%wt.) than the corresponding bulk concentration (0.7–0.8%wt.) estimated by ICP-AES. This accumulation could be ascribed to low surface energy of cobalt species compared to ceria, which in turn results in the favoured presence of ceria on the outer surface. Therefore, Ce/Co molar ratios in the 0.22–0.38 range were observed in the foam catalysts.The Co2p3/2 spectra were deconvoluted into three main contributions and two satellites. The first two main components, centred at 779.5 and 780.6 eV, were attributed to the existence of Co3+ and Co2+ ions, respectively, while the third one, centred at 782.3 eV, was assigned to the presence of Co2+ as CoO [38]. Typically, the contribution of the latter component was less than 10% of the total surface Co concentration. This oxide was presumably formed due to in situ partial reduction of surface Co3O4 species under vacuum conditions in the XPS spectrometer. Therefore, it could be assumed that this phase was not present in the catalyst formulation. The signals located at 785.2 and 789.4 eV were identified as the shake-up satellite peaks of the Co2+ and Co3+ ions, respectively [39]. It must be pointed out that the position of the main bands as well as their satellite bands did not vary markedly among all foam catalysts. Nevertheless, the deconvolution of the Co2p3/2 band in the components from Co2+ and Co3+ suggested notable differences in the oxidation state of cobalt on the surface. Hence, as a general behaviour it was observed that the Co3+/Co2+ molar ratio of the F(U) samples was slightly lower (0.80–0.93) compared with the F(G) counterparts (0.83–1.08). It was then possible to establish that smaller crystallites sizes were characterised by the presence of more oxidised cobalt species. On the other hand, it was also remarkable that for the samples prepared with the lowest amount of fuel (Φ = 0.25) were characterised by the lowest Co3+/Co2+ molar ratio. This was reasonably connected with the severe substoichiometric conditions of the combustion synthesis. In contrast, higher amounts of fuel resulted in a favoured presence of Co3+ species.On the other hand, the O1s spectra were deconvoluted into four signals (Fig. S9, Supplementary Material). The first two contributions, located at 529.7 and 531.0 eV, were assigned to the lattice oxygen species from the cobalt oxide and the ceramic substrate, respectively. The third signal, centred at 532.1 eV, was attributed to weakly adsorbed oxygen species on the surface of the samples. Note that these oxygen species could be located indistinctively on the surface of both the ceramic support and the Ce-Co active phase. Finally, the last signal, located at 533.0 eV was attributed to the presence of carbonates, water and hydroxyl species [40]. Given the high ability of lattice oxygen species of Co-based catalysts for methane oxidation [41], its relative abundance was estimated as the Olatt/Otot molar ratio for all samples. The amount of Olatt species was assumed to be proportional to the area under the signal peaking at 529.7 eV. It is noteworthy that the estimated amount of adsorbed oxygen species may be affected by air exposure. However, it is highly likely that this contamination did not result in a remarkable effect on the quantification of the amount of lattice oxygen species, since those are strongly bonded to Co or Ce atoms. Consequently, the Olatt/Ototal molar ratio of all samples would be overestimated. However, since all catalyst exhibited comparable specific surface areas, that overestimation could be assumed to be similar between all samples. Therefore the comparison among the various samples would be meaningful. Regarding the F(U) catalysts, this ratio barely varied with the Φ ratio with values around 0.10–0.11. However, a significantly higher ratio was found for the F(G) samples, which notably depended on the Φ ratio. Thus, it increased from 0.11 (Φ = 0.25) to 0.21 (Φ = 1). In sum, the use of increasing amounts of glycine as fuel favoured the presence of oxygen lattice species in the resulting Ce-Co oxide, which in turn was strongly related to the abundance of Co3+ ions (Fig. S10, Supplementary Material).To define the eventual relationship between the distribution of cerium species and the Co3+/Co2+ molar ratio, the Ce3d spectra of all samples were fitted with eight peaks corresponding to four pairs of spin-orbit doublets (Fig. S9, Supplementary Material). Following the convention adopted by Murugan et al. [42], letters U and V were used to refer to the 3d5/2 and 3d3/2 spin-orbit components, respectively. Of the four pairs of peaks, three of them (namely V, U; V’’, U’’ and V’’’, U’’’) were associated with electrons from Ce4+ while the remaining pair (V′, U′) was attributed to electrons from Ce3+ species. The Ce3+/Ce4+ molar ratios were obtained from the areas of the 3d5/2 and 3d3/2 components for each species. It must be noted that the estimation of this ratio could be affected by the possibility of cerium reduction under the conditions of spectra recording, thereby resulting in an overestimation of the proportion of Ce3+. Although it was difficult to quantify the extent of this eventual reduction, and since the samples were submitted to the same experimental analysis conditions and the ceria particle size of the foam catalysts was relatively similar (7–12 nm), it was assumed that the samples of the same set of foam catalysts (F(G) or F(U)) would exhibit a similar tendency to form Ce3+. Therefore the estimated Ce3+/Ce4+ molar ratios could at least qualitatively compared. It was found that this ratio decreased as the Co3+/Co2+ molar ratio increased, which could be explained in terms of the equilibrium Ce3++Co3+↔Ce4++Co2+ established by the charge balance requirement within the cations of the spinel lattice [43]. Hence, an increase in Co3+ population at the expense of Co2+ resulted in a decrease of Ce3+ ions in favour of Ce4+.H2-TPR analysis was used to characterise the reducibility of the foam catalysts, since this is one of the main parameters governing the performance of Co3O4-based catalysts in redox reactions. A 5%H2/Ar mixture was used as the reducing gas and the experiments were carried out with a heating ramp of 10 ºC min−1 between 50 and 900 ºC. Fig. 6 shows the corresponding reduction patterns (up to 600 ºC) of the samples prepared with the two fuels and varying Φ ratio. It should be pointed out that the observed H2 consumption would correspond to the reduction of deposited cobalt and cerium species, which is expected to occur simultaneously. In order to decouple the reduction process of Co3+, Co2+ and Ce4+ cations, our attention will be first paid to analysing the reducibility of cobalt species. Regardless the synthesis conditions, the reduction process of all catalysts was dominated by a main reduction event at around 350 ºC and a more or less perceptible signal at lower temperatures (300 ºC). Thus, the onset temperature was approximately 250 °C and 300 °C for F(G) and F(U) samples, respectively. These findings were in accordance with the sequential reduction of Co3+ → Co2+ → Co0 [44]. It must be pointed out that above 400 °C no significant H2 uptake was observed, which ruled out the presence of highly stable cobalt species in the form of cobalt aluminate [45]. This was coherent with the high chemical stability of alpha alumina that prevented the formation of this undesired spinel.After integrating the profiles of the F(G) samples ( Table 3) it was found that the total specific H2 uptakes were in all cases higher than those theoretically expected (16.6 mmol gCo3O4 −1), thereby suggesting the appreciable contribution to the overall reducibility of the cerium species present in the samples, mainly inserted in the spinel lattice as Ce4+ cations. The reduction of these species was expected to occur in the same temperature window (250–400 °C) as cobalt species and could be activated due to the transfer of hydrogen by metallic cobalt onto the ceria [46]. The H2 uptake ranged between 16.9 and 17.1 mmol gCo3O4 −1, and apparently depended on the amount of fuel used in the synthesis. These results would be in agreement with a favoured incorporation of cerium with high Φ ratios, as suggested by XRD and Raman spectroscopy. The contribution of Ce4+ reduction to the overall reducibility of the Ce-Co catalysts was evaluated by analysing a post-run sample, particularly the one prepared with Φ = 1, by XPS. Thus, it was observed that its Ce3+/Ce4+ molar ratio substantially increased from 0.23 over the fresh sample to 0.44. On the other hand, the improvement in the redox properties of the samples prepared with increasing amounts of glycine was also reflected in the shift of the reduction temperatures to lower values. Furthermore, when taking a temperature of 300 °C as a criterion, the low-temperature uptake increased with the Φ ratio, from 2.1 mmol gCo3O4 −1 for Φ = 0.25–4.1 mmol gCo3O4 −1 for Φ = 1. As for the F(U) catalysts, their total H2 uptake was comparable (16.7–16.9 mmol gCo3O4 −1), and slightly larger than the theoretical consumption. This suggested that the amount of cerium species in the lattice was comparable irrespective of the Φ ratio, in line with the results given by XRD and Raman spectroscopy. The low-temperature uptake was rather similar (1.5–1.7 mmol gCo3O4 −1) for all F(U) catalysts, and appreciably lower in comparison with their glycine-based counterparts. Analogously, a post-run sample (the one prepared with Φ = 1) was characterised by XPS. In this case, no marked differences in its Ce3+/Ce4+ molar ratio were found (0.35 for the fresh sample and 0.40 for the sample after the H2-TPR run). Finally, it must be pointed out only the diffraction signals of metallic cobalt were clearly distinguished (ICDD 00–015–0806) for both used catalysts. No signals related to cerium species were visible. This was expected due to the low Ce content as CeO2 (lower than 1%wt.CeO2) of the samples.In addition to H2-TPR, the intrinsic reactivity of the oxygen species present in each set of foam catalysts was also characterised by studying the ability of a given sample for oxidising methane (5%CH4/He) in the absence of oxygen with increasing temperature (CH4-TPRe). These experiments were conducted up to 600 ºC with a heating rate of 10 ºC min−1 followed by an isothermal step for 30 min. The amounts of evolved CO2 (m/z = 44) as the main oxidation product and CO (m/z = 28) and H2 (m/z = 2) as by-products derived from possible reforming processes during the run were measured. These results can be helpful in understanding the lean methane oxidation reaction in the light of the widely accepted Mars – van Krevelen mechanism. As shown in Fig. S11 (Supplementary Material), the process is dominated by the large formation of CO2 (and CO and H2, not shown) at 600 °C that corresponded to the full reduction of cobalt species, which eventually catalysed the conversion of methane into syngas and CO2. However, more valuable data could be extracted from the detected production of CO2 at lower temperatures (400–550 °C), since this could be exclusively ascribed to the full oxidation of methane by the active oxygen species of the Co-Ce foam catalysts. Thus, an enlarged view of the CO2 generation profile in this temperature window is included in Fig. 7. It was observed that for the F(U) catalysts the reactivity of oxygen species was relatively similar in view of their comparable peak oxidation temperature around 525–530 °C, except for the sample prepared with a Φ ratio of 0.25 (550 °C). In addition, a comparable oxygen consumption was observed from this set of samples ranging between 0.31 and 0.35 mmol O2 gCo −1. The onset temperature for methane oxidation was in the 415–455 °C range. The onset temperature was defined as the temperature at which 5% of the total CH4 uptake in the low temperature range (below 550 ºC) was consumed. Interestingly, the catalysts prepared with glycine were considerably more active as revealed by the lower onset (380–415 °C) and peak oxidation temperatures (495–535 °C). Also, more appreciable was the amount of oxygen species involved in the oxidation process over these samples that ranged between 0.31 and 0.56 mmol O2 gCo −1. All the results suggested that the use of glycine produced samples with improved properties for methane oxidation that could be associated with the favoured oxygen mobility induced by cerium insertion in the Co3O4 lattice. Thus, the most promising samples were the foam catalysts prepared with high Φ ratios (0.75 and 1.00). Fig. 8 shows the corresponding light-off curves of the oxidation of lean methane (85 l gCo3O4 −1 h−1) over the foam catalysts prepared with each fuel and varying Φ ratio. The GHSV was around 4000 h−1, calculated based on the total volume of structured foam catalyst (1.5 ml). Carbon dioxide and water were the only detected reaction products. The absence of mass and heat transfer limitations within the reactor was checked in order to ensure that they did not affect the obtained kinetic results. Taking into account that the transfer regimes for a structured catalyst significantly varies with respect to their powdered counterparts, four different criteria were checked following the recommendations given by Ercolino et al. [16] and Italiano et al. [47], namely Carberry (external mass transfer), Weisz-Prater (internal mass transfer), Mears (external heat transfer) and Anderson (internal heat transfer) criteria. The mathematical equations related to each criterion are listed in Table S2, Supplementary Material. As an example, corresponding values derived from the estimated reaction data at several temperatures (300–600 °C) for the foam catalyst prepared with glycine and Φ = 1 are included in this table. Judging from the obtained results it was verified that inter- and intra-phase concentration and temperature gradients were negligible below 500–550 ºC. Note that the contribution of heat/mass transfer limitations expectedly was significant at high temperatures (600 ºC) since the estimated values were only one order of magnitude lower with respect to the corresponding threshold.As for the foam catalysts synthesised with urea, an appreciable conversion (10%) was noticed at 400–425 °C. In view of the T50 values ( Table 4) the foam samples with a Φ = 0.75 and Φ = 1.0 showed a similar efficiency with values in the 515–520 °C range, whereas the sample with Φ = 0.5 required 530 °C for this conversion level (50%). Clearly, the poorest performance was shown by the catalyst prepared with the lowest amount of urea (Φ = 0.25). Accordingly, the conversion trend at 600 °C followed the same order, namely, 85% conversion (Φ = 0.75–1.0), 80% conversion (Φ = 0.5) and 70% conversion (Φ = 0.25). By contrast, the use of glycine as a fuel comparatively resulted in markedly more efficient foam catalysts. Hence, at 400–425 °C a conversion as high as 20% was already noticed. No substantial differences were observed among the samples with a Φ ratio of 1.0 (T50 = 450 °C), 0.75 (T50 = 455 °C) and 0.5 (T50 = 460 °C), as shown in Table 4. Thus, these three samples achieved at least 95% conversion at 600 °C. Again, the catalyst prepared with the lowest Φ ratio exhibited a considerably poorer performance (T50 = 500 °C).Despite the differences in activity found among the foam catalysts obtained with both fuels, the apparent activation energies of all examined samples were in the 70–76 kJ mol−1 range. These values were comparable to those exhibited by bulk Co3O4 oxides [48,49], and suggested that the obtained kinetic results were not affected by diffusional limitations, in line with the results reported in Table S2, Supplementary Material. The apparent activation energy was estimated by assuming a first pseudo-order for methane and a zeroth pseudo-order for oxygen [50]. The integral method was applied to estimate the apparent activation energy when considering a first pseudo-order for methane and a zero pseudo-order for oxygen. Conversions between 10% and 90% were fit to the following linearized equation for the integral reactor (Eq. 1) (1) ln [ − ln ( 1 − X ) ] = ln [ k 0 C C H 40 ( W F C H 40 ) ] − E a R T where X is the fractional conversion of methane, k0 is the pre-exponential factor of the Arrhenius equation and W/FCH40 is the weight hourly space velocity. The goodness of the numerical fit is shown in Fig. S12 (Supplementary material).Having proven that glycine was a more suitable fuel for depositing the active phases on the open cell foams by solid combustion synthesis, the determination of the optimal Φ ratio was attempted by comparing the specific reaction rate at a selected temperature of 400 °C. This reaction rate was calculated under differential conditions (conversion<20%). Therefore, it was estimated as the ratio between the experimental conversion and the weight hourly space velocity (W/FCH40). Results included in Table 4 and Fig. 8 revealed a notable dependence of the intrinsic activity with the amount of fuel for the F(G) catalysts. Hence, the normalised reaction rate notably increased from 1.9 (Φ = 0.25) to 3.8 mmol CH4 gCo3O4 h−1 (Φ = 0.5). This promotion, although less noticeable, was also evident with larger amounts of fuel. In this way, the foam catalyst prepared with the highest Φ ratio (Φ = 1) exhibited a reaction rate of 5.3 mmol CH4 gCo3O4 h−1. For comparative purposes, obtained results of the catalysts prepared with urea were included in Fig. 9 as well. As dictated by the light-off curves, the intrinsic activity at 400 °C was remarkably lower, in the 1.3–1.9 mmol CH4 gCo3O4 h−1 range. The differences in performance among the various examined Φ ratios were rather less obvious when using this fuel.A reasonable correlation was found between the low-temperature O2 consumption of the foam catalysts, as determined by CH4-TPRe analysis, and their specific reaction rate ( Fig. 10). This relationship would be the confirmation of the methane oxidation reaction followed a Mars – van Krevelen mechanism, since the catalysts that exhibited larger O2 uptakes due to their favoured mobility of oxygen species evidenced a higher catalytic activity. The reason for this behaviour seemed to lie on the larger abundance of Co3+ ions on the surface of the catalysts prepared with high Φ ratios of glycine, which in turn resulted in a more abundant presence of lattice oxygen species with high mobility, as evidenced by the complementary correlations depicted in Fig. 9 among the normalised reaction rate and the Co3+/Co2+ and Olatt/Otot at the surface. The superior performance of the glycine-based catalysts prepared with glycine with respect to their urea-based counterparts was ultimately associated with a more efficient insertion of cerium into the lattice of the spinel, thus promoting the presence of Co3+ ions within it. This induced a more marked distortion that led to improved redox properties at low temperatures. Structurally the F(G) catalysts also exhibited a well anchored, homogeneous catalytic coating on the surface of the ceramic substrate characterised by a good dispersion of both cobalt and cerium, and a relatively high porosity.The performance of the most active catalyst, namely the sample synthesised with glycine and a Φ = 1, was studied at varying GHSV in the 4000–60,000 h−1 range (equivalent to a WHSV in the 85–1275 l gCo3O4 −1 h−1 range) at 600 ºC. Results included in Fig. S13 (Supplementary material) correspond to the averaged conversion for 4 h in steps of 4000 h−1. Expectedly, a gradual decrease in conversion was found at lower residence times, from 95% at 4000 h−1, to 82% at 16,000 h−1 and 72% at 60,000 h−1. Interestingly, upon returning to the baseline GHSV (4000 h−1) and after a total accumulated time interval of 56 h at 600 °C, the conversion recovered to the same initial value (close to 95%), thereby suggesting a reasonably good thermal stability of the foam catalyst.Additionally, the effect of the presence of water (10%vol.) and carbon dioxide (10%vol.) on the catalyst stability with time on stream was investigated under isothermal conditions (550 ºC) for a total reaction interval of 285 h (85 l gCo3O4 −1 h−1). Firstly, the following feed mixtures were alternated every 25 h: 1%CH4/10%O2/N2 - 1%CH4/10%O2/10%CO2/N2 - 1%CH4/10%O2/N2 - 1%CH4/10%O2/10%H2O/N2 - 1%CH4/10%O2/N2 - 1%CH4/10%O2/10%H2O/10%CO2/N2. Finally, conversion was again recorded under a 1%CH4/10%O2/N2 atmosphere (5 h). The evolution of methane conversion under these reaction conditions is included in Fig. 11. During the first 25 h of operation, the samples showed a marked thermal stability with no evidence of deactivation. Hence, a relatively constant conversion at around 86% was noticed. After the subsequent admission of carbon dioxide to the feedstream during additional 25 h, conversion was hardly affected. Upon returning to base conditions, the same conversion was still maintained. However, the addition of water caused a significant decrease to a stable value of 58% due its adsorption on the catalyst surface. Interestingly, when water was subsequently cut off, the methane conversion was almost fully recovered, with a value similar (82%) to that observed under dry conditions. Thus, it was evidenced that this temporary inhibiting effect of water did not lead to a significant irreversible deactivation of the sample. Finally, attention was paid to examining the effect of the simultaneous presence of carbon dioxide and water for 25 h in an attempt to mimic a real exhaust gas from a natural gas-fuelled engine. Interestingly, the decrease in conversion provoked by water was not accentuated to a greater extent when combined with carbon dioxide, since a mean conversion of 56% was noted. When returning to the base conditions (1%CH4/10%O2/N2) the mean conversion along 5 h was 81%.After the first 155 h with alternating conditions, the influence of the presence of larger amounts of water vapour in the feed stream was then analysed. For this reason, varying concentrations of water vapour, from 10% to 30%vol. were admitted into the reactor during consecutive periods of 25 h. It was found that, despite the high used concentrations, the detrimental effect to the methane conversion was relatively limited. Hence, the average conversions for the various water vapour concentrations were 53% (15%H2O), 50% (20% H2O), 47% (25% H2O) and 45% (30% H2O), thus evidencing that the catalyst was relatively resistant to increased concentrations of water in the feed stream. Moreover, after returning to the base dry conditions, the achieved conversion was 78%, which pointed out that the irreversible deactivation phenomenon was also limited even after exposure of the catalyst to a feedstream containing 30%H2O. It must be pointed out that the eventual formation of CO and H2 derived from reforming processes of methane (steam and/or dry reforming) was not observed. Even in the presence of 10%CO2 and up to 30%H2O the selectivity to CO2 was 100%. In other words, it could be assumed that the reactivity of methane with oxygen (10%) was highly preferential, even when admixtured with water vapour (30%) and/or CO2(10%).The (fresh) catalyst was subjected to a similar stability test as well, but operating at under a higher space velocity, in order to assess the influence of water vapour in conditions closer to those found in real natural gas engines exhausts. During consecutive reaction time intervals of 25 h at 600 ºC, the catalytic performance was evaluated under dry and humid conditions (10–30%H2O) at 4000 h−1 (85 l gCo3O4 −1 h−1) and 40,000 h−1 (850 l gCo3O4 −1 h−1). Results shown in Fig. S14 (Supplementary material) for the first 100 h were in agreement with the previous results on stability (Fig. 10). Hence, the conversion under dry conditions at 4000 h−1 was around 95%, decreasing to 77% when adding 10%H2O and recovering again to the initial value after cutting off the admission of water. When the water concentration was raised to 30%vol. the conversion decreased to 59%. As for the second 100 h-time interval at higher space velocity, the negative effect of the addition of water was found to be less marked with respect to that observed at 4000 h−1, probably due to the water having a shorter residence time to adsorb on the surface of the catalyst. Thus, under the dry conditions the average conversion was 73% and decreased to 58% with 10%H2O and to 48% in the presence of 30%H2O.Finally, the catalyst subjected to the 285 h-stability test (at 550 °C and 4000 h−1 in the presence of H2O and CO2) was characterized in order detect any structural or chemical differences with its fresh counterpart, which could be responsible for the slight deactivation caused by the long term exposure to water vapour. Thus, XRD analysis found no abnormal crystalline phases in the aged catalyst, although the estimated average Co3O4 crystallite size was found to be appreciably larger (32 nm vs. 18 nm). On the other hand, when comparing SEM images (Fig. S15, Supplementary material), a notable deterioration of the superficial structure was detected in the used sample, with multiple cracks and rifts that spread from the numerous pores of the original foamy structure. Images taken at high magnification also confirmed the results given by XRD, with the Co3O4 crystallites exhibiting poorly defined borders and generally larger sizes (35–55 nm). These findings evidenced that exposure to water vapour induced a slight sintering.The spent catalyst was also submitted to CH4-TPRe analysis in order to assess the effect of water vapour ageing on the redox properties and mobility of oxygen species. The profile of CO2 production (m/z = 44) of both fresh and used catalysts, shown in Fig. S16 (Supplementary material), revealed a marked worsening in the reducibility of the used sample, given the increase in both the onset reduction temperature, from 392° to 412 °C, and in the peak reduction temperature, from 495° to 500 °C. However, after integration of the profiles it was found that the low-temperature O2 consumption of both samples was identical (0.56 mmol O2 gCo −1). Thus, the decrease in the reducibility of the used sample was merely a side effect of the aforementioned sintering of the Co3O4 crystallites, and not due to any detrimental effect on the intrinsic chemical properties of the catalyst. This was in line with the identical Ce/Co molar ratio (0.06) found by EDX on the surface of both fresh and used samples.In this work the intensified lean methane oxidation with novel Co3O4(10 wt%)-CeO2(1 wt%) catalysts supported over α-Al2O3 open cell foams was investigated. The structured catalysts were prepared by solution combustion synthesis using urea or glycine as fuel while varying the fuel/oxidiser ratio (Φ ratio) between 0.25 and 1.0. The textural, structural, morphological and redox properties were examined by a wide number of analytical techniques including SEM-EDX, STEM-HAADF coupled to EDX mapping, ICP-AES, WDXRF, N2 physisorption, XRD, HRTEM, Raman spectroscopy, XPS, H2-TPR and CH4-TPRe. The catalytic performance was evaluated under realistic reaction conditions in terms of a relatively high gas hourly space velocity and the simultaneous presence of water and carbon dioxide in the exhaust gas of the natural gas fuelled vehicle.Glycine was found to produce catalysts with a considerably better performance than the urea-based counterparts, with specific reaction rates being around 3 times higher. From a structural and morphological point of view, the reason behind this behaviour was closely related to the intrinsic porosity of the Ce-Co catalytic layer deposited onto the foam substrate, relative dispersion of deposited cobalt and cerium species and the Co3O4 crystallite size resulting from the used type of fuel. In particular, the catalysts prepared with glycine resulted in the formation of a highly porous catalytic coating containing relatively small, well dispersed spherical oxide crystallites. In contrast, the samples synthesised with urea did not lead to the formation of a distinct, homogeneous Ce-Co layer. In fact, some areas of the foam were not fully covered while other areas presented large flat patches of cobalt oxide. Moreover, the intimate mixture of cobalt and cerium when using glycine as fuel allowed for a more efficient insertion of Ce ions into the lattice of the cobalt spinel, which translated into a more favoured presence of Co3+ ions within the Co3O4 structure. This, in turn, led to an increased presence and mobility of the lattice oxygen species that led to a better performance for lean methane oxidation. The optimal fuel/oxidiser ratio for glycine was found to be the stoichiometric one. This foam catalyst exhibited a notable activity even at low residence times. Moreover, this sample showed a marked thermal and hydrothermal stability under isothermal conditions (550–600 ºC). While the catalytic performance was not affected by the presence of carbon dioxide, the observed inhibiting effect of water was found to be almost reversible, although exposure to humid conditions eventually caused an appreciable sintering of the Co3O4 crystallites. Andoni Choya: Investigation, Writing – original draft. Sylwia Gudyka: Investigation. Beatriz de Rivas: Methodology, Validation. Jose Ignacio Gutiérrez-Ortiz: Methodology, Formal analysis. Andrzej Kotarba: Methodology, Conceptualization. Ruben López-Fonseca: Conceptualization, 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 the Spanish Ministry of Science and Innovation [PID2019–107105RB-I00 AEI/FEDER, UE]; the Basque Government [IT1297–19]; and the University of the Basque Country UPV/EHU [PIF15/335 and DOCREC21/23]. The technical and human support provided by Advanced Research Facilities-SGIker (UPV/EHU), and Central Scientific and Technological Research Services/Atomic Spectroscopy Division (University of Cádiz) is acknowledged. In addition, authors acknowledge the use of instrumentation as well as the technical advice provided by the National Facility ELECMI ICTS, node ‘Advanced Microscopy Laboratory’ at University of Zaragoza.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.apcata.2022.118511. Supplementary material. .

A series of CeO2-modified Co3O4 catalysts supported over α-Al2O3 foams was prepared by solution combustion synthesis and examined for the lean methane oxidation. Two different fuels were used namely, urea and glycine, with varying fuel/oxidiser (Φ) ratio. The catalysts were characterised by SEM-EDX, STEM-HAADF coupled to EDX mapping, ICP-AES, WDXRF, N2 physisorption, XRD, HRTEM, Raman spectroscopy, XPS, H2-TPR and CH4-TPRe, and their activity in the abatement of methane was analysed under realistic conditions. The use of glycine produced catalysts with significantly better morphological and structural properties. Likewise, a favoured insertion of cerium cations into the Co3O4 spinel lattice was observed, which caused a significant distortion of the spinel structure, thereby leading to a higher amount of mobile oxygen species capable of oxidising methane. These beneficial structural alterations were more pronounced with higher Φ ratios.

Transition metal and nitrogen co-doped carbonaceous catalysts (M/N/C, M ​= ​transition metal) are a category of non-precious metal catalysts that are recognized as the most promising substitutes for the expensive platinum-based catalysts currently used for proton exchange membrane fuel cells (PEMFCs) and metal-air/oxygen batteries (MABs). M/N/C catalysts have high activity towards the oxygen reduction/evolution reactions (ORR/OER), are much cheaper than precious metal catalysts, and use abundant rather than scarce resources. Unsurprisingly, they have become one of the most important subjects of research in this area during the past decade.In 1964, Jasinski [1] first confirmed that cobalt cooperated with nitrogen to yield catalytic activity toward the ORR in an alkaline medium. He then proved that a transition metal catalyst together with nitrogen and carbon could, through heat treatment, achieve better ORR performance [2]. To date, it has been confirmed that co-doping with one or two transition metals, such as Fe [3], Co [4], Ni [5], Cu [6], Mn [7], and Zn [8], can result in significant catalyst enhancement, and that additional co-doping with other non-metallic elements, such as S [9], P [10], and Se [11], further improves catalyst performance.M/N/C analogue catalysts are generally prepared by the following process. First, precursors containing carbon or carbon/nitrogen are pyrolyzed at high temperature in an inert or ammonia atmosphere, then treated with acid to remove inactive metal compounds, and annealed at high temperature for additional pyrolysis and graphitization to further enhance their activity and stability [12,13]. Nitrogen doping is achieved via pyrolysis under an ammonia flow instead of in an inert atmosphere, or by adding nitrogen-containing materials (e.g., melamine, ammonia chloride etc.) as the doping agents. Washing the pyrolysis product with acid, then annealing it at high temperature, can further enhance the catalytic performance.The catalyst’s activity arises from positively charged carbon atoms caused by C–N bonds, because C has a lower electro negativity than N. These are considered the main active sites, but another type of active site can be formed by transition metal atoms coordinated with nitrogen [14–17]. The structure and active sites of M/N/C catalysts are shown in Fig. 1 .The ORR and OER play important roles in next-generation energy conversion and storage technologies, including PEMFCs and MABs [18,19]. Due to the sluggish kinetics of these reactions at the cathode, the catalysts with high activity are required. Platinum-,ruthenium-, and iridium-based catalysts are considered efficient choices for these reactions, but their scarcity and high cost hinder their application on a large scale [20,21]. M/N/C analogue catalysts hold the promise of a solution to this issue, as the requisite metals are abundant and inexpensive and have been proven to exhibit good ORR performance in an alkaline medium — even higher than Pt-based catalysts —while in an acidic medium, their performance is only slightly lower than that of Pt-based catalysts. Furthermore, their performance toward the OER can be enhanced by doping or the use of a support.The research in this area has made tremendous progress in the past decade, and in this paper, we briefly review what has been achieved.Their high activity and much lower cost has made M/N/C catalysts a focus of interest for many researchers working on PEMFCs. Significant progress has been made on their activity in an acidic medium, and the half-wave potential with a rotating disk electrode (RDE) is close to that of a commercial Pt/C catalyst, establishing a foundation for the application of M/N/C catalysts in PEMFCs [22,23]. Numerous studies have concentrated on synthesizing high-performance M/N/C catalysts for this use. In 2011, Wu et al. [24] reported preparing a catalyst with Fe and Co incorporated into a carbon–nitrogen skeleton via pyrolysis, and investigated its performance as the cathode in a H2–O2 PEM fuel cell. With a loading of 4 ​mg ​cm−2, the single cell achieved a maximum power density of 0.55 ​W ​cm−2. Since then, the additional M/N/C catalysts have been developed for the same purpose, but their performance in a single PEM fuel cell has not managed to reach 0.6 ​W ​cm−2 [25–30]. However, the new precursors and new preparation technologies continue to be explored, so the catalysts’ performance in PEMFCs should continue to increase [31–40].Porous organic compounds/polymers or metal-organic frameworks are promising precursors for preparing M/N/C catalysts, not only for enhancing catalytic activity but also to achieve better conductivity and mass transfer to counteract the sluggish ORR and the complex mass transfer process in the membrane electrode assembly (MEA). The researchers have found that the porosity of organic compounds/polymers can improve the exposure of active sites and the mass transfer process in the MEA. Chung et al. [41] used two nitrogen precursors (cyanamide and polyaniline) to prepare an Fe/N/C catalyst with atomic-level hierarchical porosity, in which the active FeN4 sites were inserted into the carbon. The hierarchical porosity boosted mass transport when the catalyst was applied in a H2–O2/air PEMFC, and the maximum power density reached 0.87 and 0.94 ​W ​cm−2 at P O2 of 1.0 and 2.0 ​bar, respectively (Fig. 2 ). Fu et al. [42] adopted an ammonium chloride salt-assisted approach to tailor Fe/N/C catalysts derived from polyaniline to increase the number of FeN4 active sites on the edge side, which boosted ORR performance. In H2–O2 and H2-air PEMFCs, the peak power density reached 0.86 and 0.43 ​W ​cm−2, respectively.In addition, because the three-dimensional structure of metal-organic frameworks (such as zeolitic imidazolate frameworks, ZIF-8) can accommodate MN4 moieties and promote mass transfer (Fig. 3 ), they have become promising precursors for preparing M/N/C catalysts with greater densities of MN4 moieties and better mass transfer results [43]. To date, PEMFCs have achieved better performance using M/N/C catalysts derived from metal-organic frameworks precursors. Our group has worked in this field and made some progress. We prepared Fe/N/C catalyst with hollow carbon nano-polyhedrons derived from hollow ZIF-8 with ferric acetylacetonate and g-C3N4. The optimized catalyst yielded excellent ORR catalytic activity in an acidic medium and satisfactory performance in a H2–O2 PEMFC, with a current density of 133 and 400 ​mA ​cm−2 at 0.8 and 0.7 ​V, respectively [44]. Its high performance was due to the polyhedral structure and the high density of Fe–N4 moieties achieved by adding g-C3N4 during the preparation process. Subsequently, we further improved this catalyst’s performance in a PEMFC by adopting a novel preparation strategy, inserting iron into the ZIF-8 framework by evaporating ferrocene during the first heat treatment, so that atomic iron could disperse uniformly with nitrogen to form active sites (Fig. 4 ) [45]. Surprisingly, our catalyst exhibited excellent performance in a H2–O2 PEMFC, reaching a current density of 1100 and 637 ​mA ​cm−2 at 0.6 and 0.7 ​V with a low catalyst loading of 1 ​mg ​cm−2 in the cathode catalyst layer (conventional loading reported in the literature was 4 ​mg ​cm−2).Shui’s group [46] prepared a concave Fe/N/C single-atom catalyst (SAC) with a high density of Fe–N4 moieties by pyrolyzing mesoporous SiO2-coated ZIF-8 and an iron source, which greatly improved both catalytic activity and mass transport (Fig. 5 ). After assembly in the MEA cathode, the Fe/N/C SAC (TPI@Z8(SiO2)-650-C) achieved a peak power density of 1.18 ​W ​cm−2with 2.5 ​bar ​H2–O2, and its current density was 0.047 ​A ​cm−2 at 0.88 ViR-free and 1.0 ​bar ​H2–O2, reaching the US Department of Energy’s 2018 precious metal-free catalyst activity target. Shui’s group [47] also synthesized Co/N/C catalysts from aZIF-8 precursor with different densities of CoN4 active sites and examined the relationship between the CoN4 active sites and the PEMFC’s power density. At optimal active site density, the catalyst reached a peak power density of 826 ​mW ​cm−2 in a PEMFC cathode. Wu’s group [48] prepared a Co/N/C catalyst with a core–shell structure by pyrolyzing a precursor of Co-doped ZIF-8, assisted by a surfactant; the peak power density reached 0.87 ​W ​cm−2 in a H2–O2 PEMFC. This high performance was attributed to the atomic dispersion of the Co active sites, the presence of micropores, and the high N content obtained from the surfactant layer, which prevented the single atomic Co from aggregating. Fig. 6 presents the major breakthroughs that have been made in the application of M/N/C catalysts in PEMFCs.Although a great deal strides have been made in the performance of M/N/C catalysts, their application in PEMFCs still faces significant challenges when it comes to stability. Currently, the catalyst performance quickly decays during PEMFC operation, especially in the initial working hours, as shown in Fig. 7 [49]. Multiple researchers have worked on determining the attenuation mechanisms and improving the stability of these catalysts. Now, it is generally believed that the attenuation factors are as follows (Fig. 8 ): (1) demetallation of the active sites [50,51]; (2) protonation of the active sites [52,53]; (3) flooding of the micropores and catalyst layer [54,55]; (4) carbon corrosion [56,57]; and (5) attacks by H2O2 or free radicals [58,59].Understanding these attenuation mechanisms can enable the development of effective strategies to improve stability, and work in this area is underway. Wang et al. [60] focused on preventing water flooding and carbon corrosion (see Fig. 9 ), so they prepared Fe/N/C catalyst by surface fluorination. The electron-withdrawing and hydrophobic properties of the surface fluorination agent (trifluoromethylphenyl) prevented water flooding and inhibited the oxidative corrosion of the carbon matrix, yielding improvement in the catalyst’s stability. Wei et al. [61] looked at eliminating H2O2 by modifying the catalyst with CeO2 in order to clear the H2O2 produced during the ORR process. Open circuit voltage tests of a single PEMFC showed that the peak power density loss of Fe/N/C with CeO2 was much less. Similarly, Bai et al. [62] enhanced Fe/N/C catalyst’s stability by combining CeO2 nanoparticles into the catalyst structure; the nanoparticles acted as H2O2 scavengers and thereby protected the active sites, but the stability results didn’t carry over into the MEA. Unfortunately, there are few reports, to date, on successful improvement of M/N/C catalyst stability in a PEMFC, so concerted efforts are still required in this area.As mentioned earlier, M/N/C catalysts have excellent ORR performance in both acidic and alkaline media [63]. And it has been proved by many research works that the MABs with M/N/C catalysts as air cathode exhibited excellent activity and stability [64–69]. The application of M/N/C catalysts in various kinds of MABs has been widely and intensively explored, including in Li-air/oxygen batteries, Zn-air batteries, Al-air batteries, and Mg-air batteries. M/N/C catalysts generally exhibit good performance in these types of batteries, comparable to that of precious-metal catalysts.Shui et al. [70] first applied the Fe/N/C as cathode catalyst in Li–O2 battery, which could reduce the overpotential, enhance battery efficiency and improve lifespan. But they did not discuss the capacity of discharge/charge of the material in detail. However, this work revealed the possibility that Fe/N/Chad potential to apply in MABs. Li et al. [71] prepared the Fe/N/C catalyst with many exposed active sites through constructed special pore-in-pore structure, the material achieved a specific capacity of 7250 ​mA ​h ​g−1 at 70 ​mA ​g−1 in Li–O2 battery.More and more attentions have been attracted on designing M/N/C as the cathode catalysts for Zn-air battery. Zhang et al. [72] constructed the Fe/N/C electrocatalyst with FeN4 sites and hierarchically ordered porous, gaining a high power density of 235 ​mW ​cm−2 and a high capacity of 768.3 ​mA ​h ​g−1 when applying the catalyst as cathode of Zn-air battery. Chenet al. [73] prepared the Fe/N/C catalyst with high density of FeNx active sites and achieved an unexpected ORR performance in alkaline medium. While applying in Zn-air battery, the peak power density could reach 266.4 ​mW ​cm−2 and a specific capacity was 795.3 ​mA ​h ​g−1. Except Fe/N/C, Co/N/C is also an effective catalyst for the cathode of Zn-air battery. Luo et al. [74] designed the Co/N/C bi-functional catalyst for the cathode of Zn-battery with 3D brush-like nanostructure, exhibiting a high peak power density (246 ​mW ​cm−2) and better cycle performance.M/N/C analogue catalysts can also be applied in other MABs, such as Al-air battery, Mg-air battery. Li et al. [75] enhanced the performance of Fe/N/C by using Cu replaced partial Fe, while it used as cathode of a Al-air battery, a higher discharge voltage and better stability achieved. Ye et al. [76] chose Fe-doped ZIF-8 as precursors to fabricate Fe/N/C catalyst with high density of FeNx active sites and mesopores, and the good power density of 72 ​mW ​cm−2 was gained at 0.72 ​V when applying in Mg-air battery.Capacity and cycling performance are two key indicators for MABs, so designing M/N/C catalysts with high capacity and good cycling stability is key for their application in these batteries [77–79]. However, M/N/C catalysts generally exhibit higher charging overpotential in rechargeable metal air/oxygen batteries due to their inferior OER performance.To realize the desired charge–discharge performance of M/N/C catalysts in rechargeable MABs, it is essential to enhance their OER performance. Notably, decorating these catalysts with other compounds that yield unique structures can greatly improve their capacity and cycling performance [80,81]. He et al. [82] developed an inexpensive, easy way to prepare an efficient catalyst for the Li–O2 battery using N-doped graphene decorated with Fe/Fe3N/Fe4N nanoparticles, as shown in Fig. 10 . Excellent discharge–charge capability was obtained when it was used in the cathode. Wang et al. [83] prepared Fe/N/C catalyst embedded with Fe–FeC3 formed from a compound MOFs system that contained MIL-100(Fe) and ZIF-8. The catalyst possessed a high initial discharge capacity of 8749 ​m Ah gCat+C −1 and a charge capacity of 8104 ​m Ah gCat+C −1 after use in a Li–O2 battery. Chao et al. [84] introduced a second transition metal (Co) into Fe/N/C catalyst. They used Fe, Co bimetallic MOF to prepare Fe, Co co-doped carbon–nitrogen composite that yielded outstanding electrochemical activity and good cycling stability in Li–O2 batteries due to its accessible active sites and unique catalyst structure, shown in Fig. 11 .Progress has also been made with Zn-air batteries. Qin et al. [85] improved the capacity and cycling performance of a Zn-air battery by preparing a N-, P co-doped carbon catalyst inserted with FePx nanoparticles and a Fe/N/C moiety. Because of the high hydrogen evolution reaction performance of FePx and the structural properties achieved through co-doping, the resulting Zn-air battery showed a high specific capacity and outstanding stability. Co/N/C catalysts have proven even more suitable for Zn-air batteries. Amiinu et al. [86] designed Co/N/C with abundant Co–N coupling at the center as a high-performance bifunctional catalyst for the Zn-air battery, gaining a peak power density of 193.2 ​mW ​cm−2 and an energy density of 853.12 ​Wh kgZn −1. A Zn-air battery assembled with this catalyst had high cycling performance and operated for 80 ​h without attenuation. The battery’s improved performance was due to the high density of active sites for the ORR/OER as well as the catalyst’s porosity and high surface area. Liu et al. [4] reduced the discharge–charge overpotential and improved the cycling performance of a Zn-air battery by applying single-atom dispersed Co/N/C catalyst in the cathode. The results showed that abundant Co vacancies enhanced the ORR and OER performance, and the large amount of exposed pyridinic N inhibited the accession of zincate ions and the precipitation of ZnO (Fig. 12 ). For these reasons, the Zn-battery achieved a power density of 155.3 ​mW ​cm−2 and excellent cycling performance.From the reported work it can be seen that excellent breakthroughs have been achieved in improving the capacity and cycling stability of MABs by using M/N/C analogue catalysts in the cathode. Although some problems remain, these catalysts have great potential for application in MABs. The studies described above suggest that M/N/C catalysts yield better results, particularly in cycling performance, when decorated with other compounds, as shown in Table 1 .In summary, the application of M/N/C analogue catalysts in PEMFCs and MABs has been widely investigated, and great progress with some breakthroughs have been made. For PEMFCs, results comparable to those with Pt/C catalysts have been reported in recent years, especially with well-designed and well-prepared Fe/N/C and Co/N/C catalysts. Due to the high density of active sites dispersed in the skeleton and the active sites exhibit high catalytic activity towards the ORR process, making M/N/C analogue catalysts being the most promising catalysts to replace precious-metal catalysts that applying in cathode of PEMFCs. However, many problems and challenges still remain to be addressed before practical application of these catalysts can occur. One issue is the poor stability/durability of M/N/C catalysts in acidic PEM surroundings, caused by the dissolution of doped Fe/Co, carbon corrosion, and protonation of the active sites. Although some strategies have been suggested to address these challenges, it is still necessary to design and explore new carbon-based catalysts with better stability and durability. On the other hand, the application of M/N/C analogue catalysts in MABs has made great strides, and their practical application is almost feasible. While applying in cathode of MABs, these catalysts generally exhibit charge–discharge performance far exceeding that of the traditional manganese oxide-based air electrode, and excellent stability/durability. Although M/N/C analogue catalysts exhibited high oxygen reaction performance in cathode of MABs, the problems and challenges are also existed. Capacity and cycling performance should be further enhanced with the real application of these catalysts in MABs, especially the cycling performance. The OER performance of M/N/C analogue catalysts will influence the cycling performance of MABs. To enhance the cycling performance, decorating the M/N/C catalysts with other compounds may be an effective strategy. In summary, to realize the practical application in PEMFCs and MABs, M/N/C analogue catalysts needs to be further improved to enhanced its performance, especially the stabilities.The manuscript has been reviewed and approved by each signed author and its publication is approved by all authors and explicitly by the responsible authorities where the work was carried out.The manuscript has not previously been published in any form, nor submitted for reviews to any other journal currently. And we will not submit it elsewhere until a decision has been made by your journal. If accepted, the manuscript will not be published elsewhere in the same form or in any other language without the written consent of your journal. Therefore, there will not be any conflict of interest for this manuscript.This work was supported by the National Key Research and Development Program of China (Project Nos. 2017YFB0102900and 2016YFB0101201), the National Natural Science Foundation of China (NSFC Project Nos. 21476088 and 21776105), the Guangdong Provincial Department of Science and Technology (Project No. 2015B010106012), and the Guangzhou Science, Technology and Innovation Committee (Project Nos. 201504281614372 and 2016GJ006).

To meet the sharp increase in demand for clean and renewable energy, it is necessary to develop new energy-conversion and storage technologies, such as proton exchange membrane fuel cells (PEMFCs) and metal-air/oxygen batteries (MABs). Due to the sluggish reaction kinetics of the oxygen reduction reaction (ORR) and the oxygen evolution reaction (OER) in the cathodes of PEMFCs and MABs, significant amounts of precious metal catalysts need to be used, driving up the cost of fuel cells and MABs and thereby hindering their commercialization on a large scale. Transition metal and nitrogen co-doped carbonaceous catalysts (M/N/C) have high catalytic activity towards the ORR and OER once the catalysts are modified with certain promoters/additives. In addition, M/N/C catalysts can be prepared from abundant, inexpensive materials, making their cost negligible compared with precious metal catalysts, a development that would efficiently decrease the cost of PEMFCs and MABs. In last decade, numerous researchers have attempted to realize these applications of M/N/C catalysts, and some exciting results have been achieved, making these promising replacements for precious metal catalysts. However, some serious problems and significant challenges remain. In this paper, we review the research on the application of M/N/C analogue catalysts in PEMFCs and MABs in the last 10 years, indicate the remaining challenges, and suggest the future research directions.

Methane, the main component of natural gas is available in large quantities and it is therefore a major energy source nowadays. Produced from bio derived sources, bio-methane has the potential to contribute substantially to future climate-neutral energy generation [1,2]. Both natural gas and biomethane are attractive fuels for decentralized power generation such as fuel cell-based combined heat and power (CHP) units. Such a CHP system has been engineered and built at Fraunhofer IMM with an electrical power equivalent of 50 kWel [3]. It relies on hydrogen generation by steam-reforming of natural gas, which is also combusted to supply heat to the reforming process. Micro plate heat exchanger technology allows the efficient coupling of both reactions in catalyst coated microchannels [4].For the combustion of various hydrocarbons including VOCs, Pt-based catalysts are known for their high activity [5]. However, in case of methane, which is the hydrocarbon most difficult to combust due to the symmetry of the molecular structure and resulting weak adsorption capacity on various supports, Pt-based catalysts have several drawbacks, namely, poor low-temperature activity and insufficient stability due to sintering and self-poisoning of metallic Pt sites by the dissociative adsorption of O2, especially under O2-rich (lean) conditions [6–8]. To overcome these limitations, Pd and mixed PtPd catalyst are suggested, which show significantly improved low-temperature activity [9–11]. On the other hand, Pd-based catalysts are quite unstable at higher temperatures which can be explained by the reaction mechanism. Generally, the Mars- van Krevelen mechanism is the widely accepted route for methane combustion over Pd [12–14].According to this mechanism, methane is adsorbed on PdO, the latter being reduced. In a second step, Pd is oxidized by O2 from the gas phase. However, PdO decomposes quite rapidly at temperatures exceeding 600 °C so that the adsorption of CH4 is inhibited, whereas metallic Pd is much less active compared to PdO [15,16]. In addition, Pd catalysts are also very sensitive to small amounts of sulphur species in the feed such as SO2 or SO3 [17]. Although adsorptive desulfurization will always be performed in practical processes upstream the combustion step, a practical catalyst should be capable of tolerating temporary breakthrough of low amounts of sulphur.Several other catalysts and support materials for low-temperature methane combustion are described, e.g. Co3O4/CeO2, core-shell systems, bimetallic Ru-Re/Al2O3, pervoskites and Ni-based catalysts but either nothing is reported about their stability at high reaction temperatures or their stability is insufficient [18–22]. For decentralized power generation units, a high-temperature stable combustion catalyst is required, because natural gas or biomethane steam-reforming is performed at T > 700 °C [23–26]. To supply the steam reforming reaction with energy, a combustion catalyst formulation is lacking, which is stable at these temperatures. Rh catalysts are often reported to be stable at high reaction temperatures, but very few papers exist dealing with their performance as catalysts for the combustion of hydrocarbons [27]. Therefore, a systematic comparison of their activity with platinum and palladium catalysts is required, which is carried out in this study with special emphasis on the stability at 600 °C. Samples were prepared by keeping the total metal loading constant at 5 wt.-%, and replacing Pt with Rh up to 4 wt.-% Rh. Because Pd-containing catalysts are more common in methane combustion, a 5 wt.-% Pd catalyst is also tested for comparison.The catalysts were prepared by the incipient wetness impregnation method. The support γ-Al2O3 (Puralox SBa-200, SASOL) was impregnated with a solution containing the calculated amounts of tetraamineplatinum(II) nitrate (Alfa Aesar) and a Rhodium(III) nitrate solution (abcr GmbH) to obtain 1–5 wt.-% Pt with up to 4 wt% Rh-containing alumina based catalysts, hereinafter denoted as xPt-(5-x)Rh. A nitric acid solution of Palladium(II) nitrate (Sigma-Aldrich) was used as the Pd precursor. The overall metal loading was kept constant for all catalysts.After calcination at 450 °C, the as-prepared powders were deposited as washcoats onto stainless steel microchannels by means of an aqueous suspension containing polyvinyl alcohol (PVA) as a binder. The plates containing the microchannels (14 channels with 500 μm width, 250 μm depth and 25 mm length) were dried at room temperature and calcined at 450 °C. From each two plates and input and output capillaries, a reactor was assembled by laser welding. Details of the microreactors applied for testing have been described elsewhere [28].The surface area was determined by nitrogen adsorption using a Sorptomatic 1990 (Thermo Fisher Scientific Inc.) automatic apparatus and calculated by the Brunauer-Emmett-Teller (BET) method. The actual metal loadings were detected by X-ray fluorescence on an ED-XRF Canberra-Packard 1510 spectrometer.For powder XRD measurements, samples were mounted on a polyacetate foil using a glue based on collodion and recorded on a Stoe Stadi P diffractometer using Cu-Kα1 radiation from a sealed tube X-ray source operating at 40 kV and 30 mA.XPS spectroscopy measurements were performed by using a multi-chamber UHV system (PREVAC, Poland). It was equipped with a monochromated Al source (XM650 X-ray monochromator) operated at 360 W and a hemispherical electron analyser (Scienta R4000). For survey scans, a pass energy of 200 eV was fixed. The background pressure of the ultra-high vacuum (UHV) chamber was 5 × 10−8 mbar. All spectra were calibrated by setting the position of the C1s line to 284.7 eV. CasaXPS (ver. 2.3.16 PR 1.6) software was used to process the recorded spectra and for deconvolution of the signals. This procedure introduced a measurement uncertainty of about +/− 0.2 eV.High resolution TEM images were recorded on a Titan G2 60–300 kV (FEI Company) equipped with a monochromator operating on a Schottky field emission gun (FEG), three condenser lenses system, the objective lens system, Cs image correction, HAADF detector and EDS (Energy Dispersive X-Ray Spectroscopy). The samples were prepared by grinding in an agate mortar to fine powders, which were transferred to a slurry in 99.8% ethanol. After homogenisation by sonication, the material was supported on a 200-mesh copper grid with lacey formvar and stabilized with carbon and left on a filter paper for ethanol evaporation. Studies of the catalysts were carried out at an accelerating voltage of 300 kV.The microreactors were placed into a metal block powered by a heating cartridge regulated by a PID temperature controller with a K-type thermocouple inserted next to the catalyst coating. A temperature program from 450 to 750 °C was carried out for each catalyst. Also, the stability at 600 °C was investigated for 40 h. Experiments were carried out at a WHSV of 400 L gcat −1 h−1 of the feed flow. The O/C ratio was adjusted to 7.4, which corresponds to a methane concentration of 5.13%. Synthetic air and methane (purity >99.95%) were provided by Bronkhorst mass flow controllers. All tests were carried out at atmospheric pressure. Product gases were analysed by an online gas chromatograph (Agilent Technologies GC 7890A system) equipped with two thermal conductivity detectors and a flame ionization detector. The conversion of CH4 was calculated according to the following equation: (1) X C H 4 = C H 4 inlet − C H 4 outlet C H 4 inlet × 100 % [CH4]inlet is the inlet concentration of CH4 and [CH4]outlet is the CH4 concentration at the outlet. The catalysts were heated in 20 mL min−1 air to the reaction temperature prior to the catalytic tests. Table 1 provides an overview of BET surface area and XRF measurements as determined for all catalysts under investigation. In Fig. S1a-e (Electronic Supporting Information, ESI), representative HR-TEM images of the fresh samples of the Pt and PtRh catalysts are shown. For all fresh catalyst samples, small metal particles of the active phase are very highly dispersed and evenly distributed on the γ-Al2O3 carrier. In addition, some larger metal particles can also be observed, but in much smaller quantity. Some amorphous areas and characteristic rod-like γ-Al2O3 crystals are observed for the support material. The small size of the noble metal particles gets obvious from Table 2 , where the average dimensions of the metal particles of each catalyst are provided. Both 5Pt and mixed PtRh catalysts exhibit similar average metal particle sizes in the range of 1.4–2.3 nm as fresh catalyst. Phase identification adapted from HR-TEM images revealed the active phase of 5Pt/Al2O3 catalyst as Pt(111) and Pt(200) facets based on the interplanar distance of 1.96 Å and 2.26 Å, respectively. For PtRh catalysts, mixed PtRh particles exposing Pt/Rh(222), Pt/Rh(200) and Pt/Rh(111) facets were identified based on the interplanar distance of 1.11, 1.93 and 2.22 Å, respectively.In Fig. S2a-e (ESI) representative HR-TEM images of the spent catalyst samples after a stability test of 40 h duration are presented. Fig. S2 (ESI) provides evident that the particles on the spent samples increased in size in the course of reaction. This indicates aggregation and subsequent sintering. However, the average particle sizes of the spent samples also shown in Table 2 reveal smaller values with increasing Rh content. Even a low amount of Rh in the 4Pt1Rh catalyst leads already to a drastically lower average particle size compared to the 5Pt catalyst. This suggests less stable anchoring of the Pt crystallites in the absence of Rh on the alumina support. Thus, the introduction of Rh leads to a more stable catalyst. Similar to the fresh samples, mixed particles in the form of Pt/Rh(111) were identified for the PtRh catalysts with the exception of 4Pt1Rh catalyst, where only Pt(111) was detected likely originating from the low amount of Rh in the sample. Fig. 1 . shows the wide-angle XRD patterns of all catalyst under investigation. Peaks for cubic γ-Al2O3 can be found at 2θ ≈ 38.2°, 45.1°, 60.5°, 66.8° and 84.7°, corresponding to the (311, 400, 333), (440, 444) facets (ICDD 01–074-2206). Pt reflexes are identified at 2θ ≈ 39.6°, 67.4° and 81.1° for catalysts with high Pt-content (5Pt, 4Pt1Rh, 3Pt2Rh) corresponding to the (111,220,311) facets, which are characteristic of face-centered cubic structure of metallic Pt (ICDD01-070-2431). Diffraction peaks at 2θ ≈ 33.3°, 39.8° and 84.5° were assigned to the Rh2O3 (ICDD 01–076-0148) component for catalysts with high Rh content (1Pt4Rh, 2Pt3Rh).X-ray photoelectron spectroscopy (XPS) was used to investigate the surface composition of the PtRh alloy present in the developed catalysts. The XPS spectra of Pt 4f for the 5Pt and PtRh catalysts are shown in Fig. 2A. Two different Pt 4f deconvolution peaks appeared at 70.3 and 75.1 eV for the 5Pt catalyst. The first binding energy is attributed to the presence of metallic Pt0 and the second binding energy corresponds to the presence of Pt4+. The 4Pt1Rh and 3Pt2Rh catalysts shows three deconvolution peaks at 70.8–71.0 eV, 73.3 eV and 75.0 eV. The first binding energy is mainly due to the presence of Pt0, whereas the latter two peaks correspond to Pt2+ and Pt4+, respectively. It must be noted that the higher Rh content in 2Pt3Rh and 1Pt4Rh catalysts provides only one deconvolution peak at 74.9 eV due to presence of Pt4+. This result indicates that the first three catalysts mainly consisted of metallic Pt species along with oxidized Pt species, while the latter two catalysts consisted only of oxidized Pt4+ species. It could be assumed that not enough Pt is present on the surface and the presence of larger amounts of Rh prevent reduction of the Pt species in the 2Pt3Rh and 1Pt4Rh catalysts. As shown in Fig. 2B, the XPS spectra of the PtRh catalysts show the Rh 3d peaks at 310.0 eV, suggesting the presence of Rh3+ in all the catalysts, which is in line with the powder XRD measurements.The following results shown in Fig. 3 were obtained for the catalysts in the temperature program test. For all tests, no CO was detected, so that full selectivity towards CO2 was achieved.It is obvious that the low-temperature activity is improved by the partial replacement of Pt with Rh. A CH4 conversion of 99.4% is already achieved at 500 °C for the 1Pt4Rh catalyst and full conversion is observed for 3Pt2Rh, 2PtRh and 1Pt4Rh catalysts at 600 °C. On the other hand, for the 3Pt2Rh and 4Pt1Rh catalysts, a small decrease of conversion is observed at 750 °C, implying that these catalysts are not stable at this temperature. However, this is not the case for the 5Pt catalyst, which shows by far the lowest activity up to 600 °C among all catalysts tested. The Pd catalyst shows a different behavior with relatively high conversion obtained at 500 °C, which increases only slightly up to 750 °C, where 96.1% conversion is achieved.The results of the stability tests as determined for the fresh samples are presented in Fig. 4 . The experiments reveal that Rh also improves the stability of the catalysts at 600 °C, whereas the Pd catalyst loses its initial activity quickly and stabilizes at 44% conversion. A high initial conversion is obtained for the catalysts containing ≥2 wt.-% of Rh, which is in line with the measurements performed at 600 °C shown in Fig. 3. Although the 5Pt catalyst is also quite stable during the test, it shows much lower conversion than the Rh-containing catalysts, as it was expected from the tests performed at different temperatures (Fig. 3). The 4Pt1Rh catalyst shows a much higher initial conversion but deactivates rather quickly. A moderate deactivation is observed for 3Pt2Rh and 2Pt3Rh catalysts, which initially show full conversion. The stability of the initially highly active 1Pt4Rh catalyst is the best of all PtRh catalysts. However, a small decline in conversion is observed after 40 h on reaction stream and from HR-TEM images (see Fig. S2e, ESI) indicating that sintering also occurs on this catalyst. A higher degree of sintering is observed with higher Pt-loading of the catalyst.The stability of the PtRh catalysts corresponds directly to the average crystallite size of the spent samples, and a higher stability is obtained with smaller particle size. Furthermore, XPS results could explain the improved activity and stability of the 2Pt3Rh and 1Pt4Rh catalysts as both consisted of only oxidized Pt4+ species compared to the other catalysts, which consisted mainly of metallic Pt species along with oxidized Pt species.The partial replacement of Pt with Rh leads to a significantly improved low-temperature activity at 450 °C reaction temperature compared to the sample containing only Pt. Mixed PtRh crystals were observed in the corresponding fresh samples and their crystal size was as low as 2 nm. Also, a stability test of 40 h duration revealed improved stability of the mixed PtRh catalysts compared to the samples containing only Pt or Pd. This improved stability during methane combustion is believed to originate from the improved stability of the crystals of the active metal species with increasing Rh-content. The synthesized PtRh catalysts are a more versatile formulation compared to Pd- or Pt- containing catalysts, because they show both higher low-temperature activity and improved thermal stability at elevated temperatures. This is not the case for Pd-containing catalyst systems, as reported in the literature [15,16] and observed in the current work. However, HR-TEM images revealed that sintering is still an issue even for the most stable 1Pt4Rh catalyst.Therefore, further measures have to be taken to avoid sintering in future work, and to make the catalyst particles more resistant against aggregation. According to the literature, two strategies could be followed. The addition of a second metal could prevent the direct contact between the catalytic active Rh (and Pt) centers [29,30]. Furthermore, anchoring of the catalytically active rhodium to a sintering stable metal oxide could help to improve the longevity of the catalyst, as indicated in the work of Fan et al. [31].None. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2020.106202.

For the combustion of methane, a partial replacement of platinum by rhodium leads to a significantly improved light-off behavior of the corresponding mixed Pt-Rh/γ-Al2O3 catalysts compared to a catalyst containing only Pt supported on γ-Al2O3. For all Pt-Rh/γ-Al2O3 catalysts, mixed PtRh crystals of low particle size were found. At 600 °C, the stability of catalysts, was found to increase with increasing rhodium content. The crystal size was determined by HR-TEM for spent catalyst samples after 40 h stability test. The samples containing higher amounts of rhodium showed smaller crystal size suggesting this as the origin of the increased catalyst stability.

The water fraction derived from the Fischer-Tropsch (FT) process contains organic, water-soluble compounds that are challenging for conventional wastewater treatment processes [1]. These compounds consist of oxygenated hydrocarbons such as alcohols that can be converted into valuable products including hydrogen by aqueous-phase reforming (APR) [2,3]. The conversion of the alcohols in the FT water fraction into hydrogen may enhance the economic efficiency of renewable fuel production through FT-synthesis and reduce the organic content in the water fraction directed to wastewater treatment.Aqueous-phase reforming takes place at low temperatures, 200 °C to 250 °C, and above the bubble point pressure of the feedstock [4], avoiding an energy demanding evaporation step. APR constitutes a suitable candidate to process wastewater with diluted organic compounds because it is energetically efficient compared to steam and autothermal reforming [5]. The energy efficiency becomes significant because evaporation of the highly diluted organic solution is avoided in APR. Consequently, a number of research groups have studied the APR of oxygenated hydrocarbons [6–8]. The mass fraction of oxygenated hydrocarbons in the water stream from the FT process is typically below 10%. A mixture of short-chain alcohols (C1-C3) is the largest group of organic constituents in the FT water fraction [9,10]. Although polyols such as ethylene glycol and glycerol have been the main model compounds applied in APR, monohydric alcohols have also been considered [3]. Methanol [11–14] and ethanol [15,16] were model compounds in APR for hydrogen production over platinum-based catalyst. Iridium supported on different metal oxides was also utilized as catalyst in the APR of methanol [17], and iridium, rhodium and rhenium supported on TiO2 in the APR of ethanol [18,19]. The APR of ethanol has been additionally conducted over nickel-based catalyst supported on hydrotalcite-like compounds [20], alumina [21], and ceria [22]. The APR of C3 alcohols has been investigated over Pt-based catalysts supported on alumina [23,24] and on polymer-derived carbon [25]. Moreover, real FT derived water fractions have been processed over Ru supported on active carbon and on metal oxides to produce alkanes via hydrodeoxygenation [26,27].In a previous work on the APR of methanol, doping of nickel on alumina with copper or cerium enhanced the hydrogen production compared to the monometallic catalyst [28]. Furthermore, nickel on ceria-zirconia catalysts exhibited high performance in terms of methanol conversion and hydrogen production [29]. Several authors have similarly described the positive effect of cerium on the catalyst activity in APR. This effect has been attributed to oxygen vacancies that may promote reforming to hydrogen and carbon monoxide, and the conversion of carbon monoxide through the water-gas shift (WGS) reaction. The addition of cerium may also enhance the stability of the catalyst [22,30–36]. Furthermore, copper has been applied as a catalyst dopant to improve the hydrogen selectivity in APR by limiting the formation of side products such as alkanes [37–39].This work focuses on comparing the APR of methanol, ethanol, propan-1-ol and propan-2-ol. Water solutions of these alcohols were selected as model feedstock because they are representative of the FT-derived water fraction. Self-prepared nickel catalyst on ceria-zirconia supports with different ceria contents, and nickel doped with cerium on γ-alumina were selected due to their high activity and hydrogen selectivity, reported in previous studies [28,29]. Furthermore, nickel doped with copper or cerium supported on ceria-zirconia were considered as potential catalysts to improve the hydrogen production in APR. The results obtained from the APR of C1-C3 alcohols elucidate the effect of alcohol chain-length, the influence of the location of the hydroxyl group in alcohols and the type of catalyst applied on the reaction pathway and the product distribution.Ceria-zirconia supports with mass percentage of ceria in zirconia equal to 17% or 25% were supplied by MEL Chemicals in powder form. Engelhard supplied the γ-Al2O3 support. The metal precursors used in impregnation were nickel (Ni(NO3)2∙6H2O, ≥97.0%), copper (Cu(NO3)2∙3H2O, 99–104%) and cerium (Ce(NO3)3∙6H2O, ≥99.0%) nitrates. These chemicals were supplied by Sigma-Aldrich. The feedstock were aqueous solutions with 5% mass fraction of either methanol (MeOH), ethanol (EtOH), propan-1-ol (1-PrOH) or propan-2-ol (2-PrOH). The chemicals were supplied by VWR Chemicals (assay on anhydrous substance is 100%), Altia Industrial (Etax Aa, assay of 99.5%), VWR Chemicals (assay of 100%) and Fluka (assay of >99.9%) respectively.The catalysts listed in Table 1 were prepared through incipient wetness impregnation, similarly as in [29]. The ceria-zirconia supports were calcined at 450 °C for 10 h in flowing synthetic air, pelletized, crushed and sieved to 200–300 μm prior to metal impregnations. The bimetallic catalysts supported on 25% ceria-zirconia were prepared through co-impregnation of nickel and copper or cerium precursors in water solutions. After impregnating the metal precursors on 17% and 25% ceria-zirconia supports, the impregnated materials were kept for 24 h at room temperature, followed by drying at 110 °C and calcination in flowing air at 500 °C for 4 h. The bimetallic catalyst supported on γ-Al2O3 was prepared through sequential impregnation of first cerium precursor followed by nickel precursor. The catalyst was dried at 80 °C under vacuum and calcined at 500 °C for 2 h in flowing air after impregnation. Prior to the APR experiments, the catalysts were reduced in situ at 450 °C and 2.5 MPa for 2 h with a H2:N2 = 1 gas flow of 10 dm3  h−1. The values of target mass percentage included in Table 1 were calculated as the mass of metal in zero oxidation state per total mass of catalyst.The equipment and methods utilized for the characterization of catalysts were detailed in [29] and briefly described here. The supports, after calcination in the case of mixed-oxide materials, calcined catalysts and spent catalysts were characterized using atomic absorption spectroscopy (AAS) and inductively coupled plasma - optical emission spectroscopy (ICP-OES) to analyse metal loadings. For the AAS, 200 mg of catalyst was dissolved in aqua regia at 120 °C and subsequently diluted with Milli-Q water. Ni and Cu loadings were determined with a Varian AA240 AAS equipment applying air-acetylene flame, and Ce loading with a Perkin Elmer 7100 ICP-OES. Nitrogen physisorption was applied to determine BET surface areas and BJH method to determine pore volumes and pore sizes distribution. Nitrogen physisorption was conducted in an Thermo Fisher Ultra Surfer after degassing the calcined catalyst samples at 200 °C for 3 h in vacuum, and the samples of spent catalysts at 120 °C for 5 h. X-ray diffraction (XRD) was conducted to identify crystalline phases and to determine the crystallite sizes of nickel species. A PANalytical X-pert PRO MPD Alpha-1 diffractometer with Cu Kα1 radiation (45 kV and 40 mA) was utilized to obtain the XRD data. The scanning was continuous and ranged from 10° to 90° (2Ɵ) with step size of 0.0131°. Based on peak broadening, Scherrer equation [40] was applied to estimate the particle size of nickel species. The X-Ray wavelength of Cu K-alpha was assumed to be 0.154 nm, and a crystallite shape-factor of 0.94 was applied, considering sphere-like catalyst particles. Attempts to identify nickel species and determine their particle size with a scanning transmission electron microscope (STEM) were made with no success. In the results form STEM, Ni species were not detected, most likely because the atomic weight of nickel is considerably lower than the atomic weight of the metals in the support, cerium and zirconium [29].Aqueous solutions prepared with Milli-Q water and 5% mass fraction of either MeOH, EtOH, 1-PrOH or 2-PrOH were processed in APR over the catalysts listed in Table 1. The experiments were conducted over 1.5 g of catalyst in a continuous fixed-bed reactor described in detail elsewhere [29]. The gaseous products were analysed with an online Agilent 490 Micro GC Biogas Analyzer with two thermal conductivity detectors (TCD), and the liquid products were analysed offline with an Agilent GC 6890 series with a flame ionization detector (FID) according to the detailed methods described in [29]. The operating conditions were set to be 230 °C, 3.2 MPa of inlet pressure, and 2.0 cm3 min−1 of aqueous solution flow.Ideally, the reforming of MeOH, EtOH, 1-PrOH or 2-PrOH results in the formation of H2 and CO (Eqs. 1–4). At low temperatures, the WGS reaction (Eq. 5) is favoured to convert CO with H2O into CO2 and H2. The Gibbs free energy changes presented in this work were calculated at 503 K with HSC Chemistry 8, software from Outotec. In addition to the APR operating conditions, potentially spontaneous reactions (Eqs. 2–4) due to slightly positive Gibbs free energy changes [41], and the type of catalyst and feedstock may facilitate different reaction pathways and the formation of side products, for instance through hydrogenation of carbon oxides (Eqs. 6 and 7). Accordingly, full reforming denotes in this work the reaction where the alcohol in the feedstock is converted into H2 and CO; and aqueous-phase reforming is considered as the process where different reactions, in addition to full reforming, such as WGS, and side reactions including methanation may take place.Methanol full reforming: (1) C H 3 O H ( l )   ↔ H 2 O   C O   g + 2 H 2   g Δ G 503 = -   24.8   k J Ethanol full reforming: (2) C 2 H 5 O H   l + H 2 O l   ↔   2 C O   g + 4 H 2   g Δ G 503 = +   8.5   k J Propan-1-ol full reforming: (3) C 3 H 7 O H   + 2 H 2 O   l   ↔   3 C O   g + 6 H 2   g Δ G 503 = +   19.9   k J Propan-2-ol full reforming: (4) C 3 H 7 O H   + 2 H 2 O l   ↔   3 C O   g + 6 H 2   g Δ G 503 = +   47.1   k J WGS reaction: (5) C O   g + H 2 O   l ↔   C O 2 g + H 2   g Δ G 503 = -   33.6   k J COx hydrogenation (6) C O   g + 3 H 2   g ↔   C H 4 g + H 2 O   l Δ G 503 = -   82.1   k J (7) C O 2   g + 4 H 2   g ↔   C H 4 g + 2 H 2 O   l Δ G 503 = -   67.1   k J The results presented in this work are based on product analyses taken at approximately 6 h on stream, when the amount of gases in the outlet stream had stabilized to nearly constant concentrations after a gradual increase. The parameters used to evaluate the experimental results are mass balance (MB, Eq. 8), conversion (X, Eq. 9), selectivity to liquid products (Sk , Eq. 10), hydrogen production rate (H2 PR, Eq. 11), hydrogen efficiency (H2 Eff, Eq. 12), hydrogen molar fraction among gaseous products (xH2 , Eq. 13), yield of gaseous compounds (Yi , Eq. 14) and yield of liquid compounds (Yk , Eq. 15). Eq. 12 includes a H2/CO2 stoichiometric reforming ratio (RR) that has been traditionally used to evaluate the efficiency of H2 production [42]. This factor considers the stoichiometric production of H2 through full reforming (Eqs. 1–4) and WGS reactions (Eq. 5). Therefore, RR equals to 3 for MeOH, 6 for EtOH, and 9 for 1-PrOH and 2-PrOH. The H2 molar fraction in Eq. 13 evaluates the fraction of H2 produced among gases, and liquid products are disregarded in this formula. In addition, when referred to H2, Eq.14 considers the amount of H2 in the outlet stream per amount of alcohol fed into the system. This equation disregards water as a reactant, although water constitutes the hydrogen source when the WGS reaction takes place. Accordingly, the hydrogen yields reported in this work might be higher than 100%. (8) M B   % = m o u t l i q ∙   w o u t j m i n l i q ∙   w i n j + X (9) X % = x i n j - x o u t j x i n j (10) S k   % = x o u t k x i n j - x o u t j (11) H 2   P R = n ˙ ( H 2 ) m c a t a l y s t (12) H 2   E f f . = n ˙ ( H 2 ) R R ∙ n ˙ ( j i n ) (13)   x H 2   ( % ) = n ˙ ( H 2 o u t ) ∑ n ˙ ( g a s o u t ) (14) Y i   % = n ˙ ( i o u t ) n ˙ ( j i n ) (15) Y k   % = x o u t k x i n j In Eqs. 8–15, j refers to the alcohol in the aqueous solution, i refers to a gaseous product and k refers to a liquid product. The m liq is the total mass of aqueous solution fed into (in) or collected from (out) the system, wj is the mass fraction of the alcohol in the aqueous solution fed into (in) or collected from (out) the system, x is molar fraction and ṅ is molar flow rate, and m catalyst is the mass of catalyst loaded into the reactor. In Eqs. 9, 10 and 15, mol fractions of liquid compounds were applied instead of molar flow rates, because the molar flow rate of the outlet liquid was unavailable due to experimental limitations to measure it accurately. Accordingly, these equations do not take into account possible changes in the total number of moles.The ceria-zirconia-supported catalysts were impregnated with either nickel, or nickel and cerium or copper, and the metal content in the catalyst was determined with AAS analysis (Table 2 ). Compared to the targeted amounts (Table 1), about 90% of the Ni target was successfully impregnated on 17CeZr and 25CeZr supports, whereas when Cu or Ce were additionally impregnated, only about 70% of the target Ni was deposited. The mass percentage of Cu, 3.8%, was also lower than originally targeted, 5%. The amount of Ce detected in NiCe/25CeZr was affected by the cerium in the support and the impregnation efficiency cannot be evaluated in this case. The alumina-supported catalyst contained as much Ni as targeted (13%). The impregnation success on the alumina support can be attributed to its larger surface area, 159 m2 g−1, compared to 17CeZr and 25CeZr supports, with 112 m2 g−1 and 99 m2 g−1 respectively (Table 2). The slight difference of surface area between 17CeZr and 25CeZr had no obvious effect on the amount of Ni impregnated on the catalyst. Table 2 additionally includes the metal loadings in the spent catalysts. The APR of MeOH induced no significant change on the metal content of ceria-zirconia-supported catalysts. In the spent alumina-supported catalyst, about 30% less Ni and a 70% less Ce was observed compared to the calcined catalysts. In addition to potential leaching of Ni and Ce, the decrease of metal mass fractions can be attributed to the weight increase of the catalyst caused by the phase change to boehmite undergone by alumina (Fig. 1 ). The metal content of spent Cu-doped catalysts was similar to the amount in the calcined catalyst. Therefore, using Cu as a promoter improved the stability of the catalyst and prevented leaching. In contrast, during the APR of C2 and C3 alcohols, leaching of 20% of the Ni in Ni/17CeZr and Ni/25CeZr was observed. In a previous study [43], nickel leaching was attributed to the acidity of the reaction medium due to carbonic acid formed from CO2. However, leaching was not observed for the most acidic feedstock applied in the present work, MeOH, which also yielded the highest amount of CO2.The surface area, and pore volume and average pore diameter of supports, and calcined and spent catalyst are also included in Table 2. Metal impregnation on 17CeZr and calcination decreased its surface area by 35%, whereas impregnations on 25CeZr and calcination caused a decrease between 15%–25%. The surface area of spent catalysts was generally lower than the surface area of calcined catalysts, which is attributed to partial obstruction of pores, confirmed by lower pore volume. The alumina-supported catalyst showed a considerable decrease of surface area during the APR experiments from 129 m2 g−1 to 22 m2 g−1. This decrease was caused by a phase change from γ-alumina to boehmite in the aqueous medium, observed in the XRD results. The structural change of the support (Fig. 1) and subsequent surface area decrease may have caused metal leaching, and collapse of pores (Table 2). The surface area of CeZr-supported catalysts decreased by 4%–13% during the APR experiments; nonetheless, the type of alcohol processed had no significant effect on the surface area and pore volume of the same catalyst. The average pore size of Ni/17CeZr and Ni/25CeZr remained close to 8 nm and 11 nm respectively, during APR. The average pore diameter increased from 10 nm in calcined NiCu/25CeZr to 12 nm in the spent catalyst. The surface area of NiCe/25CeZr was 5% lower in the spent catalyst and the pore volume and average pore diameter were unaffected by the reaction conditions. Fig. 1 presents the X-ray diffractograms of supports, calcined catalysts and spent catalysts. Compared to the pure supports, NiO peaks were identified in the X-ray diffractogram of Ni-containing catalyst at 2θ positions 36° and 43°. A CuO peak was identified for the NiCu-based catalyst at 38° 2θ. In contrast, the addition of cerium was undetected in the diffractograms of NiCe/Al and NiCe/25CeZr. After reducing the catalysts in situ, nickel remained in the metal form also after the APR experiments were carried out, regardless of the type of catalyst or feedstock applied. The diffractograms of spent catalysts presented peaks of metallic Ni at 2θ positions 44.4°, 51.9° and 77.1°. The diffractograms of spent NiCu/25CeZr additionally presented a peak at 2θ position 43.3° that corresponds to metallic Cu. Although the peaks of metallic Ni and Cu at 2θ position 44.4° and 43.3° are not completely separated, the appearance of two different peaks indicates that the complete formation of an alloy can be discarded [44]. Regarding the alumina-supported catalyst, the previously mentioned phase change from γ-alumina to boehmite (Section 3.1) is confirmed in the diffractogram of spent NiCe/Al. Boehmite can be identified in the peaks at 2θ positions 14.5°, 28.2°, 38.4°, 48.7°, 49.3°, 55.3°, 64.1° and 72.0°.The crystallite size of Ni species in calcined catalysts and in spent catalysts determined by Scherrer equation are presented in Table 3 . For the calcined catalysts, the most intense characteristic peak of NiO, 43.3° 2θ, was considered to determine its crystallite size. The peak at 44.4° 2θ, characteristic of metallic Ni, was considered for the spent catalysts. NiO crystallite size on the CeZr-supported calcined catalysts was approximately 20 nm, except for NiCe/25CeZr where larger particles were determined (28 nm). NiO crystallite size was 8 nm in the calcined NiCe/Al catalyst. APR caused no obvious effect on the nickel crystallite size of NiCu/25CeZr regardless of the feedstock applied. Accordingly, Cu promoted the stability of the catalyst, which was also indicated by the results included in Table 2. APR over the other catalysts caused different changes in the crystallite size when different feeds were processed. The APR of ethanol induced a significant growth of nickel particles in Ni/17CeZr and Ni/25CeZr. A similar effect has been described in APR over a ruthenium-based catalyst, whose metal dispersion decreased from 25% to 19% attributed to metal sintering in the APR of ethanol [19]. Moreover, a significant increase in the size of Ni particles was observed in Ni/17CeZr after the APR of methanol was conducted. As indicated in the footnote of Table 3, the Ni particle size of spent Ni/17CeZr and Ni/25CeZr was determined after the catalysts had been 12 h on stream and had been reduced twice. Accordingly, Ni/17CeZr has lower tolerance to the reduction and APR conditions that caused the increase of Ni particles compared to Ni/25CeZr. On the other hand, the alumina supported catalyst suffered obvious Ni agglomeration in the APR of MeOH due to the phase change to boehmite.To maximize the H2 production was one of the main targets of this work. Hydrogen constituted the main gaseous product in the APR of C1-C3 alcohols over different catalysts, with H2 molar fraction in the gas phase between 63% and 95% (Table 4 ). Hydrogen production and yield are useful parameters to evaluate the overall amount of hydrogen produced in the APR process independent of the reaction pathways. In the APR of MeOH, the highest values of H2 production rate were reached, between (1.9–2.4) mmol·min−1·g catalyst −1, and H2 yields, between 93% and 110% over Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr. In contrast, NiCe/25CeZr and NiCe/Al exhibited poorer performance with 70% lower hydrogen production rate and yield. The APR of other alcohols produced different amounts of H2 depending on the catalyst. Over Ni/17CeZr, the hydrogen production and yield were higher in the APR of EtOH than in the APR of 1-PrOH. Similarly, the APR of 1-PrOH resulted in lower amounts of hydrogen over NiCu/25CeZr, in this case, compared to the amount obtained from 2-PrOH. The lowest amounts of hydrogen were obtained from the APR of 1-PrOH over Ni/17CeZr, 0.27 mmol·min−1· g catalyst -1 and 24% H2 yield, and over NiCu/25CeZr with 0.15 mmol·min−1·g catalyst -1 of H2 and 13% H2 yield, and from the APR of EtOH over Ni/25CeZr 0.25 mmol·min−1·g catalyst -1 of H2 and 17% H2 yield. Over Ni/25CeZr and NiCu/25CeZr, H2 was produced in similar amounts in the APR of 2-PrOH, (0.45 and 0.50) mmol·min−1·g catalyst -1, and 41% and 45% H2 yield respectively.H2 efficiency indicates the extent of full reforming to gases and WGS (Eq. 1–4 and 5, Section 2.3). Those alcohols whose reaction pathway in APR was mainly full reforming to gases and subsequent WGS will show higher H2 efficiency. The APR of MeOH produced only gases regardless of the catalyst applied. In the APR of MeOH, H2 efficiency values were around 35% over Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr. It is worth noticing that NiCu/25CeZr was able to reach a H2 efficiency similar to Ni/17CeZ and Ni/25CeZr with 20% lower MeOH conversion. The APR of MeOH over NiCe/25CeZr and NiCe/Al resulted in low conversions, around 15%, with similarly low H2 efficiency, below 10%. In the APR of C2-C3 alcohols, different reaction pathways to full reforming and WGS to produce H2, and side reactions that consume H2 explain considerably lower H2 efficiency.Mainly gaseous products were obtained also in the APR of EtOH over Ni/17CeZr and Ni/25CeZr. However, 30% selectivity to liquid products indicates that full conversion to gases and WGS (Eqs. 2 and 5) were not the only reaction pathways, and side reactions to produce ethanal took additionally place. Ni/17CeZr and Ni/25CeZr reached also similar conversions close to 15%. However, H2 efficiency over Ni/17CeZr, 7%, was more than twice the value achieved over Ni/25CeZr, which suggest higher selectivity to the full reforming and WGS pathway (Eqs. 2 and 5) over Ni/17CeZr.The APR of 1-PrOH resulted in liquid product selectivities around 25%, over Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr. Therefore, as in the APR of EtOH, full conversion to gases and WGS (Eqs. 3 and 5) was not the only reaction pathway and side reactions to produce liquid compounds took additionally place. The conversion of 1-PrOH over Ni/25CeZr, 44%, was twice as high as over NiCu/25CeZr, and three times as high as over Ni/17CeZr. H2 efficiency was the highest over Ni/25CeZr, 7%, which suggest relatively higher selectivity to the reaction pathway that involves full reforming to gases and WGS (Eqs. 3 and 5). Over Ni/17CeZr and NiCu/25CeZr, the APR of 1-PrOH resulted in lower H2 efficiency around 3% and 1% respectively.In the APR of 2-PrOH, similar results were obtained over Ni/25CeZr and NiCu/25CeZr differing from the APR of other alcohols in the liquid selectivity. The APR of 2-PrOH resulted in higher liquid product selectivity around 65% and its conversion was comparable to that achieved in the APR of MeOH, close to 60%. However, H2 efficiency was relatively low, 5%. High selectivity to liquids and low H2 efficiency with high conversion indicates that 2-PrOH was converted and H2 was produced through a reaction pathway different to full reforming to gases and WGS (Eqs. 4 and 5).The discussion included in the previous paragraphs suggests that the APR reaction pathway of C1-C3 is more complex than that explained by Eqs. 1–5. Therefore, the following subsections will be devoted to the evaluation of the product distribution obtained in the APR of C1-C3 over different catalysts to achieve a better understanding of the reaction pathways in the APR.The APR of MeOH over different Ni-based, and Cu- and Ce-containing catalysts was conducted to evaluate the effect of metal dopants on the catalyst performance. Methanol conversion and H2 yield decreased in the order Ni/25CeZr > NiCu/25CeZr > NiCe/Al > NiCe/25CeZr (Table 4 and Fig. 2 ). Both Ce-doped catalysts, NiCe/25CeZr and NiCe/Al, showed significantly poorer performance than the other catalysts. The H2 yield over the Ce-doped catalysts was less than 40% of the H2 yield over Ni/25CeZr (Table 4 and Fig. 2). The lower performance of NiCe/Al catalyst can be attributed to the phase change undergone by γ-Al2O3 to boehmite and consequent decrease in the surface area, and metal agglomeration and leaching (Section 3.1). The results obtained over NiCe/25CeZr are surprisingly poor compared to those obtained over similar catalysts such as Ni/25CeZr or NiCu/25CeZr. The poorer results over NiCe/25CeZr reveal the negative effect of nickel particle growth (Table 3) on the performance of this catalyst, compared to the other 25CeZr-supported catalysts. Over NiCu/25CeZr, the MeOH conversion was lower than over Ni/25CeZr, and the H2 yields were similar over NiCu and Ni on 25CeZr, which explains the higher H2 molar fraction among gases over the Cu-doped catalyst (Table 4), as similarly reported in [29]. In addition, the product distribution was similar over Ni/25CeZr and NiCu/25CeZr (Fig. 2). The presence of CO2 among the gases confirms WGS reaction activity (Eq. 5) over both catalysts. The detected CH4 indicates that methanation of carbon oxides with hydrogen consumption took place, also observed over NiCe/25CeZr. Conversely, the only products observed over NiCe/Al were H2 and CO2, which indicates that CO conversion through the WGS reaction (Eq. 5) was highly promoted. No side products over NiCe/Al suggest that the selectivity was superior to that over the other Ce-doped catalyst, NiCe/25CeZr. Lower conversion over NiCe/Al hinders the comparison in terms of selectivity with the other 25CeZr-supported catalysts.The yields obtained over NiCu/25CeZr, compared to that of Ni/25CeZr, suggest that copper addition promoted the WGS reaction and methanation was less favourable. Additionally, MeOH conversion was lower over NiCu/25CeZr than over Ni/25CeZr. To evaluate the effect of Ni content on the APR of MeOH, the Ni loading was used to calculate the H2 production rate per mass of Ni using the values of H2 production rate in Table 4. Ni/25CeZr had a Ni loading of 9% mass fraction whereas NiCu/25CeZr had 7% mass fraction of Ni (Table 2). Accordingly, the H2 production rate was 27 mmol·min−1· gNi −1 over Ni/25CeZr and 30 mmol·min−1· gNi −1 over NiCu/25CeZr. These Ni-based H2 production rate indicates that lower H2 production rate over NiCu/25CeZr (Table 4) could be attributed to its lower amount of Ni compared to Ni/25CeZr; moreover, a possible negative effect of Cu on MeOH reforming could also explain it.Methanol is a simple molecule with no CC bonds, and a C/O stoichiometry of 1:1 that allows high selectivity towards hydrogen in APR [45]. Accordingly, the APR of methanol resulted in high conversions and hydrogen yields. However, longer chain alcohols present CC bonds and different C/O stoichiometry, which, along with the catalyst, has a noticeable effect on the alcohols conversion and H2 yield (Table 4 and Fig. 3 ). Fig. 3 summarizes the hydrogen yield versus alcohol conversion obtained in the APR of C1-C3 alcohols over different catalyst (data from Table 4).Conversion and H2 yield were the highest over Ni/25CeZr (Fig. 3, black) when MeOH and 1-PrOH were applied. In contrast, the APR of 2-PrOH over NiCu/25CeZr (Fig. 3, white) resulted in slightly higher conversion, and in the APR of EtOH, a noticeably higher yield was reached over Ni/17CeZr (Fig. 3, grey). Alcohol conversion and H2 yield in the APR of EtOH and 1-PrOH followed different trends over different catalysts.MeOH (Fig. 3, spheres) was converted more easily into H2 than longer-chain alcohols, although the number of hydrogen atoms contained in MeOH is lower. Conversion of 2-PrOH (Fig. 3, triangles) achieved the level of MeOH conversions; however, the H2 yields were considerably lower. The conversion of EtOH (Fig. 3, cubes) was significantly lower than the conversion of the other alcohols and similar over Ni/17CeZr and Ni/25CeZr. The APR of EtOH caused the agglomeration of Ni particles in both catalysts (Table 3), which decreased the number of active sites and likely resulted in lower conversions. H2 yield in the APR of EtOH was higher over Ni/17CeZr than over Ni/25CeZr. The conversion of specially 1-PrOH (Fig. 3, diamonds) varied considerably over different catalysts. The alcohol was converted to a larger extent and resulted in higher H2 yield over Ni/25CeZr than over Ni/17CeZr. Accordingly, higher amount of Ce in the support could have resulted in higher catalytic activity in this case.The main reaction pathways in the APR of MeOH can be deduced from the product distribution (Figs. 2 and 4 ). Methanol is accompanied by longer-chain alcohols in real water fractions derived from FT synthesis. Therefore, the product distribution from the APR of C2 and C3 alcohols was additionally studied to understand the effect of CC bonds, higher C/O ratio and higher number of hydrogen atoms in the alcohols on the product distribution. The effect of the hydroxyl group location in C3 alcohols on the bond scissions and consequent product distribution was also addressed. The product distribution derived from the APR of C2 and C3 alcohols was evaluated to enhance the understanding of the reaction pathways in APR of different alcohols over different catalysts. Fig. 4 shows the conversion of different alcohols and product yields over three different catalysts. The information presented in Fig. 4 will be discussed in this section to propose potential reaction pathways followed by different C1-C3 alcohols in APR over Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr.The APR of MeOH produces H2 and CO (Eq. 1). Moreover, CO2 is produced through WGS reaction (Eq. 5), which decreases the amount of CO while favouring the H2 yield (Fig. 4). In addition, CH4 produced in methanation (Eqs. 6 and 7), was identified among the products from APR of MeOH over the three catalysts, Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr. Ethane and propene were detected in negligible amounts over the monometallic catalysts. Undesired alkane formation was restricted over NiCu/25CeZr, which resulted in lower amounts of methane. Furthermore, no liquid products were detected in the liquid samples from the APR of MeOH over either of the catalysts.In the APR of EtOH, H2, CH4 and CO2 were the main gaseous products, in addition to negligible amounts of C2H6 and C3H6, and ethanal was the only liquid product (Fig. 4 a and b). Ni/17CeZr produced considerably larger amounts of hydrogen with similar conversion to that achieved over Ni/25CeZr. The difference in the H2 yields suggests that the APR of EtOH followed different pathways over Ni/17CeZr and Ni/25CeZr, which is also indicated by the different formation ratio of CO2 and CH4. Large amounts of H2 can be produced via full reforming of EtOH to gases and WGS (Eqs. 2 and 5). Additionally, H2 can be produced through EtOH dehydrogenation (Eq. 16) or decarbonylation (Eq. 18).Ethanol dehydrogenation (16) C 2 H 5 O H   l   ↔ H 2 O   C 2 H 4 O   ( l ) +   H 2   ( g ) Δ G 503 = + 21.6   k J Ethanal decarbonylation (17) C 2 H 4 O   l   ↔ H 2 O   C O   g + C H 4   ( g )   Δ G 503 = - 95.2   k J Ethanol decarbonylation (18) C 2 H 5 O H   l   ↔ H 2 O   C H 4   g +   C O   g +   H 2   ( g ) Δ G 503 = - 73.6   k J In the dehydrogenation of EtOH, ethanal is formed, as previously suggested in a different study [46], whereas the decarbonylation of EtOH involves the formation of CH4 and CO, as proposed in [47]. Ethanal yield was low and similar over Ni/17CeZr and Ni/25CeZr, which indicates that EtOH dehydrogenation was low regardless of the catalyst. The formation of ethanal is thermodynamically unfavourable (Eq. 16), and thus, ethanal yields were low. EtOH decarbonylation (Eq. 18) was more thermodynamically favourable, also compared to full EtOH reforming to gases (Eq. 2). However, considering the stoichiometry of Eqs. 2 and 18 and the product distribution in Fig. 4 a and b, it can be assumed that additionally, full reforming to gases took place over Ni/17CeZr and Ni/25CeZr because the H2 yield was higher than that of CO and CH4. Nevertheless, full reforming to gases was more favourable over Ni/17CeZr according to the significantly higher H2 yield and lower CH4 yield compared to Ni/25CeZr.The formation of CH4 and CO can take place via three different pathways in the APR of EtOH: (i) ethanal decarbonylation (Eq. 17); (ii) EtOH decarbonylation (Eq. 18), which also produces H2 as previously indicated; and (iii) EtOH full reforming to CO and H2 (Eq. 2) followed by hydrogenation of carbon oxides into CH4 (Eqs. 6 and 7). Considering the stoichiometry of Eqs. 2,17 and 18, the product distribution obtained in the APR of EtOH (Fig. 4) suggests that EtOH decarbonylation (Eq. 18) was more favourable over Ni/25CeZr than over Ni/17CeZr. This reaction pathway (Eq. 18) explains the lower H2 and CO2 yields and larger amount of CH4 over Ni/25CeZr compared to the product distribution obtained over Ni/17CeZr, where full reforming was more favoured. Furthermore, similarly negligible CO yields over Ni/17CeZr and Ni/25CeZr indicate that CO2 is formed via WGS reaction (Eq. 5), which allows the formation of additional H2.The APR of 1-PrOH was conducted over Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr. Hydrogen was the main product over these three catalysts. Over NiCu/25CeZr, CH4 was also formed, whereas over Ni/17CeZr and Ni/25CeZr, CO, CO2, C2H6 and a small amount of C2H4 were additionally observed. The main liquid product detected over the three catalysts was propanal (Fig. 4). Hydrogen yield was higher over Ni/25CeZr than over Ni/17CeZr due to higher 1-PrOH conversion. On the other hand, although Ni/17CeZr and NiCu/25CeZr allowed similar 1-PrOH conversions, the H2 yield was higher over Ni/17CeZr. The differences in the H2 yield suggest that 1-PrOH followed different reforming pathways over different catalysts, which is also indicated by the different formation ratio of CO2 and CH4 (Fig. 4).Hydrogen can be produced via full reforming of 1-PrOH to gases and WGS (Eqs. 3 and 5). Additionally, H2 can be produced through 1-PrOH dehydrogenation (Eq. 19) or decarbonylation (Eq. 21) [48]. In the dehydrogenation of 1-PrOH, propanal is formed, whereas the decarbonylation of 1-PrOH involves the formation of C2H4 and CO, as proposed in Ref. [47]. Propanal yield was low and similar over Ni/17CeZr and NiCu/25CeZr, and slightly higher over Ni/25CeZr due to higher conversion. Thus, 1-PrOH dehydrogenation (Eq. 19) took place at a relatively low extent regardless of the catalyst. Nonetheless, the reaction stoichiometry of Eq. 19 matches the product distribution obtained over NiCu/25CeZr. In contrast, the product distribution obtained over Ni/17CeZr and Ni/25CeZr indicates that full reforming of 1-PrOH to gases and WGS (Eqs. 3 and 5) was the main reaction pathway to produce hydrogen, which was obtained in significantly larger amounts than the other products.Propan-1-ol dehydrogenation (19) C 3 H 7 O H   l   ↔ H 2 O   C 3 H 6 O   l + H 2   g   Δ G 503 = + 9,5   k J Propanal decarbonylation (20) C 3 H 6 O   l   ↔ H 2 O   C 2 H 6   ( g ) + C O   ( g )   Δ G 503 = - 82.7   k J Propan-1-ol decarbonylation (21) C 3 H 7 O H   l   ↔ H 2 O   C 2 H 4 g +   C O   g +   2 H 2   ( g ) Δ G 503 = +   1.3   k J Over Ni/17CeZr, C2H6 was produced in stoichiometric amounts with CO, in agreement with Eq. 20. An additional source of CO could have been the 1-PrOH decarbonylation (Eq. 21) accompanied by the production of C2H4 and H2. However, low C2H4 yield indicates that this reaction did not take place to a significant extent over Ni/17CeZr. CO2 and CH4 resulted from the WGS reaction (Eq. 5) and methanation of carbon oxides (Eqs. 6 and 7) respectively. However, CO2 and CH4 low yields suggest that WGS and methanation were less favoured. Over Ni/25CeZr, C2H6 was also produced accompanied by CO via propanal decarbonylation (Eq. 20). Carbon monoxide might have been also formed in the 1-PrOH decarbonylation (Eq. 21), which explains the production of C2H4 and additional H2. Higher CO2 yield than CO over Ni/25CeZr indicates that WGS (Eq. 5) was more favourable than over Ni/17CeZr. The presence of CH4 in the gases indicates that methanation of oxides (Eqs. 6 and 7) took also place over Ni/25CeZr.Carbon monoxide was not detected in the APR of 1-PrOH over NiCu/25CeZr and CO2, C2H6 and C2H4 were observed in negligible amounts. The presence of CH4 in the gas stream obtained over NiCu/25CeZr suggests that hydrogenation of carbon oxides took place (Eqs. 6 and 7). Accordingly, we conclude that NiCu/25CeZr mainly follows the reaction pathway in Eq. 19, which has been previously suggested by Lei et al. [24]. However, that suggestion differs from the observation by Wawrzetz et al. [23], who stated that decarboxylation of propionic acid to C2H6 and CO2 was the main reaction after formation of propanal from 1-PrOH.The conversion and product distribution in the APR of 2-PrOH was similar over Ni/25CeZr and NiCu/25CeZr (Fig. 4 b and c). Considering the main reaction products, acetone and H2, 2-PrOH dehydrogenation to the ketone (Eq. 22) was assumed to be the predominant reaction pathway, which has been previously proposed in [47]. Further reaction of acetone through decarbonylation (Eq. 23) results in CH4 and CO. As observed in Fig. 4 b and c, decarbonylation of acetone was limited over both catalysts. Nonetheless, higher acetone decarbonylation over Ni/25CeZr slightly lowered the H2 yield and increased the amount of CH4 among the products.Propan-2-ol dehydrogenation (22) C 3 H 7 O H ( l )   ↔ H 2 O   C 3 H 6 O   l + H 2   ( g )   Δ G 503 = - 12.4   k J Acetone decarbonylation (23) C 3 H 6 O l + H 2   ( g )     ↔ H 2 O   2 C H 4   g + C O   ( g )   Δ G 503 = - 119.2   k J In a different study [23] on the APR of 2-PrOH over Pt/Al2O3, acetone has been reported to be the only product and no H2 had been observed, in contract to the present work. No CO was detected among the gaseous products resulting from the APR of 2-PrOH (Fig. 4 b and c). This suggest CO conversion to CO2 through WGS reaction (Eq. 5), or hydrogenation of CO, and also CO2, to form CH4 (Eqs. 6 and 7) might have taken place under the reaction conditions. Negligible amounts of C2H6 and C3H6 were additionally detected.The product distribution in the APR of different alcohols originates from the CH and OH bond cleavage of those bonds adjacent to the CO functional group [46]. For MeOH, the cleavage of these bonds led to full reforming to gases according to the thermodynamically favourable Eq. 1 with negative Gibbs free energy. For the longer-chain alcohols, Gibbs free energies of full reforming to gases (Eqs. 2–4) have positive values. Therefore, full reforming of EtOH, 1-PrOH and 2-PrOH to gases was expected to happen to a lesser extent than from MeOH. Fig. 5 shows the main reaction pathways proposed for the APR of MeOH, EtOH, 1-PrOH and 2-PrOH.The APR of MeOH proceeds through OH and CH bonds scission (Fig. 5 a). First the OH bond cleaves resulting in the formation of methoxy intermediates before decomposition to CO and H2 [46]. Every hydrogen atom in MeOH is activated to produce molecular hydrogen, which explains the high H2 yield reached in the APR of MeOH (Table 4 and Fig. 4). To maximize the H2 production, CO should be converted in the WGS reaction (Eq. 5) and limit CO bond cleavage that takes place in side reactions, such as methanation, where H2 is consumed to produce methane (Eqs. 6 and 7). NiCu/25CeZr successfully limited CO bonds scission in the APR of MeOH.The APR of EtOH proceeds through OH and CH, and CC bonds scission when full reforming to gases and decarbonylation reactions take place (Fig. 5b). The experimental results elucidated that the reaction pathways followed by EtOH in APR appear to be different over Ni/17CeZr and Ni/25CeZr. The product distribution obtained over Ni/17CeZr suggests that full reforming to gases (Eq. 2) was dominant in accordance with the larger H2 yield obtained. Lower H2 yield and relatively significant amounts of CH4 suggest that EtOH decarbonylation was more favourable over Ni/25CeZr. Accordingly, the cleavage of multiple CH bonds from the alkyl group was more favourable over Ni/17CeZr. When only OH and CH bonds cleave in EtOH, ethanal was formed. The ethanal formation pathway via alcohol dehydrogenation (Eq. 16) was less favourable than the gas formation that involved CC bond cleavage over both Ni/17CeZr and Ni/25CeZr (Eqs. 2 and 18), in agreement with the negative reaction Gibbs free energy changes and the obtained product distributions (Fig. 4).Full reforming of 1-PrOH to gases (Eq. 3) was less thermodynamically favourable than the dehydrogenation or decarbonylation of the alcohol (Eqs. 19 and 21). However, the experimental results indicates that the APR of 1-PrOH proceeds through OH and CH, and CC bonds scission when full reforming and decarbonylation reactions took place (Fig. 5 c). These pathways were the most favourable over Ni/17CeZr and Ni/25CeZr. However, conversely to the APR of EtOH, full reforming and the cleavage of multiple CH bonds in the alkyl group were more favourable over Ni/25CeZr. When only OH and CH bonds cleave in 1-PrOH, propanal was formed. This reaction pathway was less favourable over Ni/17CeZr and Ni/25CeZr. In contrast, propanal formation was the preferred reaction pathway over NiCu/25CeZr. These results over NiCu/25CeZr confirm that full reforming to gases was inhibited in the APR of 1-PrOH by the addition of Cu to the catalyst.The APR of 2-PrOH mainly proceeds through CH and OH bond cleavage to form acetone (Fig. 5 d) over Ni/25CeZr and NiCu/25CeZr. Therefore, 2-PrOH dehydrogenation was the main reaction pathway in agreement with the spontaneous Gibbs free energy of Eq. 22. The CC and CH bond cleavages involved in full reforming to gases (Eq. 4) were unfavourable. Further CC of acetone to CH4 and CO was neither a significant pathway. However, this reaction pathway took place to a larger extent over Ni/25CeZr than over NiCu/25CeZr.Catalytic APR of C1-C3 alcohols was conducted over different nickel-based catalysts. The results of these experiments allowed the evaluation of the product distribution to propose potential reaction pathways followed by different alcohols in APR over nickel-based catalyst. In addition, Cu and Ce were used as dopants to assess their effect on the performance and stability of ceria-zirconia and alumina supported catalysts. The addition of Cu to the Ni-based 25CeZr-supported catalyst promoted the catalyst stability and more selective production of H2. The addition of Ce to the Ni-based 25CeZr-supported catalyst adversely affected the catalyst stability and activity. The other Ce-doped catalyst, NiCe/Al, promoted CO-free hydrogen production, and the undesired formation of CH4 was prevented in the APR of MeOH.Focusing on Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr, these catalysts performed differently in the APR of C1-C3 alcohols. The suggested reaction pathways in the APR of C2–C3 alcohols comprises full reforming to gases, and alcohol dehydrogenation and decarbonylation. The extent in which these reactions took place depended on the type of feedstock and catalyst. In the APR of MeOH, H2 yield was high due to high MeOH conversion via full reforming to gases and the subsequent WGS reaction. Larger amounts of ceria in the support allowed a higher MeOH conversion, and Cu-doping limited CH4 formation. In the APR of longer-chain alcohols, Ni/17CeZr and Ni/25CeZr were active in the cleavage of OH, CH and CC bonds for full reforming to gases. However, side reactions such as alcohol dehydrogenation and decarbonylation were significant. Over NiCu/25CeZr, C2-C3 alcohols mainly followed the dehydrogenation pathway. Thus, Cu restricted the full reforming of alcohols to gases due to lower activity in the CC bond cleavage, which limited the H2 yield.Ni/17CeZr, Ni/25CeZr and NiCu/25CeZr are potential catalysts to process the oxygenated hydrocarbons in FT-derived water fractions. The monometallic Ni/17CeZr and Ni/25CeZr are preferred to maximize the hydrogen production. Nonetheless, NiCu/25CeZr could be additionally considered because of its improved stability during the experiments, compared to the monometallic catalyst, and when more selective production of hydrogen among gases is required.The authors thank Dr. Pekka Simell, Prof. Klaus Hellgardt and Prof. Leon Lefferts for their guidance and support. We are grateful to Aleksi Rinta-Paavola for his help with the preparation and characterization of catalysts, to Tyko Viertiö and Eveliina Mäkelä for their help with the adsorption isotherm measurements, and to Laura Lonka for her help with the APR experiments. The Bioeconomy Infrastructure and the Raw materials research infrastructure (RAMI) that permitted conducting the experimental work for this study at both VTT and Aalto University. This work was funded by Academy of Finland (AQUACAT Project no. 285398).

Catalytic aqueous-phase reforming (APR) can be applied to process the organic compounds in the water fractions derived from the Fischer-Tropsch (FT) synthesis. This work aimed at finding an active nickel-based catalyst to convert organic compounds typically found in FT-derived waters, such as alcohols, into hydrogen. In addition, this work aimed at proposing potential reaction pathways that explain the product distribution resulting from the APR of C1–C3 alcohols. Solutions with 5% mass fraction of either methanol, ethanol, propan-1-ol or propan-2-ol in water were processed in APR at 230 °C and 3.2 MPa over different nickel-based catalysts in a continuous packed-bed reactor. Methanol was successfully reformed into hydrogen and carbon monoxide with conversions up to 60%. The conversion of C2–C3 alcohols achieved values in the range of 12% to 55%. The results obtained in the APR of C2–C3 alcohols suggest that in addition to reforming to hydrogen and carbon monoxide, the alcohols underwent dehydrogenation and decarbonylation. The most stable catalyst, nickel-copper supported on ceria-zirconia, reached feedstock conversions between 20% and 60% and high hydrogen selectivity. Monometallic nickel supported on ceria-zirconia catalysts reached higher H2 yields; however, the yield of side products, such as alkanes, was also higher over the monometallic catalysts. Accordingly, ceria-zirconia nickel-based supported catalysts constitute suitable candidates to process the alcohols in the water fractions derived from the FT synthesis.

Copper surface area, ( m Cu 2 · g - 1 )Avogadro’ s number, ( - )Copper molecular weight, ( g · mol - 1 )Copper dispersion, ( % )Average surface-volume copper diameter, ( nm )Copper density, ( g · m Cu - 3 )H2 consumption from 1st and 2nd TPR, respectively, ( mL H 2 · g cat - 1 )Copper crystallite dimension from XRD, ( nm )Catalyst solid density, ( kg · m cat , s - 3 )Catalyst apparent density, ( kg · m cat - 3 )Catalyst porosity, ( m pores 3 · m cat - 3 )BET surface area, ( m 2 · g - 1 )Pore volume, ( c m 3 · g - 1 )Pore diameter, ( nm )Gas hourly space velocity, ( NL · kg cat - 1 · h - 1 )CO2 conversion, ( % )Yield of product i, ( % )Space time yield of product i, ( mmol · h - 1 · g cat - 1 )Selectivity of product i, ( % )Catalyst weight, ( kg )Molar flow rate of component i, ( mol · s - 1 )Molar fraction of component i, ( - )Root mean square error of component i, ( - )Objective function, ( - )Number of experimental data, ( - )Inlet volumetric flow rate, ( NL · h - 1 )Temperature, ( K )Total pressure, ( bar )Stoichiometric number of component i in reaction j, ( - )Rate of reaction j, ( mol · s - 1 · kg cat - 1 )Total number of reaction, ( - )Kinetic constant of reaction j, ( mol · s - 1 · kg cat - 1 )Pre-exponential factor of the kinetic constant of reaction j, (depending on model)Adsorption constant of component i, (depending on model)Equilibrium constant of reaction j, (depending on reaction)Standard enthalpy of adsorption of component i, ( J · mol - 1 )Standard entropy of adsorption of component i, ( J · mol - 1 · K - 1 )Activation energy of reaction j, ( J · mol - 1 )Gas constant, ( J · mol - 1 · K - 1 )Carberry’s number, ( - )Second Damkohler number, ( - )Effective diffusivity, ( m 2 · s - 1 )Observed reaction rate per volume of catalyst, ( mo l · s - 1 · m cat - 3 )Order of reaction with respect to component i, ( - )Gas-solid mass transfer coefficient, ( m · s - 1 )Forward rate of reaction of component i, ( mol · s - 1 · kg cat - 1 )Concentration of species i in the bulk phase, ( mol · m - 3 )Concentration of species i on the catalyst surface, ( mol · m - 3 )Statistic indicator of the F-test, ( - )Critical value of F-, from Fisher distribution tables, ( - )Variance of the lack of fit, ( - )Variance of the experimental error, ( - )Number of variables (parameters of kinetic models), ( - )Partial pressure of component i, ( bar )Adsorption term of Bussche and Froment kinetic model, ( - )Adsorption terms of Graaf’s kinetic model, ( bar - 0.5 )Adsorption terms of Seidel’s kinetic model, ( - )Total amount of reduced centers, ( - )Relative contact free energy of Cu and CeZr, ( - )Inlet reactor conditionCO2 hydrogenation to methanol reactionReverse water gas shift reactionCO hydrogenation to methanol reactionExperimental valueCalculated valueThe combustion of hydrocarbons to produce energy entails a critical global challenge that needs to be tackled with urgency. The usage of fossil fuels correlates directly to the release of greenhouse gasses – especially CO2 – into the atmosphere, which is the main responsible of global warming [1,2]. Hence, in the last century research has been focusing on the development of carbon capture and storage technologies (CCS) first and, more recently, on the alternatives for CO2 utilization (CCU) [3–6]. An interesting approach for CCU is the CO2 reduction with renewable H2 to produce valuable chemicals and/or energy carriers [7]. In this context, the CO2 conversion to methanol is particularly appealing due to the high methanol demand worldwide (i.e. about 200 kton of methanol are used every day as chemical feedstock and transportation fuel) [8]. Indeed, methanol could be used directly as an alternative fuel or as intermediate for the production of dimethyl ether, olefins, gasoline and aromatics [9–11]. The CO2 hydrogenation to methanol is a catalytic gas phase process which follows three main reactions: the direct hydrogenation of CO2 to methanol (reaction 1), the production of CO through the r-WGS reaction (reaction 2) and the hydrogenation of CO to methanol (reaction 3). (1) CO2 hydrogenation: CO2+3H2 ⇄ CH3OH + H2O ΔH0=-49.5 kJ/mol (2) Reverse water gas shift: CO2+H2 ⇄ CO+H2O ΔH0=+41.2 kJ/mol (3) CO hydrogenation: CO+2H2 ⇄ CH3OH ΔH0=-90.5 kJ/mol Among these reactions, the CO2 hydrogenation to methanol is the most desired. Inevitably, the r-WGS takes place in parallel, accelerating the H2 depletion and, at the same time, contributing to the production of water. As a matter of fact, water is the main reaction by-product, which limits the system thermodynamically and causes catalyst deactivation [12]. Depending on the catalyst, the CO hydrogenation to methanol (reaction 3) could take place simultaneously, partially balancing the negative effect of the r-WGS. Nowadays, methanol is produced industrially from syngas feedstock (i.e., mixture of CO, H2 and c.a. 3% of CO2) at pressures of 50–80 bar and temperatures of 200–300 ⁰C, over CuO/ZnO/Al2O3 catalytic beds [13,14]. Since the benchmark technology involves only traces of CO2 in the feedstock [15], the corresponding catalyst is not necessarily optimal when using pure CO2, i.e., a thermodynamically very stable molecule, as the sole carbon source. Usually CO2 adsorption is not strong enough [16] and efforts are required specifically on novel catalyst formulations [17]. Over the years, researchers have proposed a variety of different catalysts for the CO2 hydrogenation to methanol, with particular focus on Cu-based systems, in combination with different metal oxides as carrier and/or promoters [18–21]. First, important research efforts aimed at replacing the hydrophilic Al2O3 support, which could deactivate in presence of the large amounts of water produced in all the reactions [22]. In most of the catalyst formulations, the ZnO oxide still acts as main promoter, since it guarantees both a higher Cu dispersion and the formation of Cuδ+ sites at the Cu-ZnO interface [23–25]. On the other hand, various carriers/promoters have been proposed in literature such as ZrO2 [17,25–29], CeO2 [22,30–34], Fe2O3 [34–36], SiO2 [37–39], and TiO2 [30,40–43]. Most recently, the synergistic effect of CeO2-ZrO2 mixed oxides has received particular attention due to their high redox ability, improved thermal stability [44] and superior oxygen storage capacity (OCS) [45], properties that have proved highly beneficial for different reactive systems, such as the oxidation of aliphatic C2 [46], the conversion of NOx [47], the reduction of NO by propene [48] and, most recently, for the CO2 hydrogenation to methanol [49–51]. The introduction of smaller Zr4+ ions into the CeO2 tetrahedron creates a defective fluorite structure, which facilitates the adsorption of oxygen [52]. Shi et al., [49] proposed for the first time a ternary CuO/CeO2/ZrO2 catalyst for the CO2 hydrogenation to methanol. They found that a Ce:Zr mass ratio of 1 optimizes the basicity of the system in favour of the CO2 adsorption capacity. Their Cu30Ce35Zr35O catalyst showed excellent reducibility and Cu dispersion, as well as a balanced distribution of Cu0 and strong basic sites to enhance the H2 dissociative-adsorption and the formation of the H2CO intermediate, which preferentially hydrogenates to form methanol. Wang et al., [50] investigated the reaction pathway via in situ DRIFTS analysis. They showed that a calcination temperature of 450 ⁰C increases the CuO surface area and the formation of Cu-Ce-Zr sites, which favour the formation of H* and bi/m-HCOO*, responsible for the high selectivity to methanol.In any catalytic process, kinetic modelling is an essential tool to support efforts on catalyst development, to elucidate reaction mechanisms as well as to aid reactor design and process optimization. Numerous kinetic models have been proposed over the years to describe the methanol synthesis, mostly on commercial catalysts [13,53–57]. However, the majority of the kinetic models trace back to the works of Graaf et al., [58] and Bussche and Froment [59]. Both models propose a Langmuir-Hinshelwood-Hougen-Watson (LHHW) mechanism with the dissociative adsorption of H2. Graaf et al., established a dual-sites mechanism (i.e., one for CO and CO2 and one for H2O and H2) where methanol is produced from CO2 and CO simultaneously. On the contrary, Bussche and Froment considered a mechanism where Cu is the sole active site and CO2 is the only carbon source for the methanol production. Even today, literature shows disagreements on the relative contribution of CO and CO2 to the methanol synthesis. For example, Liu et al., [60] propose at least four parallel reactions: CO-CO2 exchange, CO hydrogenation, CO2 hydrogenation and WGS, while Bowker et al., [61] proposed that CO2 is the only responsible of methanol synthesis, even when feeding CO/CO2/H2 mixtures. Interestingly, Yang et al., [62] proved that the CO2 and CO contributions to methanol synthesis varies with the operating conditions. More recently, Niels et al., [63] found that CO2 is the immediate source for methanol (i.e., CO2 pathway is one order of magnitude faster), whereas the presence of CO is inhibitory at low conversion due to competitive adsorption, and beneficial at higher conversion due to the removal of water via the WGS. Finally, L.C. Grabow and M. Mavrikakis [64] showed through DFT calculations that about 2/3 of the methanol comes from CO2 in the conventional process (i.e., syngas feed). However, the situation could be completely different with CO2-rich streams and other catalyst formulations.More recently, Park et al., [65] developed a model considering three-sites adsorption, where CO2 and CO adsorb on two distinct sites. In this study the authors carried out a rate determining step analysis (RDS) based on the mechanistic hypotheses earlier proposed by Graaf et al., to find the rate expressions that best fit the experimental data. Seidel et al., [56] reviewed the elementary steps involved in the three-sites adsorption mechanism, proposing an even more complex kinetic model. This was recently simplified by Slotboom et al., [66], who reduced the number of kinetic parameters considerably.Despite the extensive literature database of kinetic models and rate expressions for the Cu:ZnO system supported on either Al2O3 (i.e., benchmark formulation) or other metal oxides, kinetic modelling of the methanol synthesis remains an intriguing research topic, with at least two important open questions: 1) what are the type and number of the catalyst active sites involved in the methanol synthesis; and 2) which is the dominant C-source for methanol formation (i.e., CO/CO2) and the corresponding prevailing reaction pathway. In addition, and to the best of our knowledge, the kinetics of this reaction on novel catalysts such as Cu-Ce-Zr mixed oxides (i.e., better performant catalysts for the conversion of CO2) has not been investigated yet.Herein, the kinetic model of methanol synthesis from CO2 and H2 over a Cu-Ce-Zr mixed oxide catalyst is investigated by means of an RDS analysis for the single-site, dual-site and three-sites adsorption kinetic model, based on the most relevant mechanistic hypotheses retrieved from literature. A total of 6 kinetic models are compared with a complete set of 96 experimental data in the range of temperature, pressure, H2:CO2 molar ratio and GHSV of 200–260 ⁰C and 10–40 bar, 3–7 and 7500–24000 NL · kg cat - 1 · h - 1 , respectively.The preparation of the ternary catalyst according to the works of Shi et al., [49] and Wang et al., [50] is followed by in depth catalyst characterization and extensive kinetic tests. Statistical analysis of the data combined with physicochemical constraints are used as tool for model discriminations. This work pays particular attention to the relative contribution of CO2 and CO to the formation of methanol (i.e., methanol synthesis from direct and indirect route, respectively) under various reaction conditions, by means of a theoretical differential analysis. The identification of the kinetic model, together with a detailed analysis of the reaction rates and the interplay between CO2 and CO hydrogenation will lead to a better understanding of this system. In this study, we will gain insights into the reaction mechanisms, identify the active sites and their role within the methanol formation, which is key for further improvement of this catalyst formulation, as well as an essential tool for reactor and process design.To elucidate on the reaction pathway involved in the CO2 hydrogenation to methanol over a copper-cerium-zirconium mixed oxides catalyst, the most relevant kinetic models reported in literature have been explored and re-parametrized. All the available kinetic models can be sorted in three groups, based on the number of active sites considered in the formulation of the mechanism. A detailed discussion is given below.The most relevant kinetic model considering a single-site adsorption mechanism is the one developed by Bussche and Froment in 1996 [59]. The most important assumption is that CO2 is the sole carbon source for methanol synthesis. As a result, only reaction (1) and (2) take place on the Cu surface of the catalyst, where both H2 and CO2 undergo dissociative adsorption. According to the authors, the rate determining steps are: 1) the CO2 dissociation on the active sites, which releases surface oxygen for the rWGS reaction and 2) the hydrogenation of the formate species for the CO2 hydrogenation to methanol. The rate equations are reported in Eq. 3–5. (3) r 1 = k 1 p CO 2 p H 2 1 - 1 K 1 eq p H 2 O p CH 3 O H p H 2 3 p CO 2 β 3 (4) r 2 = k 2 p CO 2 1 - 1 K 2 eq p H 2 O p CO p H 2 p CO 2 β (5) β = 1 + b H 2 O / H 2 / 8 / 9 p H 2 O p H 2 + b H 2 p H 2 0.5 + b H 2 O p H 2 O - 1 The Bussche and Froment model considers three adsorption constants (i.e., b H 2 O / H 2 / 8 / 9 , b H 2 and b H 2 O ) and two kinetic constants (i.e., k 1 and k 2 ), for a total of 10 parameters to be optimized.The most important and widely employed kinetic model describing the methanol synthesis is the model developed by Graaf et al., [58] in 1988. In their first publication, the authors had already recognized the lack of agreement in the literature on whether the carbon source for the methanol production is CO or CO2. As a result, they developed a model including both pathways (reaction 1, 2 and 3). All the reactions are assumed to be based on a dual-site LHHW mechanism, where CO and CO2 adsorb competitively on one site (s1) and H2 and H2O adsorb competitively on a second site (s2), with dissociation of H2. The adsorption of methanol is once again neglected. The rate equations are reported in Eq. 6–8, with the two adsorption terms (i.e., Θ 1 and Θ 2 ) related to the site s1 and site s1 described in Eq. 9–10. (6) r 1 = k 1 b co 2 C Θ 1 Θ 2 (7) r 2 = k 2 b c o 2 B Θ 1 Θ 2 (8) r 3 = k 3 b co A Θ 1 Θ 2 (9) Θ 1 = 1 + b co p co + b CO 2 p CO 2 - 1 (10) Θ 2 = p H 2 0.5 + b H 2 O b H 2 p H 2 O - 1 where A , B , and C represents the driving force of the CO hydrogenation, r-WGS and CO2 hydrogenation to methanol, respectively. As a matter of fact, the authors provided also different expressions for the driving forces terms, which depends on the particular RDS for the specific reaction. All the (48) combinations are reported in Table 1 and were tested in this study.The model from Graaf et al., includes 3 kinetic constants (i.e., k 1 , k 2 and k 3 ) and 3 adsorption constants (i.e., b co , b CO 2 and b H 2 O b H 2 ), for a total of 12 kinetic parameters.More recently, Henkel modified the model developed by Graaf et al., excluding the CO hydrogenation to methanol [54]. His reparameterization was based on two sets of experimental results, obtained from two distinct set-ups: 1) a Berty reactor and 2) a micro-fixed bed reactor, from which he obtained two different set of kinetic parameters [55]. The rate equations proposed for the CO2 hydrogenation and the rWGS are reported in Eq. 11–12, which lead to a total of 10 kinetic parameters. (11) r 1 = k 1 b CO 2 p c o 2 p H 2 1.5 1 - p CH 3 O H p H 2 O p CO 2 p H 2 3 K 1 eq 1 + b co p co + b CO 2 p CO 2 p H 2 0.5 + b H 2 O / H 2 p H 2 O (12) r 2 = k 2 b CO 2 p c o 2 p H 2 1 - p CO p H 2 O p CO 2 p H 2 K 2 eq 1 + b co p co + b CO 2 p CO 2 p H 2 0.5 + b H 2 O / H 2 p H 2 O In 2014, Park et al., [65] proposed a reaction pathway, based on the mechanism developed by Graaf et al., with the introduction of a third adsorption site exclusively for CO2. The authors considered the methanol dehydration to dimethyl ether in their reaction scheme, which was discarded in our analysis since no traces of DME were detected during the experimentation. The rate equations are summarized in Eq. 13–15. The model from Park et al., involves 14 kinetic parameters to be optimized. (13) r 1 = k 1 b CO 2 p c o 2 p H 2 1.5 - p CH 3 O H p H 2 O p H 2 1.5 K 1 eq 1 + b CO 2 p CO 2 1 + b H 2 0.5 p H 2 0.5 + b H 2 O p H 2 O (14) r 2 = k 2 b CO 2 p c o 2 p H 2 - p CO p H 2 O K 2 eq 1 + b CO 2 p CO 2 1 + b H 2 0.5 p H 2 0.5 + b H 2 O p H 2 O (15) r 3 = k 3 b CO p co p H 2 1.5 - p CH 3 O H p H 2 0.5 K 3 eq 1 + b CO p CO 1 + b H 2 0.5 p H 2 0.5 + b H 2 O p H 2 O Few years later, in 2018, Seidel et al., [56] developed an even more detailed model based on three adsorption sites, reviewing also the elementary reactions involved and the rate determining step of each reaction. The active sites are distinguished as follows: ⊙ for oxidized surface centers, assumed as active center for CO hydrogenation ⋇ for reduced surface centers, assumed as active center for CO2 hydrogenation ⊗ as the active surface center for the decomposition of H2 The rate expressions are reported in Eq. 16–18 with the corresponding adsorption terms in Eq. 19–21. (16) r 1 = ϕ 2 k 1 p CO 2 p H 2 2 1 - p CH 3 O H p H 2 O p CO 2 p H 2 3 K 1 eq Θ ⋇ 2 Θ ⊗ 4 (17) r 2 = ϕ 1 - ϕ k 2 p CO 2 1 - p CO p H 2 O p CO 2 p H 2 K 2 eq Θ ⋇ Θ ⊙ (18) r 3 = 1 - ϕ k 3 p CO p H 2 2 1 - p CH 3 O H p CO p H 2 2 K 3 eq Θ ⊙ Θ ⊗ 4 (19) Θ ⊙ = 1 + b CO p CO - 1 (20) Θ ⋇ = 1 + b H 2 O b O b H 2 p H 2 O p H 2 + b CO 2 p CO 2 + b H 2 O p H 2 O - 1 (21) Θ ⊗ = 1 + b H 2 p H 2 0.5 - 1 The parameter ϕ represents the total amount of reduced center, while 1 - ϕ represents the number of oxidized centers. Slootbom et al., [66] have recently corrected the definition of ϕ , assuming a maximum coverage of the reduced center of 90% (Eq. 22). (22) ϕ = ϕ w - 0.1 The authors used the relation of Ovesen et al., [67] for the calculation of ϕ w , as follows: (23) ϕ w = 1 2 1 - γ ∗ γ 0 where γ ∗ γ 0 is the relative contact free energy of Cu and Zn, for the benchmark formulation, and of Cu and the CeZr solution for our system. The γ ∗ γ 0 ratio is calculated according to Eq. 24–25, with the introduction of a new kinetic parameter ( Δ G 3 ). (24) γ ∗ γ 0 = 1 - K 3 p H 2 p CO p H 2 O p CO 2 1 + K 3 p H 2 p CO p H 2 O p CO 2 (25) K 3 = e x p Δ G 3 RT In this model, the adsorption constant dependency on temperature is neglected. This lead to a total of 12 parameters, if the b H 2 O b O b H 2 group is parametrized as a single constant.In 2020 Slotboom et al., [66] simplified the three-sites model, drastically reducing the amount of parameters (i.e., 6 in the simplified version). The authors revisited the elementary reaction steps of Bussche and Froment, thus, considering only CO2 as the carbon source for methanol production, with the updates from recent literature. As Graaf et al., proposed in their study for the dual-sites adsorption mechanism, Slotboom et al., provided a tool for identifying the rate determining step for both the CO2 hydrogenation and the rWGS (i.e., the CO hydrogenation is neglected). All the possible rate expressions are summarized in Table 2 , with a total of 30 kinetic models, with 6 parameters each. The adsorption term, θ ⋇ , is defined by Eq. 26. (26) θ ⋇ = b H 2 p H 2 0.5 + b H 2 O / 9 p H 2 O + p CH 3 O H - 1 The Cu-Ce-Zr mixed oxides catalyst was prepared with a CuO loading of 50 wt%, to be comparable to the benchmark formulation, and a CeO2:ZrO2 mass fraction of 1, as recently optimized by Shi et al., [49]. The catalyst was synthesized via the gel-oxalate coprecipitation method [27]. The required amount of metal nitrate precursors (i.e., Cu(NO3)2·2.5H2O, Ce(NO3)3·6H2O and ZrO(NO3)2·6H2O) were solubilized in ethanol and coprecipitated by adding an oxalic acid solution (20 wt% excess) dropwise, at room temperature and under continuous stirring. The precipitate was stirred for 3 h, aged overnight, centrifuged and washed with deionized water, dried at 95 ⁰C for 16 h and calcined at 450 ⁰C for 4 h. The catalyst was pelletized, crushed and sieved to produce 50–125 µm particle size, to be used for the characterization techniques and reaction tests. The chemical composition of the synthesized catalyst was measured via microwave plasma atomic emission spectroscopy, using an Agilent MP-AES 4200 elemental analyzer. Prior to the analysis, about 0.1 g of catalyst sample was digested in 70 v.% HNO3 solution at 80 ⁰C overnight and then diluted with 5 v.% HNO3 solution (i.e., blank solution), to reach ppm values of the metal content. The specific surface area (S.A.) and pore volume (P.V.) were determined via the BET and BJH elaboration of the N2 adsorption–desorption isotherms at −196 ⁰C, obtained using a Micromeritics ASAP 2020 gas adsorption device. Before the measurement, the sample was degassed at 250 ⁰C for 2 h. The catalyst reducibility was studied via temperature programmed reduction (TPR) analysis performed using a Micromeritics AutoChem 2920 equipment with a TCD detector. The analysis was carried out in the range 50–400 ⁰C with a heating rate of 10 ⁰C·min−1, feeding 50 mL·min−1 of a 10% H2/Ar mixture. Prior to the TPR analysis, the sample was outgassed under inert conditions as for the N2 physisorption. The copper surface area ( S Cu ), dispersion ( D Cu ) and average surface-volume diameter ( d Cu SV ) were determined via N2O oxidation followed by H2 titration method developed by Van der Grift [68]. The analysis was carried out in the same equipment used for the TPR and consists in performing a first TPR measurement, whose hydrogen consumption is indicated by X . Thereafter, the temperature was reduced to 90 ⁰C and the sample was outgassed under Ar flow for 2 h. The surface copper was oxidized feeding 50 mL·min−1 of a 2% N2O/Ar mixture for 1 h. A second TPR analysis was carried out, whose hydrogen consumption ( Y ), is indicative of the number of Cu atoms dispersed on the surface of the catalyst. The copper surface area, dispersion and diameter were calculated with Eq. 26, Eq. 27 and Eq. 28, respectively, considering a Cu/N2O = 2 titration stoichiometry and a surface atomic density of 1.4·1019 Cuat·m−2. (26) S Cu = 2 Y · N av X · M Cu · 1.410 19 (27) D Cu = 2 Y X 100 % (28) d Cu SV = 6 S Cu · ρ Cu where N av , M Cu and ρ Cu are the Avogadro’s number, the copper molecular weight and density, respectively. X-ray diffraction (XRD) analysis in the 2θ range 10-120° was performed on the reduced catalyst with a MiniFlex600 machine (Rigaku) operating with a Ni β-filtered Cu-Kα radiant at 40 kV and 30 mA and a scan step of 0.05°/min. The diffraction peaks were identified according to the JCPDS database of reference compounds. The average diameter of the Cu-crystals was estimated via the Scherrer’s equation (Eq. 29). (29) d Cu = b λ F W M H c o s ( θ ) where d Cu is the dimension of the crystallites as if they were cubes, monodisperse in size, λ is the wavelength, FWMH is the width of the peak, 2 θ is the scattering angle and b is a constant usually varying between 0.89 and 0.94. XPS measurements were performed both on the calcined and reduced catalyst, 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. The binding energies were internally calibrated setting the C1s peak position at 285 eV. The catalyst real density ( ρ cat ) was measured using an automatic gas pycnometer instrument (Ultrapyc 1200e). The apparent density of the catalyst ( ρ b , c a t ) was calculated via the catalyst porosity ( ε cat ), determined from the N2 physisorption analysis. The catalytic tests were carried out in a stainless-steel reactor (dint, 10 mm), loaded with 0.25 g of catalyst, diluted with 0.75 g of SiC, to ensure isothermal operation and prevent sintering phenomena. The catalyst and the SiC used for dilution were introduced in the reactor with the same particle size of 50–125 µm. Larger SiC particles were used as pre-heating bed, separated from the catalytic bed with c.a. 1 cm3 of quartz-wool. The reactor was placed in an electric oven and heated more precisely via a heating mantle. The temperature was measured with two thermocouples, one at the beginning of the catalytic bed and one placed at the exit of the gases. Prior to the reaction tests, the catalyst was reduced in situ at 250 ⁰C, with 50 mL·min−1 of a 50% H2/N2 mixture for 4 h. The reaction mixture was analysed with a compact gas chromatograph (Global Analyzer Solution TM, G.A.S.) equipped with a TCD detector and two packed columns (HayeSep Q 60–80 mesh and 5A molecular sieve) for the analysis of permanent gases (i.e., H2, CO2, CO and N2) and an FID detector with capillary columns (Rtx-1, MTX-1 and MTX-QBond) for the analysis of the hydrocarbons. The experimental setup is sketched in Fig. 1 . The reaction tests were performed in a range of temperature and pressure of 200–260 ⁰C and 10–40 bar, respectively, feeding H2/CO2/N2 mixtures in different proportion, to have a H2:CO2 molar ratio from 3 to 7, and a GHSV ranging from 7500 to 24000 NL · kg cat - 1 · h - 1 . The carbon balance in the reaction was respected with a maximum error of 3%. The catalyst stability was observed within a long-term (100 h) test performed at 250 ⁰C, 30 bar, H2:CO2 molar ratio of 3 and a GHSV of 9600 NL · kg cat - 1 · h - 1 . The CO2 conversion ( X CO 2 ), product yield ( Y i ), product space time yield ( STY i ) and product selectivity ( S i ) were calculated according to Eq. 30–33, where i is either methanol or CO and w cat is the catalyst weight. Methanol and CO where detected as the sole carbon species in the product mixture. (30) X CO 2 = F CO 2 in - F CO 2 out F CO 2 in (31) Y i = F i out F CO 2 in (32) STY i = F i out w cat (33) S i = Y i · X CO 2 A commercial Cu/ZnO/Al2O3 catalyst from Johnson Matthey (i.e. Katalko-51) was tested in the exact same conditions, to compare the novel catalyst with the benchmark technology.The fitting procedure was carried out entirely in MATLAB R2019a. The kinetic parameters were determined via the fminsearch optimization procedure, based on the Nelder-Mead simplex algorithm [69], which minimizes an error objective function (OF) that we defined as the sum of the root mean square errors (RMSE) between the experimental and calculated molar fraction of the carbon containing species (i.e., CO2, CO and methanol) as follows: (34) RMSE i = ∑ k = 1 N data y i , k calc - y i , k exp 2 N data (35) OF = RMSE CO 2 + RMSE CO + RMSE MeOH Where N data is the number of experimental data used for the kinetic fitting and y i exp and y i calc are the molar fractions of the component i at the exit of the catalytic bed determined experimentally and via the model prediction, respectively. The experimental data were imported in terms of y i exp , together with the corresponding boundary conditions, such as inlet flow ( ϕ o ), inlet composition ( y i 0 ), temperature ( T ) and total pressure ( P ). On the other hand, the y i calc were determined, within the algorithm iterations, via the integral analysis method, thus solving the ODEs describing the mole balance equations in a fixed bed reactor (Eq. 36). (36) d F i d w cat = ∑ j = 1 N r r j ν ji where F i is the molar flow rate of the component i , w cat is the catalyst weight, N r is the total number of the reactions involved, ν ji is the stoichiometric number of the component i in the reaction j , and r j is the corresponding reaction rate expression, which is unknown. The mole balance equations were solved under the hypothesis of steady state regime, isothermal operation, negligible pressure drop along the catalytic bed and absence of internal diffusion and external mass transfer limitation. The first three hypothesis were confirmed experimentally: 1) the reaction performance was evaluated at steady state (i.e., when no changes in the outlet composition were recorder over time); 2) the temperature difference between the gas inlet and outlet positions was less than 1 ⁰C and 3) the pressure difference between the gas inlet and outlet positions was less than 0.2 bar. The absence of mass transfer limitations was explored with preliminary experiments (details in S.I.) and was later confirmed with the Mear’s [70] and Weisz-Prater [71] testing criteria.The reaction rates ( r j ) are function of the partial pressure of the components, and parameters such as kinetic ( k j ), adsorption ( b i ) and equilibrium constants ( K j eq ). The equilibrium constants (Table 3 ) were retrieved from literature [72].The kinetic constant of each reaction ( k j ) and adsorption constants of the components ( b i ) are the parameter to be optimized throughout the algorithm. The kinetic constants were described as a function of a pre-exponential factor and an activation energy, following the Arrhenius’ law (Eq. 37) [73]. The adsorption constants, instead, were expressed as a function of the standard entropy ( Δ S ads , i 0 ) and enthalpy of adsorption ( Δ H ads , i 0 ), according to the van ’t Hoff equation (Eq. 38). However, in some of the kinetic models considered in this study, the dependency on temperature of the adsorption constants is neglected (i.e., Δ H ads , i 0 ≈ R T ) [5666]. Furthermore, to reduce the number of fitting parameters, some authors lumped the adsorption constants of some components together. (37) k j = k j , 0 e x p - E a , j RT (38) b i = e x p Δ S ads , i 0 R · e x p - Δ H ads , i 0 RT The selected algorithm (fminsearch) requires an initial guess for the fitting parameters. Kinetic constant found in literature for the Cu-Zn-Al catalyst were implemented as initial guess, assuming these are likely of similar order of magnitude that the corresponding for our Cu-Ce-Zr catalyst [5513]. This minimizes the strong dependence that the algorithm has on the initial guess itself, and therefore increases the probability of obtaining meaningful results. To increase robustness of the model results, we setup a routine that evaluated the sensitivity of the model to the initial guess. This procedure consists of running the optimization algorithm in a loop, with newly obtained results as the initial guess. Thus, the convergence was reached when the difference between the algorithm output and the initial guess was less than 1%. Once the parameters of the best fit were obtained, the covariance matrix was computed with a second algorithm based on Levenberg-Marquardt method (lsqnonlin). From the covariance matrix, the standard deviation and the 95% confidence intervals, first indicators of the quality of the fit, were determined using the nlparci function in MATLAB. However, model discrimination techniques were necessary to find the set of rate expressions that best describe our system and, therefore, to gain insight into the reaction mechanism. A model was discarded at first when the physicochemical constraints (Table 4 ) were not respected. Thereafter, the significance of the model was assessed via the comparison of the variance of the lack of fit ( s 1 2 ) and the experimental error ( s 2 2 ), where s 1 2 > s 2 2 . The F-test (Eq.40) was carried out in combination with the analysis of the p-value (Eq. 41) (i.e., probability that the data belong to the non-critical area of the Fisher distribution), assuming 95% level of confidence (i.e., α = 0.05 ). The F statistic was first calculated according to Eq. 39, where s 1 2 is the variance of the lack of fit and s 2 2 is the variance of the experimental error. The F critical was retrieved from the Fisher distribution tables, considering N var and ( N data - N var ) as degree of freedom, where N var is the number of variables (parameters). The F statistic was then compared to the F critical (F-test, Eq.40). (39) F statistic = s 1 2 s 2 2 (40) F statistic < F critical = F 1 - α ( N var ; N data - N var ) (41) p > α As a result, the kinetic models fulfilling the physicochemical constraints were evaluated according to: 1) the value of the objective function, 2) the parity plots of the experimental and calculated flow rates of the carbon species, and 3) the outcome of the F-test and p-value.The absence of mass transfer (MT) limitations was evaluated according to the criteria reported in Table 5 , where the Carberry ( Ca ) and the second Damkohler number ( Da II ) are defined per component. The order of reaction with respect to the component i ( n i ) was estimated with Eq. 42, where r i + is its forward reaction rate [74]. Such derivation is specifically defined for complex reaction rates equation such as a LHHW kinetic. The correlations used for the mass transfer coefficient ( k gs ) and the effective diffusivity ( D e ) are reported in SI. (42) n i = p i ∂ ∂ p i l n ( r i + ) Table 6 summarizes the main physical properties of the Cu/CeO2/ZrO2 catalyst. The textural properties of the catalyst are in line with the literature [49,50]. The N2 physisorption analysis revealed an isotherm of type IV with hysteresis ( Figure S1 ), which is typical of a mesoporous material (i.e., pores in the range of 2–50 nm). The TPR profile ( Figure S4 a) exhibits two peaks at 204 ⁰C and 231 ⁰C, after deconvolution. No further reduction of the support, due to H2 spillover, was measured. As a result, a reduction temperature of 250 ⁰C is believed to be sufficient to reduce all the CuO, prior to the reaction tests. The XRD spectra on the calcined and reduced catalyst ( Figure S2 ) show the typical diffraction peaks of CuO at 2θ of 35.5⁰ and 38.7⁰ and of Cu at 2θ of 43.3⁰ and 50.4⁰, respectively. The disappearance of the CuO peak in the XRD spectrum of the reduced sample ( Figure S2 b) does not necessarily indicate the presence of sole metallic Cu, as CuO crystals smaller than 3–5 nm cannot be detected, as well as the Cu that is in contact with the Ce-Zr phases via O-bridges. The more complex Ce-Zr oxide phase was analyzed via XPS ( Figure S3 ). We confirmed the presence of the Ce3+ valence, which is introduced by the zirconia phase, as reported elsewhere [49,50]. The CuCeZr catalyst of this study is characterized by a Ce4+/Ce3+ ratio of c.a. 3.53, which was calculated through the integration of the corresponding peaks of the XPS spectra. A detailed discussion on the XPS results is given in S.I. The catalyst composition according to the MP-AES method is 52 wt% of CuO, 22 wt% of CeO2 and 26 wt% of ZrO2. This composition is very close to the theorical value, indication of the reliability of the synthesis method. In Fig. 2 a the catalyst performance during CO2 hydrogenation is compared to that of the benchmark formulation (i.e., the CuZnAl from JM). The CuCeZr catalyst shows a much higher methanol production compared to CO, with a crossover temperature (i.e., T cross temperature beyond which S T Y C O > S T Y M e O H ) of ca. 240 ⁰C. On the contrary, the benchmark catalyst shows a S T Y C O larger than S T Y M e O H over the entire temperature range. Since our catalyst formulation and preparation methods is a reproduction of previous works [49,50], we also compare the performance of this catalyst with that of the original reports by Shi et al., [49] (Fig. 2 b). The catalyst synthesized in this work shows higher methanol yield with respect to the different formulations proposed by Shi et al., However, the physicochemical properties of our CuCeZr are further improved with the calcination temperature (i.e., 450 ⁰C), according to the optimization reported by Wang et al., [50]. Unfortunately, insufficient details on the results reported by Wang et al., made a direct comparison with our results unreliable. However, assuming a catalyst density of 2.56 g · c m - 3 (i.e., value we measured), the S T Y M e O H they obtained at the same conditions is c.a. 6.6 m mol · h - 1 · g cat - 1 , which compares with the value reported in Fig. 2 b. The agreement of our results with literature underline the credibility of the method. Furthermore, they emphasize the promising performance of the CuCeZr catalyst with respect to the benchmark technology, in view of the CO2 valorization to methanol. Table 7 reports the information required for the model discrimination procedure, as discussed in Section 4.1. The rate determining step analysis (RDS) is carried out only when the author(s) reported the details behind the model derivations (i.e., Graaf and Slotboom). The discrimination between the different options proposed by Graaf and Slotboom is reported in S.I. and is based on the same criteria shown here. Both the models developed by Park and Seidel did not fulfil all the physicochemical constraints, thus, the statistics analysis was not carried out. It is worth noticing that all the models which do not consider the formation of methanol from CO (reaction 3), resulted in a low p - MeOH (i.e., p -value for methanol) which indicates the tendency of the model towards a scarce prediction of the methanol outlet molar fraction. This result anticipates the importance of considering the contribution of both CO and CO2 to the methanol synthesis, especially when CO2 is the sole carbon source.The model with the lowest RMSE (which corresponds to the final value of the objective function) and the largest p -values for the carbon species is the one proposed by Graaf. As a result, we select this model (Graaf- A 3 B 1 C 3 , where A 3 B 1 C 3 refer to the specific combination of RDS) to be the most representative of our system. The kinetic parameters obtained from the fitting procedure are provided in Table 8 . The accuracy of the parameter estimation is represented by the parity plots of CO2, CO, H2 and methanol (Fig. 3 ). The orders of magnitude of all the parameters are in line with the literature, especially when compared to the values retrieved from Graaf et al. Nevertheless, given the differences in the reaction rate expressions (i.e., A 3 B 1 C 3 for our CuCeZr catalyst versus the A 3 B 2 C 3 for the CuZnAl reported by Graaf), the comparison between the two kinetic models – and catalyst – is fair when observed in terms of reaction rates, rather than kinetic constants. Such analysis is addressed in Section 5.5. On the other hand, the adsorption term corresponding to the first and second active sites (i.e., Θ 1 and Θ 2 , respectively) do not differ from the original model. Therefore, in Table 9 we compare the values of our adsorption constants to the same constants calculated by Graaf at 200 and 260 ⁰C. From the reaction rate expressions (Eq. 6–10), we see that the adsorption constants of CO2 and CO contribute also to the driving force (i.e., numerator of the reaction rate). As a result, the prediction of their effect on the reaction rate is not straightforward. On the contrary, the combined adsorption of H2O and H2 (i.e., b H 2 O / b H 2 ) contributes only to the adsorption term in the denominator, hindering the reaction rate. Since our constant is order of magnitude higher than the one derived by Graaf, this leads to the conclusion that our catalyst is either more sensitive to water or to H2 adsorption.The model discrimination allows us not only to identify a model which better predicts the performance of our catalyst, but, most importantly, to gain some insights into the reaction mechanism itself. According to the assumptions behind the model developed by Graaf et al., we can distinguish between two active centers in the structure of the CuCeZr catalyst, which is in agreement with what was hypothesized in literature [49]: 1) the metallic copper (i.e., Cu0), where the dissociative adsorption of H2 occurs and 2) the oxygen defects within the Ce/Zr interface, where the CO2 molecule adsorbs and activates. The H species spillover towards the carbon atom of the activated CO2 to begin a series hydrogenation steps, knowns as “formate” route. The reaction pathway is sketched in Fig. 4 . According to the formate path, methanol can be either synthesized directly from CO2 (direct route) or indirectly from the CO produced via the rWGS reaction. The two routes overlaps when the H2CO intermediate forms, sharing the last two steps which then lead to the formation of methanol. However, at the point where the COs1 (i.e., CO adsorbed) intermediates appears, an equilibrium between the adsorbed CO and the CO released to the gas phase explains a certain selectivity to CO. The relative contribution of CO and CO2 to the methanol synthesis depends on different factors such as temperature, H2 concentration and the distribution of the Cu0 active sites with respect to the oxygen vacancies. Nevertheless, for a fixed catalyst composition, only reaction conditions can affect the fraction of methanol produced via the direct and indirect paths. A detailed discussion on this aspect is given in Section 5.5.In Fig. 4, the limiting steps of the three reactions are also marked. In particular, the slowest steps are the formation of the H3COOs1, HCOOs1 and H3COs1 intermediates for the reaction 1, 2 and 3, respectively. This result is in agreement with the in situ DRIFT studies carried out by Wang et al., [50], where the formation of the formate (i.e., HCOOs1) through the first hydrogenation of the carbon atom is defined as “the slowest and key step”.At this stage, being the reaction rate expressions determined, we could estimate the order of the reaction with respect to CO2 ( n C O 2 ) and evaluate the Ca and Da II numbers. We observed that in our experimental conditions, n C O 2 ranges between 0.094 and 0.62. Furthermore, both the external mass transfer and internal diffusion limitation resulted to be negligible, being the maximum value of Ca and Da II of 1.4·10-3 and 7.1·10-5, respectively. This result confirms our earlier conclusion that the experiments were carried out under kinetic regime.In this section, we discuss the catalyst performance as a function of the reaction conditions explored both experimentally and via model predictions. Experimental points and simulation results are combined in the same graphs, to show at the same time the quality of the fit. A detailed analysis of the thermodynamic equilibrium is reported in S.I.. Furthermore, to underline the compatibility of our results with the adopted equilibrium constant, the catalyst performance as a function of temperature and pressure – both experimental and modeling data – are reported together with the corresponding thermodynamic limit in Figure S10 . First, from Fig. 5 we observe that the kinetic model (solid lines) describes accurately the experimental reaction performance (points), in terms of X CO 2 (Fig. 5 a), Y MeOH (Fig. 5 b) and Y CO (Fig. 5 c) as a function of the space velocity (GHSV) at various temperatures. As expected from a kinetically controlled system, the conversion decreases with the space velocity, being the contact time of the gases with the catalytic bed shorter. Furthermore, Y MeOH and Y CO show the same trend as X CO 2 , leading to the conclusion that the contact time does not affect the product distribution in the range we explored. As a result, when employing a GHSV in the range 7500–24000 NL · kg cat - 1 · h - 1 , the CO contribution to the formation of methanol appears instantaneously, so that CO does not require additional contact time to react with the adsorbed hydrogen. Indeed, if that was the case, we would have observed an optimum in Y CO as a function of GHSV. Additionally, X CO 2 , Y MeOH and Y CO all show a clear increase with temperature, resulting from the positive effect that temperature has on all the reaction rates. Finally, it is important to notice that at the lowest GHSV ( Figure S10 ), the catalyst performance approach the thermodynamic equilibrium only at 260 ⁰C (i.e., highest reaction rate), where the thermodynamic value of X CO 2 , Y MeOH and Y CO at 30 bar is 21.1 %, 8.04 % and 13.1 %, respectively. As a result, in the temperature region 200–260 ⁰C X CO 2 , Y MeOH and Y CO still displays an exponential increase with temperature (i.e., kinetic regime). Fig. 6 displays the effect of temperature and total pressure on the methanol ( Y MeOH ) and CO yield ( Y CO ). As anticipated from Fig. 5, temperature positively affects all the reactions, since the effect of kinetics (i.e., Arrhenius type) overcomes the thermodynamics. Besides, we observe that the effect of temperature on Y MeOH (Fig. 6 a) is more significant as total pressure increases (i.e., the increase in Y MeOH from 200 to 260 ⁰C is of 91% and 193% at 10 and 40 bar, respectively). On the contrary, Y CO decreases with pressure and, at the same time, it keeps the same trend vs temperature, independently on the total pressure. As a result, the temperature of crossover shifts to higher values when pressure increases: at 10 bar the crossover occurs at c.a. 216 ⁰C, while at 40 bar Y MeOH > Y CO in the temperature region we explored (i.e., 200–260 ⁰C).In Fig. 6 b the effect of total pressure in the range of 10–40 bar is underlined: Y CO and Y MeOH exhibit two opposite trends, and the effect becomes more significant at higher temperatures (i.e., faster increase/decrease vs pressure). As discussed in section 4.2, methanol is formed via two parallel routes: 1) the direct one, which involves only reaction 1 and 2) the indirect one, which involves reaction 2 and 3, in series. As a result, when feeding only CO2 and H2 or, more generally, with CO2-rich streams, the direct route is faster than the indirect, since the latter needs the formation of CO first (i.e., r 3 is negligible for low values of p CO ). As soon as CO is formed, r 3 increases, causing an increase in methanol formation and, at the same time, a consumption of CO, which acts both as a product and a reactant. Such an effect is more noticeable at greater temperatures, because of faster reactions (i.e., the effect of pressure anticipates).Besides the effect of total pressure, a higher H2 concentration in the feed (i.e., higher molar feed ratio H2:CO2) causes an increase in both Y MeOH and Y CO (Fig. 7 a), independently of temperature. However, Y MeOH increases more than Y CO , shifting again the crossover point towards higher temperatures. As shown in Fig. 7 a, the model describes quite precisely the crossover point ( T cross ). Therefore, we used the model to predict the reaction performance in a wider range of H2:CO2 (1–10). We found that T cross monotonically increases up to an asymptotic value of 258 ⁰C at around H2:CO2 of c.a. 7 (Fig. 7 b). As a matter of fact, a higher H2 concentration facilitates its adsorption on the active sites, increasing the surface concentration of Hs2. As a result, as soon as CO forms, its hydrogenation is faster than its desorption to the gas phase, which enhances the indirect pathway once again (i.e., higher Y MeOH ). However, when all the active sites for H2 adsorption (i.e., Cu0) are saturated with H2, a further increase in its partial pressure does not affect the reaction rates anymore.Once defined the reaction rates, we tested the predictive capability of the model by using the model to calculate both X CO 2 and Y MeOH and comparing those values to an independent set of experiments (i.e., experimental data not used for the kinetic fitting). The kinetic model predicts quite accurately the experimental points obtained at lower GHSV (i.e., 2880 NL · kg cat - 1 · h - 1 , last 4 points) and at lower pressure (i.e., 28 bar, first two points), with a maximum deviation of 2.1% and 2.2% for X CO 2 and Y MeOH , respectively (Fig. 8 ).In this section, we analyse in more details the reaction rates and the relative contribution of the CO2 and CO hydrogenation (i.e., direct and indirect pathway, respectively) to the methanol formation. First, we calculate the reaction rates at different temperatures, via a theoretical differential analysis (i.e., assuming conversion values lower than 5%) at 30 bar and H2:CO2 ratio of 3 (Fig. 9 a). We observe that r 1 is the highest reaction rate at temperatures below c.a. 240 ⁰C. Therefore, at low temperatures, the CO2 hydrogenation to methanol is the fastest reaction, being its activation energy the lowest (Table 8). However, r 1 is the only reaction rate showing an optimum in the temperature range we explored. It is clear indeed, that r 1 approaches the equilibrium as temperature increases, being its value very close to zero at 260 ⁰C. As a result, we observe here the two opposite effects of kinetics and thermodynamics of an exothermic reaction. On the contrary, r 2 and r 3 are quite far from the equilibrium and both display the typical exponential behaviour of kinetically controlled reactions. In addition, reaction 2 and 3 proceed with similar velocities, with r 2 being slightly faster as temperature increases (i.e., r 2 / r 3 = 1.1 at 260 ⁰C). We clearly see that the two pathways for methanol formation behave differently with temperature. As a result, the relative contribution of CO2 and CO to methanol synthesis changes as temperature increases (Fig. 9 b). At 200 ⁰C, CO and CO2 contributes almost equally (i.e., 51.5% and 47.4% at 200 ⁰C, respectively). As temperature increases, CO-to-MeOH and CO2-to-MeOH exhibit opposite trends, with CO-to-MeOH reaching a value of c.a. 100% at 260 ⁰C. This result reveals why methanol selectivity does not decay with temperature as fast as it does on the CuZnAl catalyst (Fig. 2 a) and underlines the importance of designing a catalyst in such a way that CO adsorption is strong enough, to be able to proceed with the hydrogenation steps and form methanol, rather than desorb to the gas phase and contaminate the product stream.In Fig. 10 a, instead, we report the reaction rates as a function of the H2:CO2 ratio at 200 ⁰C and 30 bar. All the reaction rates remarkably increase with the H2 concentration. In particular, when H2:CO2 goes from 1 to 10, r 1 , r 2 and r 3 increase by ca. 30, 17 and 60%, respectively. As a matter of fact, all the direct reactions exhibit a positive order with respect to H2. However, expectedly, when the Cu0 active sites are saturated with H2, a further increase in the H2 concentration corresponds to a dilution of the carbon species, such as CO2 and CO, which also influence positively the reaction rates. This explains the slight decrease of the reaction rate (more noticeable for r 2 and r 3 ) beyond H2:CO2 of c.a. 7, which is in agreement with the result reported in Fig. 7 b. For completion, in Fig. 10 b we also report the relative contribution of CO2-to-MeOH and CO-to-MeOH crosses at H2:CO2 of c.a. 1.5, with CO showing the predominant contribution beyond the crossing point. This is a clear consequence of the influence that the H2:CO2 ratio has on the reaction rate. For H2:CO2 larger than 1.5, r 3 > r 1 and the contribution of CO surpasses that of CO2, following a trend which corresponds to the reaction rates r 3 and r 1 , respectively.To underline the potential of the CuCeZr catalyst, we propose here a comparison with the benchmark formulation (i.e., CuZnAl) in terms of reaction rates. First, the model derived by Graaf et al., was implemented and validated with the experimental results obtained for the CuZnAl catalyst (details on the validation are given in S.I.). Therefore, the kinetic model we adopted for such comparison is representative of the CuZnAl system and can be used for predictive studies. As depicted in Fig. 11 a, the CO2 consumption rate ( - r C O 2 ) increases exponentially with temperature and it is quite similar for both catalysts, with the CuCeZr showing a slightly faster consumption. However, the CuCeZr catalyst converts CO2 more selectively to methanol – including both direct and indirect route – than the CuZnAl catalyst. As shown in Fig. 11 b, methanol formation rate ( r MeOH ) its higher for the CuCeZr and crosses with CO ( r CO ) only at c.a. 256 ⁰C. On the contrary, the CuZnAl shows a much faster production of CO than methanol over the entire temperature range, which indicates that the CO hydrogenation does not contribute significantly to the synthesis of methanol, being CO the main reaction product. This demonstrates that the CuCeZr catalyst allows for a delay in the selectivity decay with increasing temperature when compared to the benchmark. It is clear that, in principle, lower temperatures favour the methanol production over CO. On the contrary, a higher temperature would correspond to much faster reactions, requiring less amount of catalyst to achieve equilibrium. In the end, when the desired product – in this case methanol – comes from an exothermic reversible reaction, the choice of the optimal temperature lies on a trade-off between reaction performance and economics. However, it is clear that the CuCeZr would facilitate the conflict between the demand of high performance and catalyst/reactor costs, since it allows to achieve higher methanol selectivity and faster CO2/CO conversion at higher temperature, when compared to the benchmark formulation.In this work, we investigate the kinetics of the CO2 conversion to methanol over a Cu/CeO2/ZrO2 catalyst, which remarkably outperforms the conventional Cu/ZnO/Al2O3 in terms of methanol yield/selectivity. The cross-over temperature (i.e., T cross , defined as the temperature above which the yield to CO exceeds that of methanol) increases up to 240 ⁰C for the CuCeZr, while CuZnAl shows a higher selectivity to CO in the entire temperature range.We analyse in detail the one-site, dual-site and three adsorption sites kinetic models, based on hypothesis retrieved from literature, and accordingly derived the kinetic parameters of all the models via an optimization algorithm based on the minimization of the RMSE (root mean square error). Physicochemical constraints and statistical indicators were used as tool for model discrimination. The best performing kinetic model (i.e., dual-site model of Graaf et al.,) suggests that the reaction mechanism proceeds via the adsorption of one of the oxygens of CO2 on the oxygen vacancies of the CeO2-ZrO2 phase (i.e., 1st active site), while H2 adsorbs and dissociate on the metallic copper (i.e., 2nd active site). The adsorbed hydrogen preferentially hydrogenates the carbon atom giving rise to the formate route. According to this mechanism, methanol can be formed either directly from CO2, or indirectly from the CO produced via the rWGS. The resulting kinetic model (i.e., rate expressions and fitted parameters) predicts the experimental data quite accurately, particularly the cross-over temperature (i.e., the model predicts that T cross stabilizes at 258 ⁰C at around H2:CO2 of c.a. 7.) Further, analysis of the individual reaction rates and the relative contributions of CO2 and CO to the methanol synthesis (i.e., COx-to-MeOH) reveal that CO2 and CO contribute evenly at 30 bar, H2:CO2 of 3 and 200 ⁰C (i.e., 51.5% and 47.4%, respectively), while the pathway CO-to-MeOH takes over at higher temperatures and/or higher H2 concentration. For H2:CO2 above 1.5, the CO contribution is predominant and exhibits an optimum at c.a. H2:CO2 of 7 (at 30 bar and 200 ⁰C) , which likely corresponds to the saturation of the Cu0 sites. This analysis underlines the importance of the indirect CO hydrogenation pathway in the reaction mechanism.In conclusion, these findings lead to a deeper understanding of the reaction mechanism of CO2 hydrogenation to methanol on novel CuCeZr systems, and serve as basis for future research into this catalyst formulation. For example, a more hydrophobic surface (i.e., weaker H2O adsorption and faster desorption from Cu0 sites) could lead to faster reaction rates and lower H2 requirement in the feed. Furthermore catalyst modification that lead to stronger CO binding would facilitate CO hydrogenation and, thus increase the selectivity to methanol even at higher temperatures.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 has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 838014 (C2Fuel project).Supplementary data to this article can be found online at https://doi.org/10.1016/j.cej.2022.134946.The following are the Supplementary data to this article: Supplementary data 1

This work addresses the kinetics of the CO2 hydrogenation to methanol over a Cu/CeO2/ZrO2 catalyst studied using single-site, dual-sites and three adsorption sites kinetic models. Physicochemical constraints and statistical indicators are used as tool for model discrimination. The best performing model is used to elucidate the reaction mechanism and the relative roles of the Cu-sites and oxygen vacancies. The results show that the dissociative adsorption of H2 occurs on the Cu0 sites, while CO2 is attracted to the oxygen vacancies created by the CeO2-ZrO2 solid solution. Then, the adsorbed H interacts preferentially with the carbon atom, favouring the so-called “formate” route. The CO formed via the r-WGS reaction could either desorb to the gas phase or react via hydrogenation to methanol. Analysis of the relative contributions of the CO2 and CO hydrogenation (i.e. direct and indirect pathways, respectively) to the methanol synthesis reveals that the latter is in fact preferential at high temperatures (i.e. about 100% of methanol is produced from CO at 260 ⁰C and 30 bar), and it shows an optimum vs the H2:CO2 ratio (c.a. 7 at 200 ⁰C and 30 bar), which corresponds to the saturation of the Cu0 sites with H2. Thus, this work provides an essential tool (i.e., kinetic model) for the design of reactors and processes based on novel catalysts, and importantly, it offers a deeper understanding of the reaction mechanism as basis for further catalyst development.

Magnesium alloys are constantly at the center of attraction in the selection and design of engineering materials due to its unique physical and mechanical properties for many years such as having low density and high strength-to-weight ratio [1,2]. In addition, magnesium metal is abundant on the earth's crust. Recent improvements in the mechanical properties of magnesium alloys have increased the usability of these alloys in aerospace and automotive industry applications too [3,4]. Moreover, they can be applied to 3C products (Computer, Communication and Consumer Electronics) where lightness, thinness and performance are a priority. However, one of the major factors that considerably restricts their application fields is the poor corrosion resistance of magnesium [5–7]. Thus, there are many studies focused on improving the corrosion resistance of magnesium alloys over the last few decades [8,9].There are many alternative surface modification methods to alter and enhance the corrosion resistance of magnesium alloys, such as physical vapor deposition (PVD) [10,11], chemical vapor deposition (CVD) [12,13], conversion coating [14,15], sol-gel method [16,17], electroless Ni-P plating [18–21] and micro-arc oxidation (MAO) [22–25]. The MAO process is quite similar to conventional anodizing with the notable exception of using high voltages within the range 200 V to 600 V. MAO coatings have high adhesion strength and exhibit exceptional wear resistance and distinguished corrosion resistance compared to coatings produced by other surface treatment techniques [26,27]. The MAO is also an environmentally friendly technology that has been widely used to produce relatively hard ceramic coatings on metals and alloys. During the reaction, a high potential is applied between the electrodes in the prepared electrolytes and the ratio of electrolytes affects the structure and corrosion resistance of the layer [28–31]. Therefore, many scholars have studied the effects of different electrolyte systems, such as silicate [32,33], phosphate [34–36], etc. on properties of micro-arc oxidation layers.Electroless nickel-phosphorus (Ni-P) plating has excellent properties, such as uniform deposition, high corrosion resistance, high wear resistance, good electrical as well as thermal conductivity, good lubricity and good ductility [37,38]. Electroless Ni-P plating on Mg alloy has attracted many scholars to study. However, a potential difference between the electroless Ni-P coating and the Mg alloy substrate which induces galvanic corrosion between the electroless Ni-P coating and the Mg alloy substrate seems to be a problem [39,40]. According to some studies, a MAO layer as an intermediate layer [41,42] with an outer electroless Ni-P coating on Mg alloy can provide good corrosion resistance.The aim of this study is to investigate how electrolyte composition influences the microstructure of the MAO coating and the corrosion resistance of the resulting MAO/Ni-P bilayer coating. The morphology, structure, adhesion and corrosion behavior of the bi-layered composite coating has been investigated by scanning electron microscopy (SEM), 3D white light interferometry. To the best of our knowledge, there are no scientific studies in the literature investigating the effect of phosphate concentration composite coatings on the AZ31B Mg alloy that include MAO and electroless Ni-P coating techniques. In this study, the MAO coatings were produced and the influence of phosphate addition on the corrosion resistance of MAO/Ni-P bi-layer coated AZ31B Mg alloy is then evaluated.The AZ31B Mg alloy (50 mm × 50 mm×0.2 mm) was used as a substrate in this study. Before MAO treatment, the AZ31B plates were grounded by sandpaper from 400# to 1200#, then washed with deionized water, degreased with alcohol and dried at room temperature. The MAO operation uses a ui-polar power mode with duty cycle 30％ and constant voltage of 400 V in an alkaline solution. The electrolyte composition and processing conditions for MAO treatment are listed in Table 1 . After MAO treatment, the sample was cleaned off with de-ionized water, wiped with alcohol, and dried at room temperature. And then, the electroless Ni-P coating was deposited on the MAO coated samples. The bath compositions and operation conditions of the electroless plating process were referring to our previous study [43]. The composition of the alloy is given in Table 2 .Scanning electron microscope (SEM, JEOL JSM-IT100) equipped with energy dispersive spectroscopy (EDS) microanalysis hardware was used to observe the surface and cross-section morphologies of the MAO and MAO/Ni-P composite coated samples. The samples were coated with Pt in 60 s to increase the conductivity. The average pore size and porosity were calculated using Image J software. 3D white light interferometry (Chroma 7503) was used to analyze surface roughness of the MAO. The XRD experiments were performed on a Bruker D2 PHASER X-ray diffractometer (λ =1.54184 Å, 30 kV and 10 mA) with Cu Kα radiation. The scanning range of diffraction angle (2θ) was 10° and 90° with a step width of 0.05° and time step of 0.5 s. The Posi-test AT-M pull-off adhesion tester (Defelsko) was used to measure the adhesion strength of the epoxy coating on the coated samples. The adhesion test was conducted according to ASTM D4541. The aluminum dollies with 20 mm diameter were attached onto the surface of the coated samples using two-component adhesives and were held for 1 h at 90℃ temperature to ensure that the adhesive was completely cured. Adhesion of Ni-P coatings on MAO coated samples were investigated by the pull-off test. Before applying force to pull the trolley from the surface of the sample, cut the area around the trolley all the way to the substrate, and then use the handle to repeat the up and down movement and increase the pull-up pressure for testing until the epoxy coating is removed from the substrate. All the tests were repeated five times to ensure the repeatability of the obtained results.Electrochemical tests were carried out using a VersaSTAT 4 potentiostat/frequency to analyze the corrosion behavior of Ni-P composite coating. A three-electrode cell, with a Pt flake counter electrode, a saturated Hg/Hg2Cl2/KCl as reference electrode and the sample as the working electrode (a circle with a diameter of 1 cm is the measurement area). Potentiodynamic polarization test and electrochemical impedance spectroscopy (EIS) of samples were measured in 3.5 wt.% NaCl solution. The samples were immersed in 3.5 wt.% NaCl solution for 20 min to achieve stable open circuit potential before electrochemical measurement. The potentiodynamic polarization test was performed from −300 mV to 500 mV in reference to the OCP with a scanning rate of 5 mVs−1. The EIS data were obtained at the open circuit potential and ambient temperature with a voltage amplitude of 10 mV in the frequency range from 10-2 Hz to 105 Hz for 24 h, 48 h and 72 h, respectively. The EIS data processing was carried by ZSimpWin 3.21 software. The salt spray test was conducted according to ASTM B-117. The sample was placed in a 5 wt.% NaCl solution at a pH value range from 6.5 to 7.2. This solution was atomized into a mist and the heating chamber was kept at 35 °C.Surface morphologies of the MAO coating treated in the different baths on AZ31B Mg alloy are shown in Figs. 1 and 2 . Electrolyte composition has a crucial impact on the MAO coating structure and the final coating characteristics such as surface morphology, thickness, roughness and corrosion resistance [28,30,31]. A typical cratered structure with micro-pores (pancakes-like structure) can be observed on the surface of both samples. The micro-pore in the cratered region is a discharge channel through which molten material was ejected from the coating/substrate interface due to the high temperature and strong electric field. After ejection Mg alloy rapidly solidified upon contact with the electrolyte. The pancake-like structure is rapidly solidified around the discharge channels [44]. In Figs. 1(a) and 1(b), the surface morphology of the MAO coating treated in the phosphate-containing bath shows similar pancake-like features and micro-pores which irregularly arranged on the coating surface. Luo et al. [35] and Ma et al. [36] reported that phosphate ion concentration accelerated the reaction rate during coating formation leading to an increase in the pore size, thickness and roughness of MAO coatings. It is found in Fig. 1(a) that the size of most of the pores (98%) is about 1–5 μm and the rest of the pores (2 %) has size of 5–10 μm in diameter. On the other hand, Figs. 1(c) and 1(d) shows that the micro-pores size of the MAO coating treated in the phosphate-free bath is approximately 0.8–1.5 μm, which is smaller than that of MAO coating treated in the phosphate-containing bath. Therefore, structurally independent morphologies in terms of pore size distribution and internal cracks are due to the different plasma thermochemical interactions between the substrate metal and different ions in the electrolyte solution [28,35,36]. Moreover, cracks can be obviously observed in the MAO coating prepared from the phosphate-containing bath. It implies that the roughness of MAO coating obtained from the phosphate-containing bath is higher than that from the phosphate-free bath.The cross-sectional morphologies of the MAO coatings on AZ31B Mg alloy were examined by SEM, as shown in Fig. 2. In Fig. 2(a), the cross-sectional morphology of the MAO coating treated in the phosphate-containing bath reveals the existence of an irregular metal-oxide interface. It exhibits some porosities or empty inclusions, where the crack appeared in the coating. It can be seen that this crack feature starts from the barrier layer between the substrate and the coating, and then propagates to the outer layer of the coating. Crack is generated due to thermal stress caused by the difference in the thermal expansion coefficient of Mg substrate and coating [44]. After MAO treatment in a phosphate-free bath, the thickness of the oxide layer becomes more homogenous. It can clearly be seen that the barrier layer of this coating has no porosities and more compact than the phosphate-containing sample in (Fig. 2(b)). Therefore, the thickness of the phosphate-containing sample (10.5–13.2 μm) is larger than the phosphate-free sample (7.8–10.3 μm).The XRD patterns of MAO coatings produced from different solutions are shown in Fig. 3 . The major crystalline phase is Mg2SiO4 in the phosphate-containing solution. Depending on the short coating time and therefore the thin coating thickness obtained, the structure of Mg2SiO4 is not detected in the phosphate-free solution. This finding is in accordance with the literature [45]. For both MAO coatings, MgO and elemental Mg are observed in the XRD pattern regardless of the compounds in the solution. The peaks that are related to elemental Mg is coming from the substrate metal [46–48]. The other crystalline structures are formed due to the plasma thermochemical reactions in the discharge channels of the MAO process [49–52].The corrosion resistance of the MAO coated samples on AZ31B alloy was conducted by the potentiodynamic polarization tests. Polarization curves of the bare AZ31B and the MAO coated AZ31B plates in 3.5 wt.% NaCl solution are displayed in Fig. 4 . Corrosion current densities (icorr ) and corrosion potentials (Ecorr ) are evaluated from the intersection of the linear anodic and cathodic branches of the polarization curves, shown in Table 3 . Ecorr means the corrosion potential where current density hugely increases according to the behavior of metal dissolution until it reaches a critical value. icorr means the corrosion rate. From Fig. 4 and Table 3, the phosphate-free sample has higher Ecorr (−1360 mV) and lower icorr (1.61 × 10−3 μA/cm2) than those of the phosphate-containing sample (Ecorr : −1360 mV; icorr : 6.87 × 10-3 μA/cm2). Due to the intense chemical interaction in the phosphate-containing electrolyte generates the formation of larger discharge pores and inner cracks at the MAO coatings. These porous structures and cracks play a key role in the corrosion process of MAO coatings [28,35,36,53]. On the other hand, the phosphate-free sample has a dense and continuous ceramic oxide layer. This protective layer prevents corrosive ions from reaching into the substrate metal compared to the phosphate-containing sample. Thus, the corrosion resistance of the phosphate-free sample is better than the phosphate-containing sample.The surface morphology of electroless Ni-P coating deposited on the different MAO coatings is shown in Fig. 5 . All images show a homogeneous, uniform and nodular-like structure on the electroless Ni-P coating. No distinct particles are observed over the coating surface because no grain boundaries are seen. It indicates that the uniform surface and amorphous structure of the Ni-P deposit can be obtained under our setting coating conditions. Fig. 6 shows the polarization curves of the bare AZ31B and the various MAO/Ni-P composite coated AZ31B plates in 3.5 wt.% NaCl solution. The potentiodynamic polarization data are also listed in Table 3. The average icorr of bare AZ31B, phosphate-containing-MAO/Ni-P coating and phosphate-free-MAO/Ni-P coating are about 65.4 μA/cm2, 2.36 μA/cm2 and 0.661 μA/cm2, respectively. It is obviously seen that the icorr value of phosphate-free-MAO/Ni-P coating is lower than others, indicating that this coating would form a more compact and protective passive film. On the other hand, the Ecorr of bare AZ31B, phosphate-containing-MAO/Ni-P coating and phosphate-free-MAO/Ni-P coating are about −1530 mV, −360 mV and −360 mV, respectively. In comparison with bare AZ31B, the Ecorr value of both composite coated AZ31B samples is obviously higher than that of bare AZ31B. Because of the Ecorr value is a parameter affected by the composition of the coating, which the MAO/Ni-P composite coating is completely covered and consistent with the results of Fig. 5. More positive Ecorr and smaller icorr imply that the phosphate-free-MAO/Ni-P composite coated AZ31B sample possesses better electrochemical resistance.The salt spray test was carried out for 120 h in 5 wt.% NaCl solution to investigate the corrosion protection of the MAO/Ni-P composite coated AZ31B plates. The data of the specimens were recorded each 24 h during SST for the purpose of studying the corrosion process, as shown in Fig. 7 . The phosphate-containing-MAO/Ni-P composite coated AZ31B plate has obvious pitting corrosion after 24 h of SST. Moreover, at the end of the test (120 h), this plate had more than 4.5 % corrosion area, which indicates severe corrosion. Contrary to the phosphate-containing-MAO/Ni-P composite coated AZ31B plate, the phosphate-free-MAO/Ni-P composite coated AZ31B plate still displayed the excellent corrosion resistance after 120 h of SST.The Table 1 shows that the phosphate-free-MAO/Ni-P composite coated AZ31B plate can provide better corrosion protection capability than other counterparts. According to our previous study [43], the electroless Ni-P coating causes severe damage to MAO coating. It might be ascribed to that local failure occurs resulting from the localized poor adhesion. Then, the electrolyte easily penetrates into the defect (micro-pores or cracks) and the corrosion of the coatings will initiate. Therefore, it is necessary to identify the adhesion of MAO/electroless Ni-P coating interface.The pull-off test was carried out to investigate the effect of adhesion for MAO/electroless Ni-P coating interface, as shown in Fig. 8 . The results of the adhesion test for the phosphate-containing-MAO/Ni-P composite coated AZ31B and the phosphate-free-MAO/Ni-P composite coated AZ31B are about 1.4 ± 0.21 MPa and 7.5 ± 0.15 MPa, respectively. The adhesion between the electroless Ni-P coating and the phosphate-containing-MAO coating is found to be weak and the electroless Ni-P layer easily strips off the MAO coating. In contrast, due to good adhesion to the phosphate-free-MAO coating, the electroless Ni-P coating directly is deposited and metallurgically integrated with the MAO coating.Appropriate surface roughness and microstructure are required in order to produce coatings with superior adhesion and corrosion resistance. Fig. 9 shows the 3D white light diagram of the different MAO coated AZ31B plates. The results of the average surface roughness (Ra ) for the phosphate-containing MAO coated AZ31B and the phosphate-free MAO coated AZ31B are about 0.935 μm and 0.330 μm, respectively. Although the phosphate-free sample has a lot of micro-pores and the microstructure morphologies are smooth (Figs. 9(c) and 9(d)), there is a good adhesion between the coating and the electroless Ni-P coating. The reason is that the surface of the phosphate-free sample forms a uniform oxide layer with good structure, so that it can improve the adhesion very well. Fig. 10 shows the cross-sectional SEM and BSE images of the MAO/Ni-P composite coatings. The electroless Ni-P coating is completely covered on both MAO coatings with the same thickness (approximately 10 μm), which is consistent with the results of Fig. 5. From Fig. 1, it illustrates that the micro-pores distribution of the MAO coating treated in the phosphate-free bath is uniform and dense. Besides, the catalyst used for the electroless Ni-P coating penetrates inside the porous structure and interacts well with the pores in the interface area between MAO/Ni-P composite coatings [42]. It can clearly be seen that the micro-pores on the electroless Ni-P coating/MAO coating interface were fully covered (Fig. 10(d)). They are adhered by mechanical and physical interlocking force [54]. The electroless Ni-P coating is deposited on the MAO coating treated in the phosphate-free bath on AZ31B with increasing adhesion. This leads to the conclusion that the incorporation of the catalyst has a significant effect on the improvement of mechanical properties for both samples (shown in Section 3.3). Fig. 11 shows the EDS/elemental mapping of the cross-sectional image of the phosphate-free-MAO/Ni-P composite coated AZ31B plate. It is evident that the element of the triangular-shaped plating part is Ni and the surrounding elements are Mg, O, and Si, which indicate that the electroless Ni-P coating completely covers the micro-pores of the MAO coating. Figs. 12 (a)-12(f) show the Nyquist and Bode plots for the experimental and fitting curves for all AZ31B MAO/Ni-P composite coatings after immersion in 3.5 wt.% NaCl solution for different time periods up to 48 h (recorded every 12 h), respectively. Considering the microstructure characteristics and EIS behavior of the composite coated samples, the models chosen for the fitting of the different MAO/Ni-P composite coated samples are depicted in Fig. 13 . These equivalent circuits are commonly used on Mg alloy [55–58], where Rs is the solution resistance. The constant phase element (CPE1) and resistance (R1) represent the properties of the electroless Ni-P plating; a parallel combination of a constant phase element (CPE2 ) and resistance (R2 ) represent the characteristic of the MAO coating. A good fit was shown between the experimental data and the simulated values (Figs. 12(a)-12(f)). The results are listed in Table 4 . The phosphate-containing-MAO/Ni-P composite coated AZ31B exhibits the highest total impedance in the as-deposited state (0 h), as shown in Figs. 12(a)-12(b). After immersion for 12 h, the total impedance shows a decrease as immersion time increases (from 0 h to 48 h). The change in the phase angle is also apparent (Fig. 12(c)), shifted to a higher frequency as immersion time increases. On the contrary, the total impedance for the phosphate-free-MAO/Ni-P composite coated AZ31B shows an increase as immersion time increases, as shown in Figs. 12(d)-12(e). Since the passive layer is formed on the surface of the Ni-P coating during the immersion process, the corrosion resistance can be improved [41]. According to the results of the adhesion test (Fig. 8), the electroless Ni-P coating was found to be weak adhesion to the phosphate-containing-MAO coating and easily stripped the MAO coating. Furthermore, there were larger micro-pores and micro-cracks on the phosphate-containing-MAO coating (Fig. 1), so the chloride ions easily corrode and penetrate the coating.The absolute impedance of the composite coating at low frequency (0.01 Hz) is plotted as a function of time in Fig. 14 . The figure shows that in all cases that the impedance (|Z|) of the phosphate-free-MAO/Ni-P composite coated AZ31B is greater than the other samples. The result indicates that the phosphate-free-MAO/Ni-P composite coated AZ31B provides good protection. When compared with the corrosion resistance of these two samples, it can be seen that the phosphate-free-MAO/Ni-P composite coated sample showed consistently higher corrosion resistance in 3.5 wt.% NaCl solution at all electrochemical tests than the phosphate-containing-MAO/Ni-P composite coated sample. This is mainly due to the flat and dense microstructure of the MAO coating. Also, evenly distributed pores on the surface help catalyst to be absorbed easily at the surface of the MAO coating. Thus, the Ni-P composite coating has good adhesion and corrosion resistance. It is concluded that the firm oxide layer was essential for improving the quality of the Ni-P layer coating so that AZ31B is well protected. Fig. 15 shows the cross-section morphology of the MAO/Ni-P composite coatings after the EIS measurement with immersion in 3.5 wt.% NaCl solution for 48 h. It is evident that the phosphate-containing-MAO/Ni-P composite coated sample is peeling and the Mg matrix is dissolved, which demonstrates that the coating is damaged and the corrosive solution reacts with the Mg substrate, as shown in Fig. 15(a). In contrast, the phosphate-free-MAO/Ni-P composite coated sample remains whole and the corrosion product plugs the micro-cracks and micro-pores in the coating, which indicates that this sample has good adhesive bonding and a compact structure with few defects in MAO layer, as shown in Fig. 15(b). As a final point, it is vital to possess coating structures such as a phosphate-free-MAO coating layer before applying the electroless Ni-P coating process, which is consistent with the results of Fig. 14.In this study, the influence of phosphate addition on the morphology and the corrosion resistance of MAO/Ni-P bi-layer coated AZ31B Mg alloy have been investigated.The morphology of the MAO coating treated in the phosphate-free bath was smooth and had a dense coating layer. The pores were evenly distributed at the surface and almost no defects were observed in the cross-section. On the other hand, the phosphate-containing MAO layer had many defects and uneven pores distribution, resulting in low adhesion values to the Ni-P coating. This caused poor adhesion protection to the substrate. However, the Ni-P coating formed a strong mechanical and physical interlock with the phosphate-free MAO coating due to the uniform distribution of pore size and shape. Therefore, it had better adhesion to the Ni-P coating with a value of 7.5 ± 0.15 MPa.The potentiodynamic polarization test showed that the phosphate-free-MAO/Ni-P coating had 2.8 times lower current density than phosphate-containing-MAO/Ni-P coating. The phosphate-free-MAO/Ni-P composite coated sample was shown to elevate the impedance in the EIS analysis for long-term immersion, indicating that this intact coating structure provided better corrosion protection. Finally, the salt spray test showed that the phosphate-free-MAO/Ni-P coating still maintained better corrosion resistance for 120 h. Therefore, it is concluded that the treated phosphate-free-MAO/Ni-P coating is superior to the phosphate-containing MAO coating in terms of mechanical and chemical properties. So, it is proposed that a new method of protecting the magnesium alloy has been developed.The authors declare no conflicts of interest.This study was financially supported by the Ministry of Science and Technology of Taiwan, Republic of China, under Grant No. MOST 106-2221-E-606-013-MY3.

A bi-layer coating is deposited on the surface of the AZ31B Mg alloy for corrosion protection of the Mg alloy. The bi-layer coating is composed of a micro-arc oxidation coating (MAO), and an electroless plated Ni-P coating. The micro-arc oxidation (MAO) treatment in the electrolyte with or without the addition of phosphate ions is carried out under unipolar power mode. The microstructure and composition of the MAO coatings are analyzed by scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS). The roughness of the MAO coatings is obtained by using 3D white light interferometry. The corrosion resistances are evaluated by potentiodynamic polarization test as well as electrochemical impedance spectroscopy (EIS) in a 3.5 wt.% NaCl solution and salt spray test. The adhesion test results of the MAO/Ni-P composite coatings by drawing test machine. The results showed that the micro-pores size of the MAO coating treated in the phosphate-free bath is uniform and smaller than the phosphate-containing solution, which could decrease the roughness and enhance the corrosion resistance. The corrosion resistance of MAO coating with the phosphate-free solution for 2.8 times was better than that of phosphate-containing solution. Moreover, phosphate-free MAO/Ni-P composite coating showed the corrosion rate is more than 4 times lower than the phosphate-containing MAO/Ni-P composite coating. Furthermore, the MAO/Ni-P composite coatings prepared by the MAO coating treated in the phosphate-free bath could suppress the erosion of aggressive media during exposure after 120 h salt spray test (corrosion area is approximately 5%) and present excellent adhesion.

Plastic pyrolysis, one of the routes to recycle low value polymeric wastes, is a well-known technology to produce useful liquid fuels and presents lower greenhouse gas net emissions than other current technologies such as incineration and gasification [1]. The liquid fraction derived from plastic pyrolysis, mainly in the range of gasoline and kerosene/diesel, can be used as feedstock for refineries or fuel for diesel engine generators [2,3]. However, many studies focused on application of pyrolysis oil are limited because they used selected neat polymers such as polyethylene (PE) and polypropylene (PP) for pyrolysis. In other words, the liquid product obtained from the pyrolysis of contaminated real municipal plastic wastes cannot be used in a direct industrial application, because of the presence of chlorine coming from poly(vinyl chloride) (PVC). There are many studies in the literature focused on reducing the chlorine content in the liquid products (in particular, the light liquid fraction) such as pyrolysis in the presence of a mixture of Ca(OH)2, Ni/SAPO-11, and Redmud [4]. ZnO/CoO adsorbent showed the highest dechlorination ability of a PVC pellet using superheated steam as the pyrolysis medium [5]. Stepwise pyrolysis of PVC containing plastic wastes also reduced the chlorine content in a liquid product [6]. However, in the case of high-boiling heavy wax, few studies on enhancement of properties and applications have been reported, likely due to the difficulty to find a suitable use thus far. The high-boiling heavy wax is a dark brown in a semi-solid phase and has a wide carbon-range distribution with high chlorine content. An adequate solution for application of the worthless heavy wax is thermal, catalytic, or hydro-cracking of heavy wax into lighter hydrocarbons in the range of gasoline and kerosene/diesel [7–9]. Catalytic cracking of heavy fractions obtained from the pyrolysis of automotive waste plastics was performed using a commercial FCC equilibrium catalyst in a fixed bed reactor with the flow of nitrogen at 525 ℃ [10]. The heavy wax was diluted in atmospheric or hydrotreated vacuum gas oils and was converted into high yields of gas and gasoline fractions, which were dependant on the kinds of gas oils and polymers. Raw pyrolysis wax oil from municipal plastic wastes (refuse plastic fuel, RPF) was catalytically upgraded on the zeolites under atmospheric pressure at 450 ℃ [8]. Catalytic cracking of heavy wax oil using HZSM-5 catalyst resulted in high yields of light hydrocarbons. However, chlorine issues were not discussed in relation to catalytic cracking of heavy fractions due to the use of pyrolysis oil obtained from selected polymers such as PE and PP. For the removal of chlorine compounds in the pyrolysis oils (chlorine content, 600 ppm), metal oxides including iron oxide, ZnO, MgO and Redmud were used as catalysts [11], Among them, iron oxide and an iron oxide carbon composite were active and stable to remove organic chlorine compounds from the oils (chlorine content, < 100 ppm). This result shows that iron oxide is suitable for the refinement of low-boiling pyrolysis oil as there is no considerable change in the carbon number distribution of the oils. Chlorinated pyrolysis oils (chlorine content, 0.8∼0.9 wt.%) were refined by using Redmud, which mainly contains Fe2O3, in a batch reactor at 325 ℃ with auto-generated pressure [12]. About 45∼71 % of the light liquid fraction was produced with no olefins and very low chlorine content (< 0.1 wt.%). In comparison to thermal and catalytic cracking of oils, catalytic cracking using Redmud was slightly more effective to remove chlorine from liquid products. Most chlorine including HCl and chlorinated organic compounds was present in a gaseous phase in the case of thermal cracking, whereas chlorine is trapped physically and chemically in the form of HCl and FeCl3 in a solid phase (Redmud and heavy residue) during the catalytic cracking. Furthermore, it was reported that Redmud promoted the cracking reaction of chlorinated pyrolysis oil as well as a dechlorination reaction. However, in the aforementioned studies, upgrading and the cracking degree were not adequately evaluated because the chlorinated oils contained a large fraction of gasoline (above 80 %, C5∼C10), meaning that it is questionable whether Redmud can break up the long-chain carbon bonds in the heavy wax oil. As a result, it is necessary to achieve an integrated catalytic upgrading of chlorinated heavy wax including catalytic cracking of wax into a light oil fraction and dechlorination to obtain light oil with very low chlorine content to find suitable applications of worthless heavy wax.Here, we performed catalytic cracking of chlorinated heavy wax obtained from pyrolysis of contaminated real municipal plastic wastes using iron oxide impregnated zeolite catalysts and discussed the catalyst effects on properties of liquid products such as cracking degree of heavy wax and chlorine content in the products.Chlorinated heavy wax was obtained through the pyrolysis of RPF in a commercial rotary kiln-type pyrolyzer (COcom Co., South Korea). RPF was added to the externally heated rotary kiln reactor, which was operated at 400 ∼ 500 ℃ and the residence time of 10 ∼15 h. The chlorinated heavy wax used in this work was yellow-brown and in a solid state at room temperature (Fig. 1 ). Elementary analysis (EA), carbon-distribution analysis using a simulated distillation gas chromatography system (SIMDIS-GC), and paraffins-olefins-naphthenes-aromatics (PONA) analysis using gas chromatography (GC) were carried out to obtain information of the raw material. Each analysis method is described in the following section in detail. A thermogravimetric analysis (TGA) of the heavy wax was conducted using a TGA Q500 with air to provide the amount of non-volatile components in the wax.Pellet-type (1.5 mm diameter, clay binder) HY zeolite (SiO2/Al2O3 = 100) was purchased from TOSOH USA, Inc. and it was pretreated at 550 °C for 4 h in air prior to being used as a substrate. Iron-impregnated catalysts were prepared by the incipient wetness impregnation method using an aqueous solution of FeCl2·4H2O (Sigma-Aldrich). The pretreated HY zeolite pellets were mixed with the iron- aqueous solution (8∼9 wt% solution) in a rotary bottle (1000 mL), followed by vacuum evaporation (BÜCHI, R-100) at 70 °C and about 350 mbar until the water was gone. The Fe-impregnated catalysts were dried at 105 °C overnight, followed by calcination at 550 °C for 4 h in air. The prepared catalysts were designated as Fe[x]/HY, where x means the impregnated weight percentage of Fe based on the HY support.The physical properties of the prepared catalysts were determined using Brunauer-Emmett-Teller (BET) technique using a BELSORP-mini II (MicrotracBEL Co. Ltd.).The catalysts were pretreated at 105 °C. NH3-temperature programmed desorption (NH3-TPD) was conducted to determine the surface acidity of catalysts using an AutochemII 2920 analyzer (Micromeritics Co. Ltd). After pretreatment with He at 400 °C for 1 h, ammonia was adsorbed on the surface of the catalyst at 50 °C for 1 h and then was switched to He to remove NH3 physically adsorbed on the catalyst. The adsorbed NH3 was desorbed with an increase of temperature according to the NH3-TPD routine. An X-ray diffraction (XRD) analysis was conducted to verify the impregnated iron oxide states using a Rigaku Smartlab XRD system with Cu Kα radiation. TGA and a differential thermogravimetric analysis (DTG) of the used catalysts were conducted using a TGA Q500 with air to provide information on the volatility of coke and chlorine deposited on the used catalysts.A schematic drawing of the experimental set-up for catalytic cracking of heavy wax and the product-recovery system is presented in Fig. 2 . The received heavy wax was heated using a heating mantle (80 °C) to make it flow well and the preheated wax was fed to a reactor using HPLC pump (Eledx Laboratories, 0.001–10 ml) at 0.29 g/min. The amount of loaded wax was measured using a data logger, connected with a balance (A&D Company) during the experiments. The line between the head of the HPLC pump and the inlet of the reactor was also heated using a heating band, at temperature of 170 °C. The prepared catalysts were placed in a quartz bulb-reactor (ID = 4 mm) situated in a heating furnace. Weight-hourly-space-velocity (WHSV) was set to 2 h−1, unless otherwise stated and N2 (50 m/min) was used as a carrier gas. A K-type thermocouple was located in the reactor to control the reaction temperature (450 ℃) and the internal pressure of the reactor was displayed. The produced vapor was first condensed at 70 °C and then re-condensed at 4 °C to easily recovery the liquid products. HCl in non-condensable gas was trapped in a water trap (10 in Fig. 2) and then water was also removed in the next trap (11 in Fig. 2).The composition of the non-condensable gas was analyzed using an on-line gas chromatograph (GC, Agilent 7890A) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID) and the total flow rate of gas was determined using a gas-flow meter (Sensidyne Gilibrator2). The liquid product was recovered for 1 h after reaching a steady-state of the reaction system, unless otherwise stated. EA, SIMDIS-GC, and GC–MS analyses were conducted for the liquid products. EA including C, H, N, O and Cl content was measured by an element analyzer (Thermo Scientific, Flash 2000). A SIMDIS-GC (Agilent 24001.070) was used to obtain the carbon distribution in the liquid product and compare the characteristics of liquid products obtained from the various reaction conditions. The GC was equipped with an AC (Analytical Controls) capillary column and a FID. The sample was diluted with carbon disulfide according to ASTM D2887 prior to the SIMDIS-GC analysis. A PONA analysis of the liquid product was performed using a GC/FID (Agilent 6890 N) and a GC–MS (Shimadzu Q-2010 plus) equipped with a ZB-DHA-PONA column. The sample oil (0.1 g) was diluted 20 times with acetone and hexane, respectively. The diluted samples were centrifuged at 10,000 rpm for 2 min. (Smart 13, Hanil Science Inc.), followed by filtration with a 0.45 ㎛ syringe filter (Sartorius). The compounds from the GC–MS profile were identified by comparison of their retention times with those of the standard compounds from a P-I-A-N-O standard kit (Sigma-Aldrich), composed of 11 n-paraffins mix (0.1 mL), 37 isoparaffins mix (0.1 mL), 37 aromatics mix (0.1 mL), 30 naphthenes mix (0.1 mL), 25 olefins mix (0.1 mL), and P-I-A-N-O mix (138 n-paraffins, isoparaffins, aromatics, naphthenes, and olefins, 0.1 mL).Chlorine quantification of the liquid products and HCl-trapped water was carried out using an AC600 Semi-auto Calorimeter (LECO. Co., USA), a Metrohm Ion Chromatograph with Mitsubishi combustion module, and a Dionex Integrion HPIC System (Dionex Co., USA), respectively. For examination of chlorine and coke formation on the surface of the used catalysts, a scanning electron microscope-energy dispersive X-ray spectrometer (SEM-EDX, S-4700, Hitachi) analysis was performed.The liquid fraction (wt.%) was calculated from the amount of the liquid product based on the amount of reactant fed into the reactor during the reaction time. The solid fraction (wt.%) including coke, char, and residues which could deposit on the surface of the catalyst and in the wall of the reactor, was determined by dividing the amount of solid by the amount of reactant fed. In here, the solid weight was calculated from the difference between the two weights of the catalyst and the quartz reactor before and after the reaction. In addition, the gas fraction (wt.%) was calculated from the difference between 100 and the liquid and solid fractions. Note that the small amounts of resides and tar that accumulated in the feeding line from the pump to the inlet of the reactor were ignored and trapped HCl gas was also ignored in the calculation.A photograph (a) and the carbon-number distribution (b) of heavy wax are shown in Fig. 1. The SIMDIS-GC data showed that the heavy wax has a much higher boiling point and a very wide carbon distribution: 1.54 wt.% of gasoline fraction (C5-C10), 37.91 wt.% of kerosene/diesel fraction (C11-C22) and 60.55 wt.% of heavy wax fraction (≥C23). Based on the results of the PONA analysis, n-paraffin (87.74 wt.%) was the main components of the wax in all carbon ranges (not shown here). Table 1 shows the elemental composition of the heavy wax. The sample mainly consisted of carbon and hydrogen with a few heteroatoms and 0.14 wt.% of chlorine was detected, which should be removed for the utilization of such heavy waxes. The heavy wax had the low-level non-volatile components (0.51 wt%), based on the TGA data (up to 900 ℃ in air). A strong weight loss begun at around 150 °C and most of the heavy wax was volatilized at around 530 °C.The textural properties of the prepared catalysts are summarized in Table 2 . It was found that the BET surface area and total pore volume decreased by impregnation of Fe on the pellet-type HY zeolite. In the case of the Fe[20]/HY zeolite, the surface area and total pore volume decreased by 27 % and 31 %, respectively, compared to those of the HY zeolite. Fig. 3 displays the XRD patterns of the prepared catalysts (a to d). The diffraction peaks (2theta = 6.3, 10.3, 15.9 and 24.1°) observed in the XRD pattern of the HY zeolite could be assigned to the characteristics of a faujasite zeolite (FAU) structure [13]. For all Fe impregnated HY zeolite, iron oxide (α-Fe2O3) was observed (2theta = 33, 35, 49, 55, 62 and 64°) and its intensity increased with an increase of the Fe loading amount.The acidic properties of the catalysts evaluated by an NH3-TPD analysis are also summarized in Table 2. The amount of acid sites was measured from three distinct desorption peaks divided by the desorption temperature of 120–300 ℃ (weak acid sites), 300–500 ℃ (medium acid sites), and 500–700 ℃ (strong acid sites), respectively. All prepared catalysts had only low- and medium-strength acids. The amount of total acid of the parent HY zeolite was 0.593 mmol/g. When 3 and 5 wt.% of Fe were impregnated on the HY zeolite, the amount of total acid sites increased to 0.688 and 0.845 mmol/g, respectively. However, the total acidity of Fe[20]/HY was lower (0.595 mmol/g) than that of Fe[5]/HY. This means that weak- and medium-strength acid sites could be derived from the interaction between the surfaces of iron oxide and HY zeolites. However, excessive impregnation of Fe appeared to partially cover the acid sites on the surface of the HY catalyst. The volcano trend of the total acidity for Fe impregnated HY catalysts was in line with a previous study [14].Results of thermal cracking and catalytic cracking at 450 ℃ are shown in Fig. 4 :(a) liquid, solid and gas fraction, (b) carbon distribution in the liquid product, and (c) PONA distribution in the liquid product. As shown in Fig.4(a), thermal cracking without any catalysts showed 69.8 wt.% of liquid yield, 18.6 wt.% of gas yield and 11.7 wt% of solid. When using the HY zeolite in the cracking reaction, the liquid yield increased to 71.3 wt.% and the solid was significantly reduced to 0.5 wt.%. However, the highest gas yield (28.2 wt.%) was obtained due to the active performance of HY toward the cracking reaction. In the case of the Fe impregnated HY zeolite, the highest liquid yield (74.2 wt.%) was obtained, indicating that heavy wax oil was less cracked compared to the case of using HY zeolite, and 9.6 wt.% of solid was formed. The generated gases were methane, ethane, ethylene, propane, propylene, etc. in all cases and a trace amount of hydrogen was generated. With regard to the carbon distribution in the liquid products (Fig. 4(b)), thermally cracked liquid consisted of heavy hydrocarbons (63.5 wt.%, ≥C23) and kerosene/diesel hydrocarbons (36.0 wt.%, C11∼C22), which is similar to the fraction of raw heavy wax. Considering gas yield (18.6 wt.%), the cracking reaction of hydrocarbons occurred, but high-boiling hydrocarbons and cokes were mainly formed during the thermal cracking reaction. For the HY catalyst, most of the heavy fraction was broken down into kerosene/diesel (45.1 wt.%) and gasoline (49.4 wt.%) fractions. When using Fe[3]/HY, the largest kerosene/diesel fraction (52.2 wt.%) was achieved, but 9.9 wt.% of heavy fraction remained. For the PONA distribution of the liquid products (Fig. 4(c)), the kerosene/diesel fraction in the thermally decomposed liquid showed mainly n-olefins. The heavy fraction was a mixture of n-paraffins and n-olefins. Although the carbon distribution of the thermally cracked liquid was similar to that of heavy wax, the main type of hydrocarbon were different; n-olefin for the thermally cracked liquid and n-paraffin for heavy wax. In the case of catalytic cracking of heavy wax, regardless of Fe impregnation on HY zeolite, n-paraffins in heavy wax disappeared and various types of hydrocarbons were detected in the liquid product: i-paraffins, naphthenes and aromatics as well as n-olefins.The effect of the Fe loading amount to the HY zeolite on the product yield (a), carbon distribution in the liquid product (b), and PONA distribution in the liquid product (c) is shown in Fig. 5 . Unlike the solid yield (0.5 wt.%) of the parent HY catalyst, the solid yields of Fe impregnated HY catalysts increased although a trend was not found in the relationship between loading amount of Fe and solid yield (Fig. 5.(a)). However, a correlation between total acid sites and activity toward the cracking reaction was identified. The Fe[5]/HY catalyst containing the largest amount of total acid sites presented the highest gas yield (22.6 wt.%), whereas Fe[20]/HY with the lowest total acid sites among the Fe impregnated HY catalysts showed the lowest gas yield (12.4 wt.%). The influence of different loadings of Fe on the carbon distribution in the liquid products also demonstrates that Fe[20]/HY has the lowest cracking acitivty: 20.3 wt.% for gasoline fraction and 27.4 wt.% for heavy fraction in the liquid product (Fig. 5(b)). Herein, it should be mentioned that, although the parent HY and Fe[20]/HY catalysts showed similar total acidity (Table 2), the two catalysts showed clearly different yields and carbon distribution in the liquid product. The differences are likely due to the presence of different Brønsted and Lewis acid sites: HY zeolite with more Brønsted acid sties and Fe[20]/HY with more Lewis acids sites [14], where HY having strong Brønsted acid sties is more favorable for catalytic cracking of wax [15]. The results obtained from the PONA analyses of the liquid products (Fig. 5(c)) demonstrate that all Fe impregnated catalysts consisted of i-/n-paraffins, n-olefins, naphthenes, and aromatics and the contents of n-olefins increased with Fe loading amount. Fig. 6 shows a comparison of chlorine contents in the liquid and gas products (a) and on the surface of used catalysts (b). The longer reaction time (time on stream (TOS) = 2 h) for catalytic cracking reaction than that of thermal cracking reaction (TOS =0.5 h) was applied to effectively compare the effect of chlorine removal of the prepared catalysts than thermal cracking reaction. For the thermal cracking of heavy wax, the liquid product contained a high concentration of chlorinated compounds (0.09 wt.%) and the concentration of chlorine captured from non-condensable gas product was 0.11 ppm. When using the HY catalyst, the chlorine content in the liquid product was much lower (320 ppm) than that in the case of thermal cracking, although the reaction time of catalytic cracking was longer than that of thermal cracking. As expected, the presence of iron oxide on the surface of HY resulted in lower chlorine content in the liquid and gas products, likely due to the chlorine adsorption ability of iron oxides. The impregnation of 20 wt.% Fe yielded liquid product with the lowest chlorine content (60 ppm) among the tested catalysts. An SEM-EDX analysis was performed to observe chlorine captured by catalysts and the results are presented in Fig. 6(b). Unlike the EDX result of HY zeolite, chlorine was detected on the Fe impregnated catalysts: 1.2 wt.% for Fe[3]/HY and 3.5 wt.% for Fe[20]/HY. These results point out that iron oxide adsorbs chlorine in the form of iron chloride [11], although the characteristic peaks (2theta = 33 and 64°) of FeCl2 was not observed in the XRD pattern (Fig. 3(e)), likely due to the small amount of Cl contents on the surface of the catalysts. Lingaiah et al. [11] suggested that FeCl2 also has activity to remove chlorine compounds like iron oxide, based on their results. Therefore, it is found that Fe impregnated HY catalysts are effective to produce cracked liquid products with very low chlorine content (< 100 ppm) from chlorinated heavy wax. Fig. 7 presents the changes of liquid yield and carbon distribution in the liquid products during catalytic cracking of the chlorinated heavy wax on Fe[5]/HY catalyst at 450 ℃ and WHSV = 1.2 h−1. The liquid yield was very low (37.1 wt.%) and gasoline fraction was relatively high (33.3 wt.%) for the first hour after reaching a steady-state of the reaction system, due to relatively low WHSV and active acid sites of Fe[5]/HY catalysts. While the liquid yield reached to about 60 wt.% with TOS, gasoline fraction was gradually reduced to 22.1 wt.%, but heavy fraction increased from 6.8 wt.% to 16.5 wt.%, indicating a gradual loss of the acid sites. In case of chlorine contents, 0.01 wt.% was detected in liquid samples at TOS = 2 h and 4 h, but chlorine did not detected in other samples. This means that catalytic cracking reaction related to acid sites is more sensitive to operation time than dechlorination over iron oxides under this reaction conditions.SEM-EDX analyses of the spent catalysts revealed coke deposits as well as chlorine adsorption on the surface of the catalysts, as shown in Fig. 6(b). The regeneration of the spent catalyst is one of the important factors to promote catalyst usability. Fig. 8 presents the TGA/DTG thermal scan of the HY zeolite and Fe[20]/HY zeolite. The TGA curves show total mass loss of about 5.34 wt.% for HY and 6.78 wt.% for Fe[20]HY with two mass-loss peaks at around 72 ℃ (removal of water) and around 470 ℃ (removal of soft coke) [16]. Based on the TGA/DTG, the regeneration condition of the spent catalysts was set as 700 ℃ and 4 h in air. Fig. 9 shows SEM-EDX data of the regenerated Fe[20]/HY catalyst: chlorine was completely removed and coke was almost fully removed. This demonstrates that thermal treatment in air is effective to remove adsorbed chlorine and deposited coke on the catalyst. When using the regenerated Fe[20]/HY for the catalytic cracking of chlorinated heavy wax, a similar trend to the case of the fresh Fe[20]/HY was found: 86.1 wt.% for liquid yield, 9.6 wt% for gas yield and 4.3 wt% for solid yield, and 16.7 wt.% for gasoline, 53.0 wt.% for kerosene/diesel, and 30.3 wt.% for the heavy fraction in the liquid product.Catalytic cracking and dechlorination of chlorinated heavy wax obtained from pyrolysis of RPF were investigated using an iron oxide impregnated HY zeolite. The heavy wax as a raw material had a very wide carbon distribution (1.54 wt.% of gasoline fraction (C5-C10), 37.91 wt.% of kerosene/diesel fraction (C11-C22), and 60.55 wt.% of heavy wax fraction (≥C23)) and contained 0.14 wt% of chlorine. It was found that the largest worthy liquid fraction (gasoline and kerosene/diesel, 66.9 wt%) was achieved when using Fe[3]/HY among the impregnated HY catalysts, which is a similar result to the parent HY zeolite (67.4 wt.%). However, the presence of iron oxide on the surface of HY resulted in lower chlorine content in the liquid product than the parent HY catalyst, due to chlorine adsorption of iron oxides with a form of iron chloride. This indicated that the Fe impregnated HY catalyst had a dual function of catalytic cracking of the HY zeolite and dechlorination of iron oxide. Excessive impregnation of Fe, Fe[20]/HY, showed the lowest cracking activity of heavy wax owing to the catalyst having lowest total acid sites. Significant amount of soft coke and chlorine were deposited on the spent catalysts, and they were restored by thermal treatment in air (700 ℃). The restored catalytic activity was confirmed by using the regenerated Fe/HY catalyst for the cracking of chlorinated heavy wax. Kyung-Ran Hwang: Conceptualization, Methodology, Experiment, Analysis, Investigation, Visualization, Writing - original draft, and Writing - review & editing. Sun-A Choi: Experiment, Formal analysis. Il-Ho Choi Experiment, Formal Analysis. Kyong-Hwan Lee: Conceptualization, Supervision, Project Administration and Writing - review & editing.Data are openly available in a repository that issues datasets with DOIs.The authors declare no conflict of interest.This work was conducted under the framework of the research and development program of the Korea Institute of Energy Research (C0-2427-03).

Catalytic conversion of useless chlorinated heavy wax (chlorine, 0.14 wt%) obtained from pyrolysis of refuse plastic fuel was studied using iron oxide impregnated HY zeolite to produce a useful liquid product. It was found that the largest liquid fraction (gasoline and kerosene/ diesel, 66.9 wt.%) with very low chlorine content was achieved when using Fe[3]/HY among impregnated HY catalysts. This demonstrated that the Fe impregnated HY catalyst had a dual function of catalytic cracking of HY zeolite and dechlorination of iron oxide. Excessive impregnation of Fe, i.e., Fe[20]/HY, showed the least cracking activity of heavy wax owing to the catalyst having lowest total acid sites, but yielded liquid product with the lowest chlorine content (60 ppm) among the tested catalysts. The spent catalysts were deposited by a significant amount of soft coke and chlorine, and they were totally restored by thermal treatment in air (700 ℃).

In a world governed by a constant increase in energy demand, there is a need for new, sustainable, and ecologically acceptable power sources. Given the rate of fossil fuel consumption, researchers predict that they will be almost completely depleted in approximately fifty to a hundred years [1]. One of the proposed alternatives to the current fossil fuel-based economy is the so-called hydrogen economy [2,3]. Briefly, this approach takes energy from a non-carbon source, which can be either renewable sources like solar and wind energy or nuclear power plants, converts it, and stores it in the form of hydrogen gas. This gas can then be transported and converted back into electric energy when needed.In the heart of the hydrogen economy is the unitized regenerative fuel cell (URFC). This device is capable of running in two modes. The first one, when there is an inflow of electric energy, consists of running in electrolysis mode and converting water to hydrogen and oxygen, which can then be stored in pressurized containers. The second mode is the fuel cell mode, in which the device runs as a standard fuel cell, consuming hydrogen and oxygen and producing electricity and water. Theoretically, such a device can run indefinitely long, producing zero pollutants since the only products (depending on running mode) are hydrogen, oxygen, and water.The main problem with operating such a device, which is common to most fuel cells, is the sluggish kinetics of the oxygen reduction reaction (ORR) [4]. In the case of URFCs, the criteria for ORR catalyst selection are even more demanding because a catalyst must sustain transition from one operating mode to another and be stable for catalyzing both ORR and oxygen evolution reaction (OER) for an extended period. Although some catalysts are traditionally considered the best ones for either ORR or OER, there are some difficulties in their application in URFCs technology. For instance, Pt is regarded as the best catalyst for ORR. However, the thermodynamic potential of Pt versus standard hydrogen electrode is 1.23 V, and at such a high potential, the metal surface undergoes oxidation. This practically means that the surface of Pt catalysts consists of a Pt and PtO mix, which effectively lowers the open circuit potential, depending on Pt to PtO ratio. On OER potential large portion of Pt surface is covered with PtO, which inhibits its activity toward OER catalysis.On the other hand, ruthenium (Ru) and iridium (Ir) are considered the most efficient OER catalysts. However, these catalysts’ low natural abundance and high cost rule them out of commercial applications. These and some other issues have led researchers to synthesize and characterize new catalysts that can catalyze both ORR and OER and endure harsh conditions in URFCs.Recently, a number of different catalysts based on carbon-supported transition metal nanocomposites and metal-oxide-supported nanocomposites have been reported as substitutes to noble-metal electrocatalysts [5,6]. Carbon-supported materials are still highly investigated catalysts due to the high porosity of carbon support. However, the durability of carbon under OER polarization conditions remains a challenge due to high potentials and a highly oxidative environment, resulting in a change in the material’s morphology, composition, and structure and irreversibly affecting its activity toward OER [7]. Nam et al. [8] reported ternary Ni46Co40Fe14 dispersed in carbon (C@NCF-900) as a highly efficient bifunctional electrocatalyst with half-wave potential, E1/2, of 0.93 V vs. RHE and low potential to achieve a current density of 10 mA cm-2, E10, in OER operation mode (1.66 V vs. RHE). Fu et al. [9] reported N-doped carbon nanofibers decorated with NiCo as a bifunctional electrocatalyst superior to commercial Pt/C in terms of ORR catalysis and RuO2 in terms of OER activity. Hollow-structured carbon-supported NiCo2O4/C was further reported by Wang et al. [10] as an electrocatalyst of superior durability under both ORR and OER conditions compared to the commercial Pt/C in alkaline solution with a potential of 1.67 V required to achieve 10 mA cm-2 in OER operation mode and a potential difference between E1/2 and E10, ΔE, of 0.96 V, lower than that of commercial Pt/C (1.06 V).In this work, four different catalysts to be used in URFC technology were prepared. Catalysts were synthesized via simultaneous supercritical carbon dioxide deposition method, and their electrocatalytic ability was assessed. Introducing another metal along with Pt can influence the electronic structure of Pt by changing interatomic distance in Pt-Pt bonds, resulting in a catalyst of superior activity and stability for the two investigated reactions. Moreover, lowering the Pt content in the catalyst also lowers its price. Graphene nanoplatelets (GNPs) were used as support to increase the surface area and improve catalyst/support durability. Grafting metal nanoparticles onto GNPs provides a structure with uniformly distributed, abundant active sites, decreasing the electrical contact resistance between neighboring metal nanoparticles, thus delivering good electrical contact and fast transport of reactants and their plentiful incidence with active catalytic sites. The unique structure of graphene may improve electrode durability by strengthening the interaction between the catalyst particles and the graphene support. Additionally, the degree of graphitization also plays a significant role in the oxidation resistance of carbon-based materials. The higher graphitic content has also been linked to a stronger interaction between metal and carbon support. An increase in the degree of graphitization results in stronger π-sites (sp2 hybridized carbon) on the support (which are the anchoring sites for the electrocatalyst), strengthening the metal-support interaction.Nickel (II) hexafluoroacetylacetonate hydrate (C10H2F12NiO4.xH2O, MW: 472.79, 98% purity), copper (II) hexafluoroacetylacetonate hydrate (C10H2CuF12O4.xH2O, MW: 477.65) and iron (III) tris (2,2,6,6-tetramethyl-3,5-heptanedionate) (C33H57FeO6, MW: 605.65) supplied by Sigma-Aldrich, were used as metal precursors. 1,5-dimethyl platinum cyclooctadiene (Me2PtCOD) supplied by Strem Chemicals, was used as Pt precursor. GNPs (759 m2 g-1 area) used as support material was purchased from XG Science (xGnP® Grade C). The CO2 and N2 gases were purchased from Habaş.GNPs-supported Pt-M (M= Ni, Fe, Cu) catalysts were synthesized simultaneously through supercritical carbon dioxide (scCO2) deposition technique, Fig. 1.Briefly, 0.1 g of GNPs was placed in a pouch made of filter paper and placed in a custom-made high-pressure stainless-steel vessel having sapphire windows along with a stirring bar. For the desired metal loading onto GNPs support, the amounts of the corresponding precursors were determined by using the adsorption isotherm of the precursor onto the GNPs [11]. Next, a predetermined amount of organometallic Pt and other metal precursors were added, and the vessel was closed. To provide a supercritical environment, the vessel was heated to 333 K and CO2 gas was introduced into the vessel up to 24 MPa with a syringe pump. These conditions were maintained for a period of 24 h to ensure that the system reached equilibrium. At the end of the 24th hour, the vessel was depressurized, and the pouch was removed. Subsequently, the resultant material was placed in a tube furnace to convert metal precursors to their metal forms. The conversion was carried out thermally at 400 °C under N2 flow for 4 h. Consequently, GNPs supported Pt-M catalysts were obtained [11,12].The metal loading over GNPs was analyzed by inductively coupled plasma mass spectrometry (ICP-MS) technique using an Agilent 7800 ICP mass spectrometer. The morphology and elemental distribution of the catalysts were characterized with transmission electron microscopy (TEM, Hitachi HighTech HT7700) and energy-dispersive X-ray spectroscopy (EDX). Catalyst compositions and surface oxidation states were characterized by X-ray photoelectron spectroscopy (XPS) using a Thermo Scientific Kα X-ray photoelectron spectrometer.Electrochemical investigation of the synthesized catalysts was performed using Gamry Interface 1010 galvanostat/potentiostat in a standard three-electrode electrochemical cell connected to Gamry rotator (Gamry RDE710 rotating electrode) and with Pt mesh and saturated calomel electrode as counter and reference electrode, respectively. For easier comparison of results, all potentials are expressed versus the reversible hydrogen electrode (RHE). Catalytic inks for working electrodes preparation were made using Nafion as a binder [13]. High purity gases (O2 or N2, Messer, 99.9995 vol%) were bubbled into the 0.1 M KOH electrolyte/electrochemical cell to control the atmosphere.ORR and OER stability tests were performed in the chronoamperometry mode. OER tests were performed in Arbin Instruments equipment by applying a 30-second OER pulse (potential of 1.7 V) followed by a 120-second pulse at the ORR potential. In this way, O2 formed during OER pulse was reduced during the ORR mode, thus preventing material peeling from the substrate due to the intense bubble formation and, at the same time, switching between the two modes of URFCs was simulated. Fig. 2 shows the metal concentrations of the prepared catalysts determined by ICP-MS. It can be seen that the second metal loading is much lower compared to the Pt loading in all catalysts. This may be explained by the solubility of metallic precursors in the supercritical medium. The Pt loading in the plain Pt/GNPs catalyst was 15.8 wt%. The addition of the second metal resulted in an increase in the total metal loading, as well as in the Pt loading, either in sequential or simultaneous supercritical carbon dioxide deposition of the metals [11,12]. Additionally, other factors such as the relative size of the metal ion and the ligand, the number of ligands attached to the metallic-ion, CO2-philic tendency, and the polar nature of the precursor in scCO2 can influence the relative compositions of the metallic phase yield in PtM/GNPs [14].To investigate the distribution of the different elements in the catalysts, TEM, EDX, and elemental mapping measurements were performed, and the results were presented in Fig. 3. The TEM images indicate that the uniform and well-dispersed metal nanoparticles with a size range of 2–3 nm are tightly anchored onto the GNPs support. The elemental mapping results reveal that the metal distribution on the GNPs is consistent with the loading amount of the catalysts. As expected, the corresponding EDX spectra show the C, Pt, Ni, Fe, and Cu signals from the respective catalyst. The observed dominant peak at around 8 keV for PtNi/GNPs and PtFe/GNPs catalysts can be attributed to the Cu TEM grid.To examine the surface nature of the synthesized catalysts, XPS measurements were performed. Fig. 4a shows the XPS survey spectra of the catalysts. The observed peaks were found to be compatible with the prepared catalyst structure and literature reports [15–17]. Additionally, the presence of these peaks in the structure suggests that the GNPs supported Pt-M catalysts were successfully prepared with scCO2 deposition technique. Fig. 4b shows the Pt 4 f XPS spectra of the prepared catalysts. In Pt/GNPs catalyst, the spectra comprised two peaks allocated to the Pt 4 f7/2 and Pt 4 f5/2 of metallic Pt, respectively. This result illustrates that the organometallic precursor was efficiently reduced to its metallic form. In comparison with Pt/GNPs catalyst, the Pt 4 f core level peaks of the PtNi/GNPs and PtCu/GNPs bimetallic catalysts slightly shifted to lower binding energies, remaining the same for the PtFe/GNPs catalyst, which has the lowest second metal content. The observed shifting could be attributed to the electron interaction between Pt and secondary metal atoms (Ni, Cu). Changes in the d-band center are believed to accompany similar variations in the surface of core-level shifts in the same direction [18]. Thus, the negative shift in the binding energy of the Pt 4 f core level reflects the downshift of its d-band center relative to the Fermi level. The binding energy strongly influences the adsorption/desorption capability of reaction species on the catalyst surface. According to the Hammer and Norskov model [19], it can be said that the downshifted d-band center in the catalysts would weaken the chemical adsorption strength of oxygenated intermediates such as OHads in alkaline solution. This would cause Pt-OHads bonds to break easily, leading to more electrochemically active sites available for the oxygen reduction reaction (ORR). On the other hand, the downshift of the d-band center can also weaken the chemical adsorption strength of active oxygen, and block O-O band breaking, which deteriorates the ORR kinetics. Therefore, to maintain the balance between the two opposite effects, there should be an optimum Pt 4 f binding energy corresponding to the best catalytic activity.In addition, high-resolution XPS scans of the catalysts were also carried out, as shown in Fig. 5. In the high-resolution Ni 2p spectrum of PtNi/GNPs, displayed in Fig. 5a, the main peaks are observed at binding energies of 855.6 eV for Ni 2p3/2 and 873.5 eV for Ni 2p1/2; these are the characteristics of PtNi and two shake-up satellites [20]. The spectrum shows that Ni is in an oxidized state, corresponding to NiO and Ni(OH)2, and NiOOH, respectively. The major component is Ni(OH)2. These oxygen species (e.g., NiO, Ni(OH)2, NiOOH) present in the Pt-Ni catalysts represent surface and subsurface oxidative states and not crystalline oxidative states as no such oxide peaks are apparent in the XRD patterns of the catalysts [12]. Namely, XRD analysis of the samples synthesized following the same procedure has been previously performed [12]. XRD pattern of Pt/GNPs catalyst revealed distinct diffraction peaks corresponding to reflections from (111), (200), (220) and (311) planes of Pt in the face-centered cubic (fcc) crystal structure as well as a broad peak (at 2θ = 26°) corresponding to the (002) plane of the graphite-like crystalline structure of the GNPs. Diffraction peaks of the PtM/GNPs shifted to higher 2θ values relative to Pt/GNPs, evidencing the successful synthesis of PtM (M = Ni, Fe, Cu)/GNPs. Moreover, no other diffraction peaks were observed for any of the catalysts, confirming the formation of alloys of Pt and secondary metals.Furthermore, the obtained XPS results are comparable with the results of PtNi alloy-graphene catalyst, as discussed previously by Li et al. [21]. In the high-resolution Fe 2p spectrum of the PtFe/GNPs, the peaks centered at around 711 and 725 eV can be assigned to the Fe 2p3/2 and Fe 2p1/2, respectively [22]. The Fe 2p region is deconvoluted to resolve each component. This confirmed the existence of zero-valent Fe0, oxidized Fe2+, and Fe3+ with a broad satellite peak. Similarly, in the case of PtCu/GNPs, two kinds of Cu species are observed in a close view of the Cu 2p spectrum, at ca. 932.1 and 952.2 eV, attributed to Cu 2p3/2 and Cu 2p1/2, respectively [23]. After deconvolution, the spectrum shows that although most Cu is in the form of metallic Cu (Cu0, 932.1 eV), a signal from Cu2+ (934.3 and 954.3 eV) also exists. The presence of Cu2+, further confirmed by a satellite peak at 943.7 and 940.1 eV, can be ascribed to the easy oxidation of surface Cu atoms in the air [24].To investigate double-layer capacitance (Cdl), cyclic voltammograms (CVs) were recorded in the 200 mV range around the open circuit potential (OCP), Fig. 6. Capacitance was then determined as slope of Δj = f(ν) plot, where Δj = ja - jc (mA cm-2) and ν is electrode polarization rate (mV s-1), Fig. 6 inset. Pt/GNPs sample showed the highest value of Cdl (8.2 mF cm-2) among the studied samples. PtFe/GNPs showed the highest Cdl value (7.2 mF cm-2) among the bimetallic Pt-M samples, comparable to that of Pt/GNPs. Samples containing PtNi and PtCu showed somewhat lower values of double-layer capacitance, i.e., 4.5 and 5.1 mF cm-2, respectively. Slight distortion of CVs with the increase in the polarization rate is observed for PtNi/GNPs and PtCu/GNPs, indicating the existence of ohmic resistance in parallel with Cdl [25]. It is worth mentioning that all catalysts tested in this work showed a higher Cdl value than commercial Pt/C (40 wt% Pt) catalyst (3.1 mF cm-2) tested under the same condition [13], with these higher values of Cdl indicating larger active surface of the synthesized catalysts than that of commercial Pt/C catalyst.Activity toward ORR was investigated by performing a series of linear scan voltammetry (LSV) experiments in the ORR potential region using different rotation rates, Fig. 7. It can be seen that all tested catalysts show relatively high diffusion-limited current densities (j d ), which are comparable to that of commercial Pt/C (40 wt% Pt) catalyst (−6.43 mA cm-2 at 1800 rpm) [13]. jd of − 4.65 mA cm-2 and − 4.37 mA cm-2 was reached during ORR at PtFe/GNPs and PtCu/GNPs, respectively, while a slightly lower value was observed for PtNi/GNPs sample (−3.65 mA cm-2). jd value of − 3.71 mA cm-2 was reached during ORR at Pt/GNPs reference sample.Tafel slope, representing the reaction’s sensitivity toward applied potential, was calculated from LSV plots at 1800 rpm, Fig. 8a and Table 1 . Tafel analysis was performed in the current range from the onset potential, Eonset, to ca. 70–80% of the limiting current density. Pt/GNPs and PtFe/GNPs showed two values of Tafel slope each: 65 mV dec-1 and 102 mV dec-1 for Pt/GNPs vs. 81 mV dec-1 and 66 mV dec-1 for PtFe/GNPs, depending on the potential range. Dual values of the Tafel slope indicate a change in the ORR mechanism at these catalysts with a change in potential. The other two samples showed only one value of Tafel slope: 102 mV dec-1 in the case of PtNi/GNPs and 67 mV dec-1 in the case of PtCu/GNPs.Half-wave potential, E1/2, was also determined from the LSV studies. Combining it with the potential at which OER reaction achieves a current density of 10 mA cm-2, we were able to calculate the difference between these two potentials, ΔE, a key parameter in benchmarking catalysts for URFCs application (see below) [26]. As expected, the catalyst containing just Pt showed the highest value of E1/2 (0.98 V), followed by PtNi and PtFe catalysts (0.94 V and 0.91 V, respectively). The lowest value of E1/2 was observed for PtCu/GNPs catalyst (0.90 V). These values are comparable/somewhat higher than that reported in the literature (Table 1), indicating somewhat higher ORR activity of the herein studied catalysts.Koutecky-Levich analysis was used for assessing the number of exchanged electrons during ORR. Constructed Koutecky-Levich plots (Fig. 8b) represented straight lines that indicate a first-order reaction. The highest value of n was calculated for PtNi/GNPs sample (n = 3.93), indicating that four electrons are involved during the O2 reduction at this catalyst. Values calculated for the rest of the catalysts are between 3.61 and 3.77 electrons, Table 1. All calculated n values are comparable with that for ORR at commercial Pt/C (40 wt% Pt) catalyst (3.97) [13] and with the n values found in the literature for ORR, Table 1.Although there is still a big debate about the precise mechanism of ORR, it is widely accepted that in alkaline media it can occur by either four- or two-electron mechanism. The first one concerns bidentate molecule adsorption (adsorption of two O atoms) or a direct four-electron pathway to generate OH- (Eq. 1) [7]: (1) O 2 + 2 H 2 O + 4 e − → 4 O H − (2) O 2 + H 2 O + 2 e − → H O 2 − + O H − (3) H O 2 − + H 2 O + 2 e − → 3 O H − (4) 2 H O 2 − → 2 O H − + O 2 Eqs. (2–4) correspond to the so-called two-electron pathway in which O2 is first converted to HO2 - and further generates OH- in alkaline electrolytes. However, the reactivity of each of these steps is strongly dependent on the O2 adsorption energy, dissociation energy of O-O bond, and on the binding of OH- to surface.Thus, the superior activity of PtFe/GNPs toward ORR may result from its lower Gibbs free energy of oxygen adsorption determined by the geometric and electronic effect [35]. Fe is believed to represent active adsorption sites that “capture” oxygen and increase adsorption capacity towards oxygen, so the presence of copious Fe sites on the catalyst’s surface can reduce the oxygen adsorption Gibbs energy in the first step of the ORR. The interaction between Fe and Pt can further impact ORR activity by affecting the catalyst’s structure, including the Pt-Pt bond distance [35]. Namely, the addition of Fe causes the lattice contraction decreasing the interatomic distance that influences the adsorption and transfer of oxygen-containing species in ORR. Lattice distortion further affects the overlap of orbital, modifying the electronic properties on the active site and thus affecting the surface reactivity and the catalyst's performance.Stability tests for all catalysts were performed in chronoamperometry mode with constant O2 bubbling for 4 h, Fig. 9. Relatively good stability in the ORR operation mode was observed for all tested catalysts with no significant drop of current densities with time. Steady current densities indicate a stable performance of the synthesized catalysts in long-term exploitation in ORR operating mode.Similar to ORR, OER is also a multi-electron transfer process, which in alkaline media could be represented by Eqs. 6 – 10, (6) M + O H − → MOH + e − (7) MOH + O H − → MO + H 2 O ( l ) + e − (8) 2 MO → 2 M + O 2 ( g ) (9) MO + O H − → MOOH + e − (10) MOOH + O H − → M + O 2 g + H 2 O ( l ) + e − where M is the catalyst’s active site, and MOH, MO, and MOOH are the reaction intermediates. The reactivity of each step is again highly dependent on the adsorption energy of oxygen species. It is often reported that the formation of MO from MOH or the evolution of MOOH from MO are the rate-determining steps (RDS) for OER [7].OER kinetics at four studied catalysts was investigated using LSV experiments at 1200 rpm, Fig. 10a. Key parameters as exchange current density, j 0 , overpotential for reaching benchmark current density of 10 mA cm-2, η 10 , the current density at an overpotential of 400 mV, j 400 , and Tafel slope were determined to further characterize the investigated catalysts.One can see that the best performing catalyst in terms of the reached current densities at a given potential is PtFe/GNPs, followed by PtNi/GNPs and PtCu/GNPs catalysts. The lowest current density was achieved with Pt/GNPs catalyst, concluding that Fe, Ni, and Cu are indeed active sites for OER. Oxide film forming under OER polarization conditions most likely occurs on Pt as indicated by Damjanovic et al. [36], leading to the degradation of Pt-based catalysts performance and a low value of reached current density. Namely, contrary to the ORR, the OER occurs on the oxidized surface being constrained either by the strong adsorption of OOHads or by the weak adsorption of Oads [37]. The optimal performance is achieved for metal oxides with a moderate binding energy of the OER intermediates. This corresponds to the top of the volcano plot constructed by plotting the overpotential to reach 1 mA cm-2 as a function of the difference in the free binding energy between O* and OH. One can observe that, for example, NiO is located closer to the optimal value when compared to PtO2.Another important difference between ORR and OER is that during the ORR only the outer surface atoms are active once the reaction comes to be diffusion-controlled with increasing overpotential [37]. Opposite, reaction sites in the “inner” surface turn to be active with increasing OER overpotential, i.e., the reaction proceeds within the inner catalysts layers as well. Therefore, the porosity of a catalyst plays a crucial role in its performance during OER. N2-sorption analysis carried out in our previous study [12], revealed the highest electrochemical active surface area (ECSA) of 136 m2 gPt -1, in case of PtFe/GNPs, followed by PtNi/GNPs (132 m2 gPt -1) and PtCu/GNPs (122 m2 gPt -1). Notably, lower ECSA ranging between 43.6 and −87.2 m2 gPt -1 were determined for Pt/GNPs catalysts for different metal loading.The present study aimed to develop bifunctional oxygen catalysts that are vital for the operation of unitized regenerative fuel cells and rechargeable metal-air batteries. These are often obtained by combining highly active ORR catalyst (typically Pt) with highly active OER catalyst (typically RuO2 and IrO2). Within the present study, Pt was combined with low-cost transition metals instead of expensive RuO2 or IrO2.For further comparison of the catalysts’ OER performance, Tafel slope values were evaluated and found to be relatively high, varying from 280 mV dec-1 for PtFe/GNPs to 490 mV dec-1 for PtCu/GNPs (Fig. 10b), suggesting impediment of electron transfer within the studied catalysts during OER.In terms of other calculated OER parameters ( Table 2), results are in good agreement with the initial observations based on OER LSV experiments. Namely, the PtFe/GNPs catalyst, which achieved the highest current density in the investigated potential region, also showed the lowest overpotential to reach a current density of 10 mA cm-2 (0.572 V), followed by PtNi/GNPs (0.652 V). Higher values of this parameter were calculated for Pt/GNPs and PtCu/GNPs with η10 of 0.765 and 0.660 V, respectively. The highest current at overpotential of 400 mV was achieved with PtCu/GNPs (3.27 mA cm-2), followed by PtFe/GNPs (2.89 mA cm-2) and PtNi/GNPs (2.64 mA cm-2) catalysts. As anticipated from the initial LSV results, and due to the mentioned oxide film forming, the lowest value of j400 was achieved with the control Pt/GNPs catalyst, which only has Pt as an active site for OER.After determining the potential at the current density of 10 mA cm-2, we were able to calculate ΔE for each catalyst. It was found that this parameter has the lowest value of 0.890 V for PtFe/GNPs catalyst. Although all of the studied catalysts show mildly higher overpotential to reach benchmark current density of 10 mA cm-2 (Table 2), ΔE values are comparable or lower than some materials reported in the literature [10,27–30,33,34], indicating the promising performance of studied catalysts in URFC application. Fig. 10 c and Table 2 present the results of electrochemical impedance spectroscopy (EIS) experiments for the studied catalysts. A small difference in the electrolyte resistance, Rs was attributed to the small changes in cell geometry and distance between electrodes. Furthermore, PtFe/GNPs showed the lowest value of charge transfer resistance, Rct, of 53 Ω (Fig. 10c inset), followed by PtNi/GNPs catalyst with significantly higher Rct of approximately 450 Ω. For Pt/GNPs and PtCu/GNPs catalysts, Rct was found to be even higher (ca. 480 Ω and 550 Ω, respectively).OER stability test for all catalysts was performed for 10 h by switching between ORR and OER polarization conditions. For clarity, only OER curves are shown in Fig. 10 d. A significant initial drop in OER currents was observed for all studied catalysts within the first hour of the experiment. In the next nine hours, currents continued to drop, though slower. This suggests the relatively low stability of the studied catalysts for OER. It is worth mentioning that it is not necessary degradation of catalyst that led to the current drop in the stability test. The partial loss of catalytic ink from the electrode due to intense bubble formation in OER mode is one of the possible causes, suggesting further investigation to optimize ink preparation and composition to withstand the harsh conditions when switching between URFC’s two operation modes.Pt and PtM (M = Ni, Fe, Cu) supported on GNPs were synthesized by simultaneous scCO2 deposition method. XPS confirmed the successful preparation of the catalysts via scCO2 deposition method by showing the presence of the basic elements (C 1 s, O 1 s, Pt 4 f, and Pt 4d) in all samples with the signals of Ni 2p, Fe 2p, and Cu 2p in the PtNi/GNPs, PtFe/GNPs, and PtCu/GNPs catalysts, respectively. TEM analysis revealed the formation of metal nanoparticles of 2–3 nm size uniformly distributed over GNPs. The catalysts were tested for ORR and OER as electrode reactions in URFCs. PtFe/GNPs showed promising performance for ORR regarding the high diffusion-limited current density, low Tafel slope, number of exchanged electrons close to 4, and high double-layer capacitance. The same material showed the best performance toward OER, evidenced by the highest current density and the lowest Tafel slope. Furthermore, this material performance was comparable to that of commercial Pt/C electrocatalyst containing double the amount of Pt. Dušan Mladenović: Investigation, Formal analysis, Visualisation, Writing – original draft. Elif Daş: Investigation, Formal analysis, Writing – original draft. Diogo M.F. Santos: Visualisation, Writing – review & editing. Ayşe Bayrakçeken Yurtcan: Conceptualisation, Investigation, Writing – original draft. Šćepan Miljanić: Conceptualisation, Supervision. Biljana Šljukić: Conceptualisation, 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.The authors would like to thank the Ministry of Education, Science and Technological Development of Republic of Serbia (contract number: 451-03-68/2022-14/200146). The Portuguese Foundation for Science and Technology (FCT) is acknowledged for contract IST-ID/156/2018 (B. Šljukić) and a research contract in the scope of programmatic funding UIDP/04540/2020 (D.M.F. Santos).

Pt and Pt-M (M = Ni, Fe, Cu) nanoparticles supported on graphene nanoplatelets (GNPs) were synthesized by simultaneous supercritical carbon dioxide deposition method. Morphology analysis by TEM revealed the formation of metal nanoparticles of 2–3 nm size uniformly distributed over GNPs, while XPS was used to determine their oxidation states. Four materials were tested as electrocatalysts for ORR and OER in unitized regenerative fuel cells and rechargeable metal-air batteries. PtFe/GNPs exhibited favorable ORR kinetics in terms of the highest diffusion-limited current density, the lowest Tafel slope, and a high number of exchanged electrons (n = 3.66), which might be attributed to its high double-layer capacitance and, thus, high electrochemically active surface area. Furthermore, this material performance was comparable to that of commercial Pt/C electrocatalyst containing double the amount of Pt. The same material showed the best performance toward OER as evidenced by the highest current density, the lowest value of exchange current density, and overpotential to reach a current density of 10 mA cm-2, as well as the lowest Tafel slope.

Developing new routes for the effective utilization of non-petroleum carbon resources (such as coal, biomass, natural gas, and waste) to produce clean fuels and value-added chemicals is of great interest due to the shortage of petroleum resources and the environmental problems [1]. Aromatics, acting as significant bulk chemicals, are widely applied for the production of solvents, medicines, dyes and polymers in the chemical industry [2]. Currently, aromatics are mainly obtained by the petroleum refining process, which is environmentally unfriendly and energy-intensive [3,4]. Catalytic conversion of syngas (a mixture of CO/H2) derived from aforementioned non-petroleum resources into aromatics has attracted extensive attention because of its prominent function in sustainable development [5]. By means of the well-known Fischer–Tropsch synthesis (FTS), various hydrocarbons can be produced from syngas, but almost no aromatics can be obtained due to the limitation of Anderson-Schulz-Flory distribution (ASF) [6,7]. The composite catalysts which integrating Fe-based FTS catalyst with H-ZSM-5 zeolite (such as Fe–Pd/H-ZSM-5 [8], Fe–MnO/GaZSM-5 [9], Na–Zn–Fe5C2/H-ZSM-5 [10], and Fe3O4@MnO2/H-ZSM-5 [11]) is a useful means to optimize the aromatization performance. However, their aromatics selectivity is still unsatisfactory.Recently, an effective oxide-zeolite (OX−ZEO) bifunctional composite catalysts design strategy has been put forward to directly synthesize aromatics from syngas (STA reaction), among which CO and H2 are activated into C1 oxygenated intermediates over metal oxides, while the zeolite is mainly responsible for C–C coupling [12–19]. For example, Wang et al. [12]. Reported a composite catalyst combining Zn–ZrO2 oxide with H-ZSM-5 zeolite could obtain 80% aromatics selectivity at 20% CO conversion. Bao et al. [13]. Presented that a composite catalyst integrating ZnCrOx oxide with H-ZSM-5 zeolite could achieve about 73.9% aromatics selectivity at 16% CO conversion. Nevertheless, the catalytic performance still needs to be further improved to meet the demands of industrial production, which will depend on the in-depth comprehending of the structure–performance relationship.Previous studies have pointed out that the structure (such as crystal size) of the H-ZSM-5 zeolite in composite catalysts has a great influence on the catalytic performance for STA reaction [19,20]. Wei et al. [19] reported that the H-ZSM-5 with short size along the b-axis presented low molecular-diffusion resistance, leading to high selectivity of tetramethylbenzene. Xie et al. [20] observed that the light aromatics selectivity is closely related to the crystal size of H-ZSM-5, which could be enhanced by increasing the size of the b-axis. In comparison, there is little knowledge about the influence of the oxide structure in composite catalysts for STA reaction. Generally, the oxide with a smaller size can result in more defect sites (oxygen vacancies) [17,21], and it has been reported that oxygen vacancies over metal oxide surfaces are conducive to CO hydrogenation activation [1,16,22]. Therefore, we wondered whether oxide with a smaller size will have a beneficial effect on the catalytic performance. However, little in-depth research about the size effect of oxide component in composite catalysts for STA reaction has been reported.In this work, we investigate the size effect by selecting ZnCr2O4 spinel oxide, which is often used for syngas conversion, as a probe oxide, mixing with H-ZSM-5 zeolite as a composite catalyst for STA reaction. To achieve this goal, a series of ZnCr2O4 spinel oxides with distinct size were synthesized, and we compared the catalytic performance of this series of ZnCr2O4&H-ZSM-5 composite catalysts as a function of size of ZnCr2O4. We further reveal the intrinsic reason of the size effect by in situ DRIFTS characterizations. Moreover, by decreasing the crystal size of ZnCr2O4 oxide, the space-time yield (STY) of aromatics can reach as high as 4.79 mmol gcat −1 h−1, which outperforms the previously reported some typical composite catalysts (Fig. S1 and Table S1).A series of ZnCr2O4 spinel oxides with diverse sizes were synthesized by employing a conventional co-precipitation method. Typically, 48 g of Cr(NO3)3·9H2O and 17.85 g of Zn(NO3)2·6H2O were dissolved in 100 mL of deionized water to make a salt solution, and 47.10 g of ammonium carbonate (NH4)2CO3 was added in 100 mL of deionized water to prepare a precipitant solution. After that, two peristaltic pumps were used to simultaneously inject the above two solutions into a beaker at 70 °C under persistent stirring to form a precipitate. Meanwhile, the pH value of the blended solution was maintained at around 7.0 by adjusting the flow rate of the two peristaltic pumps. After precipitation, the synthesized mixture was further aged for 5 h at the same temperature. Then the suspension was filtered and washed with deionized water several times to obtain a filter cake, which followed by drying overnight at 120 °C and finally calcined in air with a heating rate of 2 °C min−1 for 6 h. The resulting powder oxides with different sizes were denoted as ZnCr2O4-T, where T representing the calcination temperature, which varied from 400 to 700 °C. For example, ZnCr2O4-400, ZnCr2O4-500, ZnCr2O4-600 and ZnCr2O4-700 represent samples calcined at 400, 500, 600 and 700 °C, respectively.H-ZSM-5 zeolite with a Si/Al ratio of 96 used in this work was the same as our previous study [17,23]. The detailed physical properties of this H-ZSM-5 zeolite are exhibited in Fig. S2 and Table S2, which embraces a typical nano-sized MFI structure. All the composite catalysts were acquired by physical mixing of the two components (oxides and zeolites) with a mass ratio of 3:1. For example, for the preparation of the ZnCr2O4-400&H-ZSM-5 composite catalyst, ZnCr2O4-400 oxide and H-ZSM-5 zeolite were firstly placed in an agate mortar and triturated together into powder, then followed by pressing under 30 MPa, crushed, and finally screened to granules between 40 and 60 mesh sizes (0.3–0.45 mm).The powder X-ray diffraction (XRD) data were obtained on a PANalytical X'Pert PRO X-ray diffractometer equipped with a Cu Kα radiation source radiation (λ = 1.5406 Å). XRD patterns were measured in the 2 theta range of 5-90°. The average crystal size of ZnCr2O4 spinel oxides was calculated according to the Scherrer equation by selecting the two representative diffraction peaks at 30.3° and 35.8°, which correspond to the (022) and (113) lattice planes of spinel cubic ZnCr2O4, respectively.The chemical composition was determined on a Philips Magix-601 X-ray fluorescence (XRF) spectrometer. The micropore volume, BET specific surface areas and average pore width were recorded by N2 adsorption–desorption at −196 °C using a Micromeritics ASAP 2020 system, the samples were pre-degassed in vacuum at 120 °C for 24 h before measurement. The morphology of catalysts was observed by utilizing a Hitachi SU8020 field-emission scanning electron microscope (FE-SEM). A JEM-2100F microscope was used to acquire the transmission electron microscopy images (TEM) and high-resolution transmission electron microscopy images (HRTEM). The average particle size of each sample was estimated by counting at least 50 particles. The X-ray photoelectron spectroscopy (XPS) was conducted on a Thermofisher Excalab X+ spectrometer with monochromatized Al Kα as the exciting radiation, the binding energy of all the data was calibrated by utilizing the C 1s of 284.8 eV as the reference. The CO-temperature programmed desorption profiles (CO-TPD) were carried out by using a Micromeritics AutoChem II 2920 analyzer equipped with a TCD, helium and 10% CO–He were used for reference and adsorption, respectively. The H2-temperature programmed reduction profiles (H2-TPR) were performed on a Micromeritics AutoChem II 2920 analyzer equipped with a TCD, argon and 10% H2–Ar were used for reference and reduction, respectively. The NH3-temperature programmed desorption profiles (NH3-TPD) were tested by a Micromeritics AutoChem II 2920 equipped with a TCD detector.The in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements were performed on a Bruker Tensor 27 instrument equipped with a MCT detector to detect the change of intensity of surface intermediate species. Typically, a diffuse reflectance infrared cell with a ZnSe window was loaded with 50 mg of the sample, which was pretreated with 30 mL min−1 of H2/N2 flow (H2/N2 = 1/5) under 0.1 MPa at 310 °C and followed by sweeping with pure N2 at a flow rate of 30 mL min−1. After that, the temperature was heated up to 390 °C and the background spectrum was recorded. Then 5 mL min−1 of syngas (H2/CO/Ar = 47.5/47.5/5) was introduced into the infrared cell under 0.1 MPa at 390 °C and the in situ DRIFT spectra were acquired at 16 scans with a resolution of 4 cm−1.All the catalytic reactions were carried out in a high-pressure fixed-bed stainless steel tubular reactor (internal diameter of 8 mm). The reaction effluent products were heated to maintain in the gas phase and analyzed by an Agilent 7890B online gas chromatograph, which equipped with a thermal conductivity detector (TCD) connected to a TDX-1 packed column and a flame ionization detector (FID) connected to a PoraPLOT Q-HT capillary column. A concentration of 5% Argon contained in syngas was utilized as an internal standard gas. CO conversion and CO2 selectivity were calculated according to the following equation based on the carbon atoms number. (1) Sel C O 2 = C O 2 o u t l e t C O i n l e t − C O o u t l e t × 100 % CO 2outlet : moles of CO2 at the outlet (2) Conv  C O = ( C O i n l e t − C O o u t l e t ) C O i n l e t × 100 % CO inlet : moles of CO at the inlet CO outlet: moles of CO at the outlet.The selectivity of hydrocarbons (CnHm), MeOH and DME among the carbon products (excluding CO2) were calculated according to the total carbon atoms of the products detected by FID detector. Sel C n H m = n C n H m t o t a l c a r b o n a t o m s o f p r o d u c t s d e t e c t e d b y F I D × 100 % Sel M e O H = n M e O H t o t a l c a r b o n a t o m s o f p r o d u c t s d e t e c t e d b y F I D × 100 % (3) Sel D M E = n D M E t o t a l c a r b o n a t o m s o f p r o d u c t s d e t e c t e d b y F I D × 100 % n CnHm: carbon atoms number of CnHm n MeOH: carbon atoms number of MeOH n DME: carbon atoms number of DME n CnHm: carbon atoms number of CnHm n MeOH: carbon atoms number of MeOH n DME: carbon atoms number of DMEA series of ZnCr2O4 oxides were prepared by a co-precipitation method at different calcination temperature [24], which were named as ZnCr2O4-T, where T representing the calcination temperature (see Experimental section). The XRD patterns (Fig. 1 a) show that all ZnCr2O4 oxides are of the typical cubic ZnCr2O4 spinel phase (PDF #98-009-5832), but as the calcination temperature decreases from 700 °C to 400 °C (from ZnCr2O4-700 to ZnCr2O4-400), the full width half maximum (FWHM) broadens from the XRD analysis (Fig. 1a), suggesting that lowering the calcination temperature may lead to a decrease in the crystal size of ZnCr2O4 spinel oxides [25]. The calculation results from Scherrer equation further verify this speculation, the average crystal size of ZnCr2O4 spinel oxides changes dramatically from 32.1 to 7.9 nm (Table 1 ). The variation of the crystal size can attribute to the different crystallite growth rate in different calcination temperature [26,27].The N2 physical adsorption–desorption experiments (Fig. 1b) present that reducing the crystal size of ZnCr2O4 spinel oxides (from ZnCr2O4-700 to ZnCr2O4-400) can result in the formation of hysteresis loops, indicating the existence of mesopores, which may be caused by the stacking of oxides particles (Fig. S3). In addition, it can be seen from Table 1 that ZnCr2O4-700 with a larger size exhibits a very low BET specific surface area (8.63 m2 g−1) and pore volumes (0.02 cm3 g−1). Surprisingly, reducing the crystal size of ZnCr2O4 spinel oxides (from ZnCr2O4-600 to ZnCr2O4-400) can greatly increase the BET specific surface areas (∼1.6–15 times) and pore volumes (∼1.6–20 times).The FE-SEM images (Fig. S3) and TEM images (Fig. 1c–f and Fig. S4) further provide convincing evidence that this series of ZnCr2O4 spinel oxides are consist of irregular nanoparticles with different particle sizes. The corresponding average particle sizes of ZnCr2O4 spinel oxides estimated by the TEM are 33.03 nm (ZnCr2O4-700), 16.63 nm (ZnCr2O4-600), 11.87 nm (ZnCr2O4-500) and 7.29 nm (ZnCr2O4-400), which is almost completely consistent with the average crystal size calculated by the Scherrer equation (Table 1). In addition, the lattice spacing of 0.208, 0.251, 0.295, and 0.480 nm in HRTEM images (Fig. S5) can be ascribed to the (004) (113) (022), and (111) planes of ZnCr2O4 spinel phase (PDF #98-009-5832), respectively. This further indicates that this series of ZnCr2O4 oxides are of the typical cubic ZnCr2O4 spinel structure.Generally, the oxide with a smaller size and larger specific surface area can result in more surface defect sites (oxygen vacancies) [17,21], which are widely considered as active sites for CO hydrogenation activation [1,16,22,28,29]. To explore the concentration of oxygen vacancy over oxides, we performed X-ray photoelectron spectroscopy (XPS) measurements. As clearly depicted in Fig. 2 a, two diverse signal peaks can be recognized from the O 1s XPS spectra of ZnCr2O4 oxides. One peak at a lower binding energy of 530.0 ± 0.3 eV can be regarded as the lattice oxygen atoms (Olattice) [30], while another peak situated at a higher binding energy of 531.0 ± 0.3 eV can usually be attributed to the oxygen atoms in the vicinity of the oxygen vacancy (Ovacancy) [29,31]. Obviously, the ratios of the Ovacancy peak for these oxides are distinct, indicating that their corresponding oxygen vacancy concentration could be quite disparate. On the basis of the calculated results from the deconvolution of the O1s XPS signal (Table 1 and Table S3) [32], the ZnCr2O4-400 exhibited the highest oxygen vacancy concentration, followed by ZnCr2O4-500, ZnCr2O4-600, and ZnCr2O4-700. It is not surprising that ZnCr2O4-400 owns more surface vacancies than the other three oxides due to its smaller size and larger specific surface area.Furthermore, we performed CO temperature-programmed desorption (CO-TPD) experiments to investigate the adsorption and desorption behaviors of CO on the ZnCr2O4 oxides. As shown in Fig. 2b, the low-temperature peak (<300 °C) can be attributed to the weak adsorption of CO in the bulk phase, whereas the high-temperature peak (>300 °C) can be linked to the CO strongly absorbed at the oxygen vacancies [33–35]. The amount of CO desorption follows this order: ZnCr2O4-400 > ZnCr2O4-500 > ZnCr2O4-600 > ZnCr2O4-700, indicating that the oxygen vacancy concentration also follows the same order, which is quite consistent with the XPS results (Fig. 2a and Table 1). Moreover, it has been widely reported that stoichiometric ZnCr2O4 spinel (normal spinel) is hard to reduce [19,36]. However, the existence and increase of the oxygen vacancies could promote the reducibility of ZnCr2O4 spinel oxides [19,36]. Therefore, the sharp reduction peak in the temperature range of 200–350 °C (Fig. 2c) could be assigned to the ZnCr2O4 spinel oxides with oxygen vacancies [19,36]. Distinctly, the reduction peak intensity also follows this order: ZnCr2O4-400 > ZnCr2O4-500 > ZnCr2O4-600 > ZnCr2O4-700, which further demonstrates the difference of oxygen vacancy concentration (Table 1).The above characterization results demonstrate that we have successfully prepared a series of ZnCr2O4 spinel oxides with different size and surface defect sites. Then, we wondered whether these oxides with their unique structural properties would affect the activity and product selectivity for STA reaction. Therefore, ZnCr2O4 spinel oxides with distinct size were blended with the same H-ZSM-5 as composite catalysts for STA reaction at 390 °C, 3.0 MPa, oxides: H-ZSM-5 = 3:1 (mass ratio) and GHSV = 1500 mL gcat −1 h−1. The results in Table 2 and Fig. S6 indicate that this series of ZnCr2O4&H-ZSM-5 composite catalysts exhibit obviously different catalytic performance. From ZnCr2O4-700&H-ZSM-5 to ZnCr2O4-400&H-ZSM-5, the CO conversion, aromatics selectivity and STY of aromatics are significantly promoted. Since the zeolite components for all the composite catalysts are completely consistent, it can be reasonably inferred that the difference in reaction results is mainly related to the difference in the structural properties of the ZnCr2O4 spinel oxides.Therefore, we compared the catalytic performance of this series of ZnCr2O4&H-ZSM-5 composite catalysts as a function of size of ZnCr2O4 spinel oxides. As presented in Fig. 3 a, the CO conversion and aromatics selectivity are greatly affected by the crystal size of ZnCr2O4. For example, the CO conversion reaches as high as 32.6% for ZnCr2O4 with a crystal size of 7.9 nm, but it dramatically declines to 3.4% when it grows to 32.1 nm. Correspondingly, the aromatics selectivity exhibits a similar trend, i.e., it is 76% over the former ZnCr2O4 but decreases to 50.4% over the latter. Meanwhile, the crystal size of ZnCr2O4 also significantly affects the STY of aromatics (Fig. 3b). It reaches as high as 4.40 mmol gcat −1 h−1 for 7.9 nm ZnCr2O4 crystals in contrast to only 0.31 mmol gcat −1 h−1 for those with a crystal size of 32.1 nm, enhancing by about 14.2 times. The same variation trend is also found for average particle size (Fig. S7).Furthermore, we also noticed that the smaller the crystal size is, the higher oxygen vacancy concentration exists, which correlates monotonically with the CO conversion, aromatics selectivity and the STY of aromatics (Fig. 3). Oxygen vacancies are widely considered as active sites for CO hydrogenation activation in syngas conversion [1,16,22,28,29], which can influence the formation of surface intermediate species over oxides. This could be the intrinsic reason for the size effect of oxides on the catalytic performance for STA reaction.Therefore, in order to obtain further insights into the size–performance relationship, the in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed to monitor the evolution of surface intermediate species over different ZnCr2O4 oxides. As illustrated in Fig. 4 a–d, the adsorbed surface formate species (2957, 2870, 2743, 1578 and 1358 cm−1) [21,37,38] and carbonate/bicarbonate species (1492, 1380 and 1305 cm−1) [21,29,38] are clearly identified over all the ZnCr2O4 oxides after exposing to the syngas atmosphere under reaction condition, it has been widely reported that these active C1 oxygenated species are regarded as the crucial intermediate species. These active oxygenated intermediates can be further transformed to form MeOH, DME and olefins intermediates [21,23,37], or/and carbonyl compounds intermediates [38,39], which can be consumed by H-ZSM-5 (Fig. S8) and continuously converted to generate aromatics in H-ZSM-5 finally [21,23,38,39]. Noticeably, when tracking the intensity variation of the IR signals for these surface intermediate species, it can be seen that the formation of these intermediate species over ZnCr2O4-400 is extremely faster and the corresponding intensity is also significantly stronger, followed by ZnCr2O4-500, ZnCr2O4-600 and ZnCr2O4-700 (Fig. 4a–d), which is well consistent with the above-mentioned catalytic performance (Table 2 and Fig. 3).According to the results found above, we summarize the reasonable reason for the size effect of oxides on the catalytic performance for STA reaction: (1) The ZnCr2O4 oxides with smaller size and larger specific surface area can result in more surface defect sites (higher oxygen vacancy concentration), which are widely considered as active sites for CO hydrogenation activation in syngas conversion [1,16,22,28,29], thus resulting in higher CO conversion. (2) Syngas can be activated over oxygen vacancies of oxides to form C1 oxygenated intermediate species [21,27,37,38], which can be further transformed by H-ZSM-5 to produce aromatics [21,23,38]. ZnCr2O4 oxides with smaller size and higher oxygen vacancy concentration can lead to the rapid formation of more C1 oxygenated intermediate species, thus resulting in higher aromatics selectivity and higher STY of aromatics. The above results demonstrate that smaller ZnCr2O4 particles undoubtedly have a beneficial effect on the catalytic performance for STA reaction. Additionally, based on the understanding of the size–performance relationship, ZnCr2O4-400 with a smaller size mixing with H-ZSM-5 can achieve as high as 4.79 mmol gcat −1 h−1 of aromatics STY, which outperforms the previously reported some typical catalysts (Fig. S1 and Table S1) [12–19,40–44].In conclusion, we investigated the size effect of ZnCr2O4 spinel oxide in oxide-zeolite composite catalysts for syngas to aromatics (STA) reaction. The CO conversion, aromatics selectivity and space-time yield (STY) of aromatics are all significantly improved when the crystal size of ZnCr2O4 oxide decreases, which can mainly ascribe to the higher oxygen vacancy concentration and thus the rapid generation of more C1 oxygenated intermediate species. Based on the understanding of the size–performance relationship, ZnCr2O4-400 with a smaller size mixing with H-ZSM-5 can achieve 32.6% CO conversion with 76% aromatics selectivity. The STY of aromatics reaches as high as 4.79 mmol gcat −1 h−1, which exceeds the previously reported some typical catalysts. These results demonstrate the importance of the oxide size in oxide-zeolite composite catalysts for STA reaction and may be helpful to design more efficient catalysts for conversion of syngas to aromatics.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 acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 21978285, 21991093, 21991090), and the “Transformational Technologies for Clean Energy and Demonstration”, Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDA21030100). We acknowledge Mr. Yijun Zheng and Mrs. Yanli He for their assistance in the structural characterization of catalysts.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.2021.07.003.

Direct conversion of syngas to aromatics (STA) over oxide-zeolite composite catalysts is promising as an alternative method for aromatics production. However, the structural effect of the oxide component in composite catalysts is still ambiguous. Herein, we investigate the size effect by selecting ZnCr2O4 spinel, as a probe oxide, mixing with H-ZSM-5 zeolite as a composite catalyst for STA reaction. The CO conversion, aromatics selectivity and space-time yield (STY) of aromatics are all significantly improved with the crystal size of ZnCr2O4 oxide decreases, which can mainly attribute to the higher oxygen vacancy concentration and thus the rapid generation of more C1 oxygenated intermediate species. Based on the understanding of the size–performance relationship, ZnCr2O4-400 with a smaller size mixing with H-ZSM-5 can achieve 32.6% CO conversion with 76% aromatics selectivity. The STY of aromatics reaches as high as 4.79 mmol gcat −1 h−1, which outperforms the previously reported some typical catalysts. This study elucidates the importance of regulating the size of oxide to design more efficient oxide-zeolite composite catalysts for conversion of syngas to value-added chemicals.

Substituted thiocarbamides and their transition metal complexes are one such group of compounds that have promise anticancer activity (Pandey et al., 2019; Pandey et al., 2018). These compounds are more effective anticancer agents, most likely due to the presence of intramolecular hydrogen bonding in the structural framework, which increases lipophilicity and improves hydrogen bonding interactions with DNA (Mahendiran et al., 2018). The coexistence of hard nitrogen, oxygen, and soft sulphur donor atoms in the structural motif of substituted thiocarbamides (Pandey et al., 2019; Almalki et al., 2021; Al-Qahtani et al., 2021; Alkhamis et al., 2021; Abu-Dief et al., 2021) gives rise to structural variety in transition metal complexes. Moreover, ligands contains triazole moiety and their derivatives show an essential function as an antimicrobials (Emam et al., 2020; Gaber et al., 2020), anticancer (Gaber et al., 2020) and antitumoral (Matesanz et al., 2020). The significance of thiourea chemistry was improved by association with metal ions which strengthen antimicrobial activity (Maalik et al., 2019), antioxidant activity (Maalik et al., 2019; Rahman et al., 2020) and anticancer activity (Maalik et al., 2019; Abbas et al., 2020). Also, thiourea derivatives offers excessive promise as a metal transition sensor ionophore (Yahyazadeh and Ghasemi, 2013; Fakhar et al., 2016; Razak et al., 2020). Because of dyes' resistance to aerobic digestion and their stability toward heat, light, and oxidizing agents (Robinson et al., 2002; Han and Yun, 2007), many challenges arise when attempting to handle wastewater-containing dyes. Such dyes are toxic, carcinogenic, and mutagenic (Anliker, 1979; Chung and Stevens, 1993). Catalytic oxidation has recently been reported as a viable technique for curing colored water (Santos et al., 2009). Metal complexes are important cellular components that participate in a variety of biochemical processes in living organisms. Minerals have a variety of properties, such as reactivity to organic substrates, altered coordination patterns, and redox activity, to name a few. As a result, developing special coordination complexes, whether drugs, is regarded as a primary goal in the development of effective diagnostic tools (Hambley, 2009). Several bioactive minerals are being studied for their potential use in the development of new pharmaceuticals. Transition metals such as Co(II), Cr(III) and Zn(II) are required for many biological processes such as electron transfer and catalysis, and they are commonly found in enzyme or protein active sites (Thompson and Orvig, 2003).As a result of the extensive range of biological properties (Pandey et al., 2019; Pandey et al., 2018; Emam et al., 2020; Gaber et al., 2020; Matesanz et al., 2020) of thiocarbamides and triazole and derivatives, the preparation of the thiocarbamide derivative and its Cr(III), Ni(II) and Zn(II) chelate is recorded in the current research. According to the spectral and theoretical studies, various types of chelation have been suggested for metal complexes. Furthermore, the isolated Zn(II) complex's catalytic activity was reduced during the decomposition of organic Erichrome Black T (EBT) dye. Moreover, zeta potential estimation, molecular docking calculations, antimicrobial and antioxidant activities of the investigated compounds were evaluated.Solvents, metal(II) chlorides, benzoyl isothiocyanate and 1H-1,2,4-triazol-3-amine have been purchased from Sigma-Aldrich.PerkinElmer-2400 series-II analyzer have been used for partial elemental analysis. Ordinary methods (Jeffery et al., 1989) were being used to estimate the content of chloride and metal in investigated complexes. The Infra-red spectra were measured by KBr pellets via FT-IR spectrophotometer “Mattson 5000″. The electronic spectra are analyzed by via UV / Vis Spectrophotometer (Unicam). The NMR spectra of 1H and 13C with Brucker 400 MHz were handled on ligand H2L that is identified in the solvent (DMSO). The photoluminescence spectra of investigated compounds were made in DMSO solution on excitation by using a LS50B PerkinElmer Fluorimeter. Jenway 4010- conductivity meter was used to assess the molar conductivity of prepared DMSO solution of metal complexes (0.001 mol/dm−3). The spectrum of X-ray diffraction was described in detail utilizing Cu, Wavelength 1.5406 Å source on diffractometer ”the Bruker AXS Advance“. Mass spectra were obtained with ionization mode (EI) in the range of m/z = 40–1000 with Varian Mat 311. The Gouy procedure for scientific magnetic susceptibility to Sherwood has been used to determine the effective magnetic moment µeff, at room temperature per metal atom. The zeta potential measurements for the H2L and its metal complexes in water were performed via Malvern Zeta-size Nano at 25 °C. The thermal analyzer TGA-50H was used for thermal evaluation (TGA / DTG) under measured conditions such as the temperature rises by rate equal to 10 °C/min.The H2L ligand derived from thiocarbamide moiety and metal chelates have been synthesized in accordance with the procedures described in Scheme 1 . The products obtained were crystallised a number of times with absolute EtOH then with Et2O and then desiccated over anhydrous calcium chloride. TLC was performed to verify the purity of the H2L ligand. The physical and analytical results are summarized in Table 1 .Unfortunately, we could not get single crystals from the investigated compounds, thus structure optimization data have been measured with DMOL3 application in the Materials Studio software (Delley, 2002). Optimized complex frameworks were estimated using the DFT (Modeling and Simulation Solutions for Chemicals and Materials Research and Studio, 2011) process. Computations of DFT semi-core pseudopods (dspp) have been generated via double number basis sets and functional polarisation (DNP) (Warren, 1986) that was much additional successful than the duplicate gaussian basis groups (Kessi and Delley, 1998). Furthermore, The optimistic interchange-correlation feature was based on the functional (GGA) and (RPBE) (Hammer et al., 1999).The organic dye was oxidatively degraded at a pH 7.0 (buffered aqueous solution) in an air atmosphere in the presence of known doses of green oxidizing agent H2O2. 0.2 mg of the Zn2+ complex as a catalyst was added to 10 ml of the dye solution (30 mg/l), followed by an appropriate dose of H2O2 (30%) and stirring. At the end of each experiment, the flask contents were filtered, and the concentration of the dye in each filtrate was measured at λmax of 530 nm (Hassani et al., 2015). At a constant dose of H2O2 (0.2 ml) and constant dye concentration (30 ppm), the effect of time was investigated by running the reaction for 5-, 10-, 15-, 20-, and 30-minute intervals. Initially, the effect of temperature was investigated by performing the reaction at 30, 45, and 60 °C. Finally, the effect of H2O2 dose was investigated using 0.2, 0.3, 0.4, 0.5, and 0.6 ml of H2O2 and constant dye concentration (30 ppm and 30 °C, respectively).Hana Instrument 8519 digital pH meter was used to perform pH-metric measurements. Titrations were carried out at 298, 308, and 318 K. The Van Uitert and Hass relation (Uitert and Haas, 1953) is used to correct pH-meter readings in a 50 percent (v/v) dioxane-water mixture: - log H + = B + log U H o + log γ ± Where log U H o and log γ ± are the correction factors for the solvent composition and ionic strength, respectively and B is the reading.The titrations of solution mixtures towards standardized free carbonate NaOH solution (9.65x10-3 M) in 50 percent (v/v) water - dioxane at constant ionic strength are used in the experiment (1 M KCl solution). Figure S1, Supplementary Materials, depicts this. The solution mixtures were made in the following manner: a. 1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 10 ml bidistilled H2O + 12.5 ml dioxane. b. 1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 2.5 ml (5 × 10−3 M) H2L + 10 ml bidistilled H2O + 10 ml dioxane. c. 1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 2.5 ml (5 × 10−3 M) H2L + 9.5 ml bidistilled H2O + 10 ml dioxane + 0.5 ml metal ion (Ni2+) (5 × 10−3 M). 1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 10 ml bidistilled H2O + 12.5 ml dioxane.1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 2.5 ml (5 × 10−3 M) H2L + 10 ml bidistilled H2O + 10 ml dioxane.1.25 ml HCl (1.04 × 10−2 M) + 1.25 ml KCl (1 M) + 2.5 ml (5 × 10−3 M) H2L + 9.5 ml bidistilled H2O + 10 ml dioxane + 0.5 ml metal ion (Ni2+) (5 × 10−3 M).The antimicrobial activity of H2L and its respective complexes was examined against Aspergillus flavus fungus and Candida albicans fungus (ATCC 7102) as well as the bacteria G-bacteria: Escherichia Coli (ATCC 11775) and G+: Staphylococcus Aureus (ATCC 12600) by a modified disc dispersion methodology (Scheme S1, additional materials); Kirby-Bauer testing (Pfaller et al., 1988). The solution from individual substance, of investigated compounds and standard drug (Amphotericin B Antifungal Agent and Ampicillin Antibacterial Agent) in DMSO solution, were arranged for testing against spore germination. The inhibition regions diameter was expressed in millimetres (Abu-Dief et al., 2021; Aljohani et al., 2021; Abdel-Rahman et al., 2016; Abu-Dief et al., 2020).Mammary gland (MCF-7) breast cancer, and hepatocellular carcinoma HepG2 liver cells have been established using a technique stated by Mauceri, H.J.et al (Mauceri et al., 1998). Percentage of relative cell viability was determined by: Therelativecellviability % = A 570 o f T r e a t e d S a m p l e s A 570 o f U n t r e a t e d S a m p l e s × 100 2 ml of 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) solution (60 mM) and 3 ml of MnO2 solution (25 mg / ml) was added to each of the studied compounds as well as all solutions have been prepared in 5 ml of buffer solution with pH 7, using 0.1 M of aqueous phosphate. The mixture was stirred, centrifugated, filtered, and the absorption at λ734 nm of the produced green–blue solution (ABTS radical solution) was attuned to about 0.5. The spectroscopical grade of phosphate buffer / MeOH was then applied to the 50 ml (2 mM) solution of the examined species (1:1). Absorption was measured and the decrease in colour was stated as a percentage of inhibition. L-ascorbic acid is a regular antioxidant as a positive control. while sample without ABTS was a negative control (Lissi et al., 1999; El-Gazzar et al., 2009; Aeschlach et al., 1994).The inhibition% of free radical ABTS was calculated by the equation: I % = ( A b l a n k - A s a m p l e ) / ( A b l a n k ) × 100 A large number of studies have shown that epidermal growth factor receptor (EGFR) is a potential therapeutic target for the treatment of various tumors (such as colorectal and breast tumors) (Avdović et al., 2020). The inactivation of this receptor can affect the spread of cell cancer and promote cancer cell apoptosis. thus, we analyzed the potential inhibitory effects of ligand, H2L and its metal complexes.Molecule Operating Environment (MOE) software is used to evaluate the binding affinity between EGFR protein and investigated molecules (Integrated Computer-Aided Molecular Design Platform, Molecular operating environment, Chemical Computing Group, 2019). Use the specified minimization algorithm to minimize the energy of all 3D molecule structures. Then the charge of the atom is processed and the minimized potential energy is corrected. The inspected target substances are saved as a new database in MDB format (Shah et al., 2020). The crystal structure of the EGFR receptor (PDB: 3W2S) is taken from the protein database (Sogabe et al., 2013). MOE has been used to adjust the structure of molecularly bound proteins by removing ligands, adding hydrogen, and minimizing the energy of 3W2S. The minimized energy structure is used more as a binding receptor. The largest active center of 3W2S (PHE 856, LYS 745, ASP 855, CYS 797, ARG 841, ASN 842, MET 793, LEU 718, GLY 719, SER 720, VAL 726, ASP 800, and PHE 997) is the MOE position obtained by the site finder algorithm. The docking uses various features (initial re-scoring methodology: London dG with poses 10, final re-scoring methodology: GBVI/WSA dG with poses 5, placement: triangle matcher, and refining: rigid receptor) to identify and evaluate the compounds' connections with 3W2S. The S value is a rating value that measures the affinity of a compound to the receptor and is calculated by the standard MOE rating function. The RMSD is also used to compare the binding conformation with the binding reference configuration.Quantitative and spectrometric results of the ligand, H2L and its coordinated metals suggest that the metal–ligand stoichiometry is 1:1 for all complexes except the Ni(II) complex, that has 2:1 (M: L) stoichiometric ratio. The synthetic procedure of the planned complex structure is displayed in Scheme 1. The designed complexes were stable, and insoluble in the most popular organic solvents, excepting DMSO and DMF. The molar conductivity of the synthetic compounds is of a non-electrolytic behavior.The H2L 1H NMR spectrum (Figure S2, Supplementary Materials) in DMSO solution showed two signals related to SH and NH protons at 11.855 and 8.445 ppm, respectively. the presence of SH signal indicated that H2L solution was in the form of thiol. All previous signals have vanished by addition D2O (Figure S3, Supplementary Materials). The aromatic ring protons have been found between 7.334 and 7.880 ppm.The 13C NMR spectrum (Figure S4, Supplementary Materials) revealed three signals related to the C-S and C = O carbons at 168.30 and 166.00 ppm, respectively (Hosny et al., 2021; Hosny et al., 2018). Signals occurring at 151.86, 153.62 and 158.78 ppm attributed to the C = N carbon atoms of the hetero ring and the open chain, separately.The 1H NMR spectrum of Zn(II) complex (Figure S5, Supplementary Materials) in DMSO solution showed the signal related to NH proton at 8.449 ppm as well as the disappearance of signal attributed to SH proton which confirmed the proposed structure of Zn(II) complex.A brief evaluation has been made between the Infra-Red spectral data of the H2L ligand and its complexes in order to investigate the coordination action of H2L towards the metal ions. The most significant IR bands of absorption were displayed in Table 2 and Figure S6, Supplementary Materials. The IR spectrum of H2L shows distinct vibrations at 1612, 1638, 1697 and 2053 cm−1, related to ν(C = N)ring, ν(C = N)*, ν(C = O) and ν(SH), respectively (Rakha, 2000; Abdel-Monem et al., 2018; Abdel-Monem and Abouel-Enein, 2017; Hosny et al., 2018; Abdel-Rhman et al., 2019). The intense vibration at 2053 cm−1, due to the v(SH) group, indicates that, H2L is existing as thiol form in the solid phase. The vibrations referring to ν(NH) groups cannot be determined since a large destruction of 3160–3400 cm−1 overshadows their asymmetrical and symmetrical vibrations. The bending system δ of the (C = N)ring appeared at 623 cm−1, which has changed to a larger wavenumber when it is participate in complexing.In the Cr(III) complex, the H2L coordinates the Cr(III) ion as neutral bidentate via the recently formed groups of (C = N)* and (C = N) hetero ring. This type of complexation is indicated by the presence of ν(SH) and ν(C = O) at the higher and same wavenumbers, separately. This shows that these locations are not participate in coordination. Furthermore, the change of ν(C = N)* to the lower wavenumber and the change of ν/δ(C = N)triazole ring to lower and higher wavenumbers (Rastogi and Sharma, 1974), respectively, indicated mutual coordination between these locations. In addition, in Zn(II) chloride complex, the H2L works as mono-negative tridentate via deprotonated-SH, (C = O) and (C = N) ring groups. This statement was concluded by the absence of ν(SH) (Pandey et al., 1993), the change of ν(C = O) to less wavenumber (Abdel-Monem and Abouel-Enein, 2017) and the change of δ and ν of (C = N) ring to higher and lower wavenumbers, respectively (Abdel-Monem et al., 2018). Finally, the H2L operates as a binegative tetradentate throughout the binuclear Ni(II) complex. This proposal is verified by the disappearance of both ν(C = O) and ν(SH) (El-Sawaf et al., 2020) with the parallel presence of ν(C = N)** and ν(C-O) (Liu et al., 2012) as well as the changing of ν/δ(C = N) vibrations (Rastogi and Sharma, 1974; Pandey et al., 1993).Recent vibrations in regions 410–470 and 511–585 cm−1 were related to ν(M−N) (Ferraro and Walker, 1965) and ν(M−O), respectively. As well, the wide vibrations at ≈ 3432–3380 cm−1 confirm the existence of H2O in the complex (Chubar et al., 2003). The values of weight loss from the TGA data were used to distinguish between coordinated and crystallised H2O.The electronic spectra of ligand (H2L) as well as metal complexes were recorded in DMSO solution have been revealed in (Table S1, supplementary materials and Figure S7, supplementary materials). The ligand H2L presented bands in the ranges 242 and 216 nm, that could have been allocated to the (π → π*)Ar and (π → π*)ring transitions, respectively (Tossidis et al., 1987). The band at a value of 300 nm is due to the carbonyl moiety n → π* transition (Tossidis et al., 1987). The electronic spectra of Cr(III) complex shows three absorption bands at 610, 450 and 336 nm attributable to 4A2g(F)→4T2g(F)(ν1), 4A2g(F)→4T1g(F)(ν2) and 4A2g(F)→4T1g(P)(ν3) transitions, respectively characteristic for octahedral Cr(III) complexes (Parmar et al., 2010). Furthermore, the magnetic moment value, (µeff. = 3.36B.M.) can be taken as an extra indication for the octahedral geometry of Cr(III) complex. The complex, [Ni2(HL1)Cl2(H2O)2].4H2O complex have a magnetic moment values 2.63B.M., which is lesser than the measured value for a single nickel atom of d8-octaedral and/or tetrahedral complexes and larger than the diamagnetic square-planar complexes. This value may indicate the presence of Ni(II) complex in mixed stereochemistry (El-Asmy et al., 1990). This interpretation is also verified by the two bands at 346 and 484 nm assignable to 3T1(F)→3T1(P) and 3T1(F)→3T2(F) transition, respectively compatible with the tetrahedral configuration as well as one band at 396 nm is appearing because of forbidden d–d transition, reliable with the square planar geometry of Ni(II)-complex (Saha et al., 2016). Tauc’s equation, αhυ = A(hυ - Eg)r was used to calculate optical band gaps for Cr(III) and Ni(II) metal complexes, where (r = ½ or 2 for indirect, and direct transitions, respectively), (A): energy independent constant, and (Eg): optical band gap (Hosny et al., 2020). Eg Values are estimated from the plot of (αhυ) with (hυ) ( Fig. 1 ). According to the curves, the direct band gaps for Cr3+, and Ni2+ complexes are 4.64, and 4.66 eV, respectively. while, the indirect band gaps are 4.22, and 4.26 eV, respectively This information reveals that these complexes are magnetic insulators with insulating properties, high-spin frameworks, and antiferromagnetic ordering on a regular basis, i.e., metal cations in close proximity have opposite spin. Furthermore, the electronic structure of these complexes is distinguished by a predominance of metal d-orbitals in both the valence and conduction bands. (Cipriano et al., 2020).The mass spectrum of H2L (Figure S8, Supplementary Materials) revealed that the molecular ion peak [M]+ appeared at the value of m / z equal to 247, that was identical to the M.wt. of ligand. The fragmentation path of the H2L ligand was given in Scheme 2 .The photoluminescence spectrum of H2L show emission broad band at 349 nm. Moreover, its Cr(III), Ni(II) and Zn(II) complexes, show emission broad bands at 356, 349 and 359 nm, respectively, Fig. 2 . These bands could be assigned to as L-M charge transfer (Singh et al., 1999; Etaiw et al., 2018). The emission bands of Cr(III) and Zn(II) complexes indicate that both are traditional blue complexes. Furthermore, the intensity of fluorescence in all complexes is substantially lower than that of free ligand. This could be because the transition metal (M)–fluorophore (F) interaction is too strong, resulting in fluorescence quenching (Zhao et al., 2010).The XRD patterns of separated [Ni2(HL)Cl2(H2O)2].4H2O complex is depicted in Figure S9, Supplementary Materials and its (2θ)° value for peaks, the peak indexing, and inter-planar spacing (d-values) were showed in Table S2, supplementary materials. The lattice parameters of Ni(II) complex has been evaluated by using match software ( https://www.crystalimpact.com/match/ ). The [Ni2(HL)Cl2(H2O)2].4H2O complex has a triclinic space group with P −1 and lattice parameters a = 12.34 Å, b = 12.50 Å, c = 24.45 Å, α = 100.09° β = 90.31°, γ = 95.43° whose unit cell volume is 3772.94 Å3. The lattice parameters of [Ni2(HL)Cl2(H2O)2].4H2O complex exhibits a good harmony with the Crystallography Open Database (COD) No. 4,332,969 (Hernández-Molina et al., 2006).Such lattice parameters were determined by using the next relationship: triclinic 1 d hkl 2 = h a h a cos γ cos β k b 1 cos α l c cos α 1 + k b 1 h a cos α cos γ k b cos α cos β l c 1 + l c 1 cos γ h a cos γ 1 k b cos β cos α l c . 1 cos γ cos β cos γ 1 cos α cos β cos α 1 - 1 The crystalline-particle parameters were calculated using standard equations in the FWHM method (Velumani et al., 2003). the parameters were particulate sizes = 0.2897 Å lying in the nanometer scale, (2θ)°=23.99, d spacing = 3.7102 Å, FWHM = 5.1099, the crystal strain (ε) = 1.5346 and the dislocation density (δ) = 11.9127 Å−2. The regular-crystal lattice of the Ni(II) complex could be calculated from the minimised quantities of dislocation‐density (δ) and the crystal strain (ε) (El-Metwaly et al., 2020).The Zeta-potential measurements provide information regarding the stability of the colloidal suspension; the colloidal suspension is stable when the forces generating particle mutual repulsion whichplaysa prominent role. The higher absolute value of the zeta potential exhibitsthe greater the repulsion between the particles forming the suspension and thus the higher stabilityof suspension, whereas low values of ZP (±5 mV) indicate more flocculation between the particles and thus a higher tendency for instability (Bhagat et al., 2019). In this research, [Ni2(HL)Cl2(H2O)2].4H2O complex had the largest positive-positive repulsion with a potential of 20.3 mV suggesting the stability of this colloidal suspension. But, the ligand (H2L) and Cr(III) complex displayed potential of −9.34 and −6.91 mV, respectively which are in the negative range, this results shows the lower stability of the suspension of the ligand (H2L) and its Cr(III) complex. Furthermore, Zn2+ complex demonstrated a potential of 2.69 mV, indicating the instability of Zn(II) complex colloidal solution as the particles of this suspensions prefer to flocculate.Prediction of coordinated or crystallised H2O molecules may be performed utilizing TGA data (Rakha et al., 1989; Zaky et al., 2014) (Figure S10, Supplementary Materials). It can be assumed that there is an alignment between the TGA data and the proposed molecular formula. For instance, the Cr(III) complex has 5-degradation stages. The first one seems to have a weight loss of 8.05 percent between 35 and 124 °C, suggesting the elimination of two hydrated water molecules. The step two has a weight loss of 19.67 percent in the 124–352 °C temperature range, demonstrating the elimination of the H2O co-ordinating molecule and two HCl molecules. The 3rd step (from 352 to 448 °C) has a weight loss of 41.36 percent, referring to the elimination of the molecule C4H3ClN5S. The fourth stage in the 448–547 °C range has a weight loss of 6.27 percent, relating to the elimination of C2H4. Finally, unoxidized carbon and metal oxide existed as a residue. Table S3, Supplementary Materials, demonstrates the steps of decomposition of complexes.Coats–Redfern (Coats and Redfern, 1964) and Horowitz–Metzger (Horowitz and Metzger, 1963) techniques were accustomed to predict the kinetic and thermodynamic variables of the designed complexes (Figures S11-S16, supplementary materials). Tables 3 and 4 demonstrate the various kinetic parameters (A, Ea, ΔS*, ΔH* and ΔG*) of the separated complexes. we may notice that: (i) A resemblance was found between the data collected from both approaches. (ii) The good stability of the complex was proved by the high value of the activation energy. (iii) The positive value of ΔG* shows that the degradation stage is non-spontaneous process, also, the + ve value of ΔH* suggested endothermic operations (Abu-Dief et al., 2019; Abu-Dief et al., 2020). (iv) The negative ΔS* of certain degradation steps indicate that the activated fragments have a more orderly composition than the un-decomposed fragment and the degradation reactions become slow (Moore and Pearson, 1961). Although + ve values can indicate that the disorder of the decomposed fragments rises much faster than the un-decomposed fragment (Kenawy et al., 2001). A resemblance was found between the data collected from both approaches.The good stability of the complex was proved by the high value of the activation energy.The positive value of ΔG* shows that the degradation stage is non-spontaneous process, also, the + ve value of ΔH* suggested endothermic operations (Abu-Dief et al., 2019; Abu-Dief et al., 2020).The negative ΔS* of certain degradation steps indicate that the activated fragments have a more orderly composition than the un-decomposed fragment and the degradation reactions become slow (Moore and Pearson, 1961). Although + ve values can indicate that the disorder of the decomposed fragments rises much faster than the un-decomposed fragment (Kenawy et al., 2001).The optimized ligand, H2L and its metal complexes structures that labeled with the atom symbol and its number were shown in figure S17, supplementary materials and are listed in tables S4-S11, supplementary materials. Coordination induces a minor difference in bond angles and lengths existing in the thiocarbamide structure of H2L; the major initiatives in the angles of H2L were N(14)-C(13)-N(11), N(15)-N(14)-C(13), S(12)-C(10)-N(8), O(9)-C(7)-N(8), O(9)-C(7)-C(7)(2), N(11)-C(10)-N(8), N(17)-C(13)-N(11), S(12)-C(10)-N(11), C(13)-N(11)-C(10), N(8)-C(7)-C(7)-C(2), and N(17)-C(13)-N(14). Analyzing the result of H2L and the separated complexes, the following observations can be stated: (i) As predicted, the Cr(III) structure has angles close with those predicted for octahedral complexes with sp3d2 hybridization (El-Gammal, 2010; El-Morshedy et al., 2019). In addition, the optimized structure of the Zn(II) complex tends to be tetrahedral (Moore and Pearson, 1961). moreover, the Ni(II) complex produced mixed geometry (tetrahedral and square planar) including sp3 and dsp2 hybridization. (ii) (C = O), (C-O), (C-S), (C = N)azomethine, and (C = N)ring moities have larger bond lengths than those found in the ligand (H2L) attributed to the formation of metal–oxygen and metal-nitrogen bonds (Moore and Pearson, 1961). i. The bond angles of coordination atoms of ligand moiety will be changed in all complexes due to the formation of chelate rings (Fukui et al., 1954). (iii) In the Zn(II) complex, metal ion is tri-coordinated to the H2L ligand in a tetrahedral geometry with bond angles; Cl19-Zn18-S12 = 132.55°, Cl19-Zn18-O9 = 117.666°, (iv) N(14)-Zn18-O9 = 106.487°, S12-Zn18-O9 = 91.074°, N14-Zn18-S12 = 92.107° and Cl19-Zn18-N14 = 112.157° which give a small deviation from tetrahedral geometry. As predicted, the Cr(III) structure has angles close with those predicted for octahedral complexes with sp3d2 hybridization (El-Gammal, 2010; El-Morshedy et al., 2019). In addition, the optimized structure of the Zn(II) complex tends to be tetrahedral (Moore and Pearson, 1961). moreover, the Ni(II) complex produced mixed geometry (tetrahedral and square planar) including sp3 and dsp2 hybridization.(C = O), (C-O), (C-S), (C = N)azomethine, and (C = N)ring moities have larger bond lengths than those found in the ligand (H2L) attributed to the formation of metal–oxygen and metal-nitrogen bonds (Moore and Pearson, 1961).The bond angles of coordination atoms of ligand moiety will be changed in all complexes due to the formation of chelate rings (Fukui et al., 1954).In the Zn(II) complex, metal ion is tri-coordinated to the H2L ligand in a tetrahedral geometry with bond angles; Cl19-Zn18-S12 = 132.55°, Cl19-Zn18-O9 = 117.666°,N(14)-Zn18-O9 = 106.487°, S12-Zn18-O9 = 91.074°, N14-Zn18-S12 = 92.107° and Cl19-Zn18-N14 = 112.157° which give a small deviation from tetrahedral geometry. Fig. 3 displays the computed IR spectrum of the ligand, H2L in the vacuum and its observed spectrum. A slight variation between the observed and the computed can be found since the observed spectrum was evaluated for the solid material. Figure S18, Supplementary Materials is the relationship chart between the computed and the observed wavenumbers demonstrate the linear relationship according to the given equation νcal = 0.878 νExp + 119.321 whereas R2 = 0.9866.By the aid of Density functional theory (DFT), we have been able to determine different quantum variables as ELUMO, EHOMO, dipole moment, binding energy, and compounds' total energy (Table 5 ) (Liu et al., 2012; Yousef et al., 2012; Govindarajan et al., 2012; Abu El-Reash et al., 2013; Pearson, 1989; Padmanabhan et al., 2007; Gaber et al., 2018). Figure S19, supplementary materials includes the energy of frontier molecular orbitals (FMOs, that includes both orbitals of HOMO and LUMO). The data designated that: (i) the stability of studied metal complexes was demonstrated by the high Ea value and it was verified by the negative EHOMO and ELUMO value (Gaber et al., 2018; Abu El-Reash et al., 2013). (ii) Generally, the HOMO orbital was dispersed on O(9), S(12), N(8), N(11), N(15), N(14), and N(17) atoms, which are the expected position for nucleophilic attacks in the metal ion. (iii) The stability of metal complexes than the ligand, H2L has been explained from the total energy measurements (Aljahdali and El-Sherif, 2013). the stability of studied metal complexes was demonstrated by the high Ea value and it was verified by the negative EHOMO and ELUMO value (Gaber et al., 2018; Abu El-Reash et al., 2013).Generally, the HOMO orbital was dispersed on O(9), S(12), N(8), N(11), N(15), N(14), and N(17) atoms, which are the expected position for nucleophilic attacks in the metal ion.The stability of metal complexes than the ligand, H2L has been explained from the total energy measurements (Aljahdali and El-Sherif, 2013).Physical and electrostatic potential behavior is estimated by the theoretical or diffraction strategies. MEP was expressed based on electronic density (Zalaoglu et al., 2010) since has been used as a parameter in the description of nucleophilic and electrophilic attack locations and also the interaction of hydrogen bonds. Figure S20, Supplementary Materials shows the MEP which illustrated for the compounds in the study that shows that the greenish color region pointed to the neutral electrostatic potential field, whereas the blue region is the favored position for the nucleophilic attack that had the lowest zone of e's (Tanak et al., 2011). However, the reddish color section related to the region rich in e's as well as the position needed for electrophilic attack.Mulliken atomic charge has a major part to play in the mathematical interpretation of the molecular construction. Figure S21, Supplementary Materials and Tables S12-S15, Supplementary Materials show the distribution of charges of H2L, while oxygen and nitrogen atoms provide a negative value, however most carbons and hydrogens atoms provide positive values. This may be due to the e-donating potential of oxygen and nitrogen atoms.The catalytic activity of [Zn(HL)Cl]0.0.5H2O complex was investigated in the oxidative degradation of an organic dye, such as EBT dye, using H2O2 as the oxidizing agent “due to its green character”. The oxidation did not occur in the absence of the H2O2 or Zn(II) catalyst, but when both oxidant and catalyst were used, the dye degraded. As a result, the impact of H2O2 dose was studied in combination with the impact of time and temperature to determine the ideal conditions for the reaction. The effect of time was investigated at a constant dose of H2O2 (0.2 ml) and constant dye concentration (30 ppm) by running the reaction for 5-, 10-, 15-, 20-, and 30-minute intervals. Early, the effect of temperature was investigated by performing the reaction at room temperature (30 ˚C), 45, and 60˚C. Finally, the effect of H2O2 dose was investigated by using 0.2, 0.3, 0.4, 0.5, and 0.6 ml of H2O2 and constant dye concentration and temperature (30 ppm, and 30 ˚C, respectively). The results of the experiment are shown in (Fig. 4 ), which shows that the degradation of the dye increases with the time of the reaction, with about 45 percent of the dye removed after 30 min. By increasing the H2O2 dose, the degradation first improved and then began to be constant or decrease, as shown by self-quenching of OH radicals according to the succeeding OH equation: H2O2 + OH ̇→ H2O + HO2 HO2 ̇ + OH ̇ → H2O2 + O2 After 10 min, at a constant dose of H2O2 (0.2 ml) and constant dye concentration, the effect of temperature on the reaction was investigated (30 ppm). The study indicate that the capacity of dye removal increases with temperature, with only 57 percent of the dye remaining after 10 min at 60 °C compared to 85 percent at 30 °C. Table 6 shows the stoichiometric protonation constants of the investigated Ligand (L). The ligand compound investigated here has two protonation constants, corresponding to the protonated N-triazol ring and C-SH groups. As shown in Scheme 3 , the N-triazol ring has the highest pKa value (pKa1 = 8.60, at 25 °C) and the Sulphur group has the lowest (pKa2 = 4.74, at 25 °C). Fig. 5 depicts the species distribution of the ligand (L). The ligand (L2+) from Sulphur group to form HL+ tends to lose it protons after the pH is raised to the pH range of 4.71–4.74. As the conditions become more alkaline, the second proton aims to be deprotonated to a free ligand (L).The stability constants of binary chelated ligand (L) with Ni(II) metal ion as an example of divalent transition metal ions. The comparison of the titration curves of free ligand with the complexed ligand shows that adding Ni(II) ion to the free ligand solution lowers the pH. As a result, the curves associated with complexes are found at lower pHs than that of the free ligand because they require more alkali to raise the pH to the level of the free ligand. The release of protons from the coordinated ligand is an implication of complex formation. The stoichiometric stability constants associated with the inspected ligand's Ni(II) complex were estimated in 50 %(V/V) Dioxane-Water at various temperatures and are shown in Table 6. Table 6 shows the logarithms of the stability constants for all complex systems evaluated by potentiometric equilibrium titration process (1) and (2) (simplicity charges are omitted): (1) M + L ⇌ M L β = M L M L (2) M + 2 L ⇌ M L 2 β = M L 2 M L 2 The Ni-L process distribution diagram (Fig. 6 ) is investigated with the goal of investigating the changes in concentration of the Ni(II) complex with pH. The Ni-L complex is generated at pH 5.5 with a maximum of 95 %, while at pH 8, the complex Ni(L)2 is formed.The protonation of the ligand and its Ni2+ complex is attributed to the data of thermodynamic parameters that are relative to the temperature data shown in Tables 7 and 8 . ΔS and ΔH values were determined by establishing a correlation between equilibrium constant values (ln K) and temperature reciprocal values (1/T) (ln K = - ΔH/RT + ΔS/R) resulting in an intercept ΔS/R, and a slope –ΔH/R (Figs. 7 and 8 ). There are several conclusions, which are summarized below:a. The reaction involving ligand protonation is exothermic and has a net negative ΔG (Table 7).b. The data in Table 8 show that the values of log10 K1 - log10 K2 for binary complexes are positive, implying that the first ligand molecule coordinates to the metal ion is preferable to its bonding to the second (El-Sherif et al., 2012). This may demonstrate the significance of the steric effects caused by the addition of the second molecule of the ligand i.e., the NiL2 (1:2) species cannot be formed until the NiL(1:1) species is formed. This could be due to: (i) an improvement in free metal ion Lewis acidity (M+n) when compared to a 1:1 chelated ion (ML+ n–1) and (ii) steric weakness resulting from the addition of a 2nd bulky ligand to the chelated ion ML+ n–1.c. All of the negative values correlated with the complex formation of the Gibb free energy reflect the random existence of the Ni-L complex formation reactions.d. The negative heat content (ΔH) values indicate that the complex formation operation is exothermic, implying that the chelation operation works better at low temperatures.e. The ligand complex (ΔS) values are positive, suggesting that the metal complex formation is entropically favorable (El-Sherif and Eldebss, 2011), and the complexation mechanism is related tothe production of hydrogen ion (H+) and H2O molecules (Jeragh et al., 2007). During the production of metal chelates, the ligand displaces water molecules from the metal ion's main hydration sphere. The number of particles in the process thus increases, implying that the system's randomness increases with the next equation. M H 2 O n a q 2 + + L a q - ⇌ M L a q + + n H 2 O f. indicate that the complexation mechanism is spontaneous and exothermic, implying that the process of complex formation is entropically favorable.It can be concluded that log K1 > log K2 indicates that Ni(II) ion empty locations are more readily available for binding the 1st ligand than the 2nd one. The chelation mechanism-induced reaction is exothermic, spontaneous, and entropically favorable.The antimicrobial effects of the Schiff base ligand and metal complexes were undertaken toward S. aureus, C. Albicans, E. coli and A. flavus ( Table 9 ), while Amphotericin B and Ampicillin were being used as guideline for anti-fungal and antibacterial behavior, respectively. The Cr(III) complex displayed the highest inhibitory effect towards all microorganisms under investigation based on the estimated area diameter (mm/mg Sample). In addition, all compounds were evaluated no action against A. flavus fungus stain. Also, the Zn(II) complex did not show any activity against those fungus stain. Various antibacterial potency can be due to variations in the composition of the cell wall of the microorganisms (Koch, 2003).All prepared compounds have been evaluated for ABTS- antioxidant behavior Table 10 . Ni(II) complex displayed significant antioxidant potency with percent inhibition = 70.50% relative to ascorbic acid. Whereas ligand, H2L and other M2+-complexes exhibited moderate potency.2-various cell lines, HePG2 (liver carcinoma), and MCF-7 (breast carcinoma), have been used to determine the cytotoxicity of the prepared compounds under investigation (in vitro). The represented Fig. 9 demonstrates the relationship between concentration and cell viability. By such plots, IC50 values (IC50 is the concentration that inhibits 50 percent) can be determined as shown in Table 11 . The results may be summarized in the following points: a. The Ni(II)-complex demonstrates effective strong activity towards MCF-7 and HePG2 cells with IC50 values of 10.96 ± 1.0 and 8.31 ± 0.9 µM, respectively, identifying that Ni(II) complex acts as chemo-therapeutically substantial (Yousef et al., 2014). b. Cr(III)-complex exhibited moderate activity toward MCF-7 and HePG2 cell lines with IC50 values 41.03 ± 2.9 and 27.71 ± 2.1 µM, respectively. c. The weak action of Zn(II) complex towards both cancer cells with IC50 = 54.30 ± 3.5 and 61.97 ± 3.7 µM. d. Finally, H2L exhibited weak action towards HePG2 cancer cells with IC50 values equal to 54.14 ± 3.3 µM and also moderate potency towards MCF-7 cancer cells with IC50 values 33.82 ± 2.5 µM. The Ni(II)-complex demonstrates effective strong activity towards MCF-7 and HePG2 cells with IC50 values of 10.96 ± 1.0 and 8.31 ± 0.9 µM, respectively, identifying that Ni(II) complex acts as chemo-therapeutically substantial (Yousef et al., 2014).Cr(III)-complex exhibited moderate activity toward MCF-7 and HePG2 cell lines with IC50 values 41.03 ± 2.9 and 27.71 ± 2.1 µM, respectively.The weak action of Zn(II) complex towards both cancer cells with IC50 = 54.30 ± 3.5 and 61.97 ± 3.7 µM.Finally, H2L exhibited weak action towards HePG2 cancer cells with IC50 values equal to 54.14 ± 3.3 µM and also moderate potency towards MCF-7 cancer cells with IC50 values 33.82 ± 2.5 µM.It is believed that molecular binding is very important in drug discovery. The investigated compounds related to the most suitable active site of 3W2S of EGFR (PHE 856, LYS 745, ASP 855, CYS 797, ARG 841, ASN 842, MET 793, LEU 718, GLY 719, SER 720, VAL 726, ASP 800, and PHE 997) that predicted by the site-finder algorithm in MOE (Fig. 10 and Figures S22-S24). The largest binding pocket was assigned and all hits were docked against the most active site using the MOE docking software, Table 12 . As a glance in this table, the S values of investigated compounds are close to each other. Thus, the inhibitory activity may be compared according to the type and number of interaction bonds of the tested compounds with EGFR protein. According to the interaction with EGFR, the inhibitory activity order is [Ni2(HL)Cl2(H2O)2].4H2O > H2L > [Cr(H2L)Cl3(H2O)].2H2O > [Zn(HL)Cl]0.0.5H2O. This order is harmonious with experimental data. It has been observed that the contact between H-donor and H-acceptor is the most common type of interaction with EGFR receptor while Zn(II)-complex doesn’t show any interactions with EGFR receptor. Based on the results tabulated, it can be deduced that the Ni(II)-complex (Figure S23, Supplementary Materials) has the highest inhibitory activity of the EGFR protein which is similar to experimental data. In this complex, two nitrogen atoms of ligand build two H-acceptor interactions with LYS 745 and LEU 858 of EGFR (with distances 2.70 and 3.49 Å) also oxygen atoms of water molecules build two H-donor interactions with ASP 837 and ASN 842 of EGFR (with distances 2.52 and 2.98 Å). While in EGFR- H2L interaction (Fig. 10), there is one H-donor and one H- acceptor interactions with nitrogen and oxygen atoms with PHE 856, and LYS 745, respectively (with distances 3.21 and 2.96 Å). Finally, Cr(III)- complex shows only π-cation interaction of five-membered ring of ligand with LYS 745 of EGFR (distance = 4.01 Å).In the present manuscript, a new thiocarbamide derivative (H2L), was produced by the reaction of benzoyl isothiocyanate with 3-amino triazole. Its Cr(III), Ni(II) and Zn(II) complexes were synthesized and characterized using various spectroscopic techniques. The ligand operates as a neutral bidentate, mono-negative tridentate and binegative tetradentate in the Cr(III), Zn(II) and Ni(II) complexes, respectively. The photoluminescence spectra of ligand and its metal complexes exhibits that fluorescence quenching of complexes than free ligand. The suggested frameworks of these complexes have been optimized using the DFT analysis. Coats-Redfern and Horowitz-Metzger methods have been used to calculate the kinetic parameters (Ea, A, ΔH*, ΔS* and ΔG*) for titled complexes of all thermal degradation stages. The catalytic activity The Zn(II) complex demonstrated promising activity in the degradation of organic dyes, indicating that it can be used as a starting point for developing catalysts in such features. The greater cytotoxicity and ABTS-antioxidant activity were observed in the Ni(II) complex relative to the other studied compounds. Whereas Cr(III) complex exhibits the highest antimicrobial activity towards E. coli, S. aureus and C. albicans. According to molecular docking interaction, Ni(II) complex exhibits the highest inhibitory activity to the EGFR protein that agree with the experimental anticancer data.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.I wish to thank Dr. Mohammed M. El-Gamil, expert of Toxic and Narcotic Drug, Forensic Medicine, Ministry of Justice, Egypt for his help in preparing metal complexes and for his instructions for theoretical studies.Supplementary data to this article can be found online at https://doi.org/10.1016/j.arabjc.2022.104104.The following are the Supplementary data to this article: Supplementary data 1

Throughout this research, the thiocarbamide derivative (H2L), and its Cr(III), Ni(II) and Zn(II) complexes have been reported. The thiocarbamide moiety was established with a reaction of benzoyl isothiocyanate and 1H-1,2,4-triazol-3-amine. Structural elucidation of such compounds was achieved using elementary examination, spectral and magnetic experiments. The octahedral construction of the Cr(III) complex, the tetrahedral geometry of the Zn(III) complex and the mixed geometry (tetrahedral and square planar) of the Ni(II) complex have been verified by the optimization of structure using DFT. The action of Zn2+ complex in the oxidative degradation of an organic azo-dye was investigated, and it showed promising results. The thermal degradation behavior of thiocarbamide metal complexes were studied as well as the calculation of the kinetic data for title compounds (Ea, A, ΔH*, ΔS* and ΔG*) of thermal degradation steps has been tested utilizing two different techniques. Liver carcinoma (HePG2) and breast carcinoma (MCF-7) cytotoxicity as well as ABTS-antioxidant activities demonstrated the effective inhibitory of the Ni(II)-complex relative to other tested compounds. The antimicrobial activity of the compounds suggests that Cr(III) has the highest activity. Furthermore, the Molecular Operating Environment (MOE) program was used to calculate the binding affinity between the EGFR protein and the compounds under investigation.

Biomass is an effective energy carrier, contributing to the growing demand for clean and everlasting energy sources for the sustainable development of society. Biomass can be converted to biofuel through biochemical technologies (e.g., fermentation and anaerobic digestion) and thermochemical technologies (e.g., pyrolysis, liquefaction, gasification and torrefaction) [1]. One of the leading technical barriers to industrialising biomass-derived energy is the high energy input of conversion processes, especially for biomass with high moisture content. Wet bio-feedstocks require energy-costly drying operation, reducing the efficiency of the energy conversion process. However, supercritical water gasification (SCWG) technology may overcome this problem as the wet biomass can be directly gasified without an energy-intensive drying step. For high moisture biomass, the energy required for the drying process is higher than that of heating the water to a supercritical point. It was reported that the total efficiency of heat utilisation of biomass SCWG is higher than that of thermal gasification when the moisture content exceeds 27% [2]. SCWG of wet biomass can produce H2-rich syngas, which has a high heating value and can be used as a cleaner alternative of fossil fuels.Biomass gasification is a process that converts feedstock materials into gaseous products such as syngas at high temperature conditions (above 700 °C), with a controlled amount of oxygen and/or steam but without combustion. One of the primary benefits of SCWG technology is associated with the high-pressure/high-temperature water that is used as the reaction medium. The physical properties of water drastically change when the pressure/temperature conditions are above its critical point. As the reaction medium, supercritical water (SCW) offers several advantages, such as low viscosity, high diffusion coefficient, and complete miscibility with varying organics and gases, thereby enhancing the mass transfer and reaction rate in the reactor [1,3].Cellulose is one of the main structural components of lignocellulose biomass, constituting 40–50 wt% of lignocellulosic biomass on a dry weight basis [4]. Moreover, it is reported that the contribution of cellulose to H2 production during the gasification process is more than that of hemicellulose and lignin [5]. Therefore, it is essential to investigate the conversion mechanism of cellulose during the SCWG process. Extensive experimental studies on SCWG of cellulose have been carried out [1,3,6]. Cellulose comprises glucose monomers linked together by β-1,4 d-glucopyranose bonds, forming strong intramolecular and intermolecular hydrogen bonds [3]. Cellulose undergoes rapid hydrolysis and decomposes to its monomer (e.g., glucose) at very short residence times under elevated pressure/temperature. Then glucose undergoes hydrolysis to liquid-phase organic intermediates, followed by the slower formation of small quantities of stable light gases [7]. Therefore, cellulose is one of the most refractory substances that are difficult to dissolve in hot water [8], requiring harsh operating conditions to convert it into biofuel. It is widely accepted that a higher operating temperature favours the formation of H2 [1,5]. However, heating the feedstock and water to supercritical conditions is an energy-consuming process. Lowering the reaction temperature of SCWG and improving the conversion efficiency are essential for promoting the commercial utilisation of SCWG.The use of catalysts in SCWG is one of the most promising approaches to improve the gas yield while minimising the heat requirements, which can reduce the operating costs of the process [9]. Nickel-based catalyst is one of the most effective transition metal catalysts in biomass gasification for improving the gas yield and preventing the formation of tar (heavy hydrocarbons produced during SCWG, which can contaminate equipment and lead to increased maintenance costs) [10]. Extensive studies have been carried out to investigate the catalytic effect of Ni on the gaseous product yield in the biomass SCWG process [9,10]. It is widely accepted that Ni could effectively promote water gas shift reaction and steam reforming reaction [11,12], which are the two main reactions occurred in SCWG to produce hydrogen. Nickel (Ni) is known for its tendency to catalyse the cleavage of C–C, C–O, and O–H bonds [13,14], which promotes the formation of various carbonaceous products. The cracking products can be effectively dehydrogenated to produce more hydrogen [15]. Kumar and Reddy [16] investigated the impacts of Ni, Ru, and Fe on the gas yield during SCWG of banana pseudo-stem. The results showed that Ni has the highest activity in H2 generation. Ruppert et al. [14] studied the thermochemical conversion of cellulose for hydrogen production with Ni/ZrO2. They considered that the organic intermediates probably undergo dehydrogenation on the metal surface, hence increasing H2 yield. The cleavage of C–C and C–O bonds can occur to form various carbonaceous products. However, the proposed mechanisms were primarily based on the analysis of products detected during the SCWG process. Detailed structural changes at the molecular level, such as radicals and intermediates in cellulose dissociation and steam reaction, can hardly be captured through experimental methods. The exact mechanism of Ni catalytic thermal decomposition of cellulose has not been fully understood and further investigation is needed.Molecular dynamics (MD) simulation provides an opportunity to investigate the underlying mechanisms of catalytic SCWG of biomass at an atomic level. MD with ReaxFF can simulate the cleavage and forming of chemical bonds to identify elementary pathways. ReaxFF MD simulations were adopted to study the SCWG of lignin [17–20]. The structural evolution of lignin and the chemical reactions of forming CO, CO2, CH4, and H2 were obtained. Zhang et al. [21] conducted a molecular study on SCWG of glucose under microwave heating. They found that the external electric fields promote glucose decomposition to produce formaldehyde and hydrogen-free radicals, increasing H2 yield. The ReaxFF approach has been successfully employed to study the SCWG of biomass catalysed by the metal catalyst. The work of Monti et al. [22] showed that the ReaxFF approach was able to obtain an atomic-level characterisation of the crucial steps of the adsorption of the lignin molecules on the Palladium catalyst, including their fragmentation and desorption. The SCWG of lignin with Pt and Ni nanoparticles was studied by using ReaxFF simulation [23]. It was found that the Pt and Ni reduce the degradation temperature, accelerating the aromatic ring-opening process. The ReaxFF simulation study of Fe-catalysed SCWG of lignin revealed that Fe iron with a low oxidation state contributes to the formation of CO, while iron with a high oxidation state was beneficial to increasing CO2 yield [24]. The evolution of the lignin decomposition catalysed by Ni was investigated with ReaxFF simulation [25]. The results indicated that Ni could potentially accelerate the scission of C–O bonds and destroy the conjugated π bond of the aromatic ring during the ring-opening process. The generation process of H2 molecules occurring on the Ni surface was presented. The thermal stability of carboxymethyl cellulose on the Fe2O3 surface was studied by Saha et al. [26]. It was found that cellulose can be adsorbed on the metal surface via the formation of bonds between Fe and oxygen atoms. The chemisorption would bulge the Fe slightly out of the Fe2O3 surface.Although significant experimental work has been performed on Ni catalytic SCWG of biomass, there is a lack of detailed understanding on the chemical processes involved at the molecular level. Further fundamental modelling studies are required to deepen the understanding of the catalytic and micro reaction degradation mechanisms during the SCWG process. This could provide a basis for optimising operating conditions and developing high-efficiency catalysts, thereby promoting the utilisation of SCWG technology. To the best of our knowledge, the ReaxFF simulation of Ni catalysed SCWG of cellulose has not been carried out. This study investigated the effect of Ni on cellulose depolymerisation and ring-opening process. The effects of temperature and cellulose-to-water mass ratio (C/W) on gaseous products were investigated. The detailed gaseous product generation pathways were analysed. Besides, the influence of temperature and C/W on carbon deposition behaviour on Ni nanoparticle (NiNP) was investigated.All MD simulations in this work were conducted using the ReaxFF force field [27]. The description of connectivity-dependent interactions in the ReaxFF force field is based on bond order formalism. Bond order is determined by interatomic distance using an empirical formula, including contributions from σ, π and ππ bonds. The chemical reactions during the time intervals can be analysed based on the interatomic potential and the bond order. Nonbonded interactions, such as Coulomb and van der Waals interaction, are calculated independently. The charge equilibration (QEq) method adjusts the partial charge on individual atoms. The following equation calculates the energy of each particle: (1) E s y s t e m = E b o n d + E o v e r + E u n d e r + E l p + E v a l + E t o r + E v d W a a l s + E c o u l o m b where E bond , E over , E under , E lp , E val , E tor , E vdWaals , and E coulomb stand for bond energy, overcoordination energy penalty, undercoordination stability, lone pair energy, three-body valence angle energy, four-body torsional angle energy, van der Waals energy, and Coulomb energy, respectively. In this study, the C/H/O/Ni parameter set [28,29] developed for modelling hydrocarbon chemistry catalysed by Ni was adopted to study the cellulose SCWG catalysed by Ni nanocatalysts. The verification of the adopted force field was carried out in our previous work [30].Cellulose (C6H10O5)n is a polysaccharide consisting of a linear chain of several hundred to many thousands of β-1,4 linked d-glucopyranose units. Fig. 1 shows the model used in the MD simulations. The model construction starts with a monomer, and the unimolecular d-glucopyranose was built and optimised using the Materials Studio [31] Forcite module. Ten d-glucopyranose monomers were connected to form a polymer, as shown in Fig. 1(c). Face-centred cubic lattice of NiNP was created on a web-based crystallographic tool [32]. The minimum surface energy of corresponding Miller indices of (111), (100) and (110) was adopted from the work of Chen et al. [30]. The melting temperature of NiNP depends on its size, and the melting temperature decreases with decreasing radius of NiNP. The melting temperature of 3 nm NiNP simulated with the ReaxFF force field is around 1700 K [30], which is lower than the simulation temperature (1800 K∼2200 K) in this study. The simulated melting temperature of 4.0 nm NiNP is around 2000 K by using the same ReaxFF force field [30]. Although the 4.0 nm NiNP would melt to some degree when the temperature is above 2000K, the NiNP still keeps a spherical shape. Therefore, NiNP with a diameter of 4.0 nm was adopted to maintain the integrity of NiNP during the simulation in this study.Eight reaction systems S1–S8 were built to investigate the Ni catalytic SCWG of cellulose, as listed in Table 1 . Different cellulose molecule numbers have been tested to eliminate the effect of atom number on the simulation results. Ten cellulose molecules were adopted to ensure the validity of simulation results. To observe the reactions that occur on the Ni surface, the catalyst to biomass ratio considered is relatively high. Cases S1–S6 are used to study the effects of temperature and catalyst on the SCWG of cellulose. Cases S5, S7, and S8 are used to study the effect of cellulose-to-water mass ratio (C/W) on the catalytic SCWG of cellulose. The system pressure would affect the yield of gaseous products. An increase in pressure will shift the methanation reactions ( CO + 3 H 2 ↔ CH 4 + H 2 O , CO 2 + 4 H 2 ↔ CH 4 + 2 H 2 O ) to the right, thereby enhancing the formation of CH4 [1]. Therefore, the simulation box dimensions were adjusted to keep the same pressure in systems S7 and S8 with varying water molecules.Sorensen and Voter [33] pointed out that an elevated temperature could accelerate the reaction process and thus significantly extend the simulation time scale, which has become a familiar and effective strategy in ReaxFF MD simulation [24,34]. For reaction rates described by the Arrhenius equation, increasing the temperature would increase the reaction rates but not the activation energy barrier. Salmon et al. [35] studied coal pyrolysis using ReaxFF simulation at an elevated temperature. They compared the simulated product distribution with experimental results and concluded that elevated temperature did not influence the reaction pathways during coal pyrolysis. Accordingly, elevated reaction temperatures were chosen in this work to study the catalytic mechanism of Ni during the SCWG of cellulose.Initial configurations of all models were built by using Packmol [36]. Cellulose and water molecules are distributed randomly into the cubic box, and the box dimensions are listed in Table 1. In catalytic SCWG (e.g., cases S4–S8), NiNP was fixed in the centre of the unit cell and water and cellulose molecules are distributed around, as shown in Fig. 1(e). After system energy minimisation, the simulation cell was relaxed at 300 K for 20 ps. Subsequently, the equilibrated system was heated to the final reaction temperature with a heating rate of 15 K/ps. Then, the simulations would last at the target temperature for 2 ns. All simulations were performed using an isochoric-isothermal NVT (fixed atom numbers, volume, and temperature) ensemble. A time step of 0.25 fs was assigned. The trajectories and species information were outputted every 100 steps. The linear and angular momentum of NiNP was zeroed every 10 timesteps.The periodic boundary condition was applied in all directions. The initial velocities for all atoms were generated randomly following the Maxwell-Boltzmann distribution. Nosé-Hoover thermostat and barostat were adopted to control the system temperature and pressure with a temperature and pressure damping constant equal to 100 times and 1000 times of the time step, respectively. A bond order of 0.3 was employed to identify chemical bonds between pairs of atoms [28,30]. All simulations were repeated three times with different initial configurations and velocity distributions. The ReaxFF MD simulations were performed with the REAXC package [37] in the Large-scale Atomic/Molecular Massively Parallel Simulation (LAMMPS) [38].The first step in cellulose conversion involves its depolymerisation to oligomers or d-glucopyranose [4], which undergoes hydrolysis to form liquid-phase organic intermediates via scission of C–C and C–O bonds. Guo et al. [9] established the mechanism of Ru catalytic gasification of d-glucopyranose. Hydroxyl groups are adsorbed to the catalytic Ru surface predominantly through oxygen atoms. The reactant undergoes dehydrogenation on the catalyst surface, followed by subsequent cleavage of C–C or C–O bonds, which results in syngas production. However, the detailed adsorption and degradation process on catalyst surfaces are not readily accessible by experiments.The depolymerisation and ring-opening percentage of cellulose during the heating period in cases S2 and S5 are shown in Fig. 2 (a) and (b). The depolymerisation and ring-opening percentage are computed by the following equations: (2) D e p o l y m e r i s a t i o n p e r c e n t a g e = C l e a v a g e o f β - 1 , 4 l i n k a g e I n i t i a l n u m b e r o f β - 1 , 4 l i n k a g e × 100 ( % ) (3) Ring - opening percentage  =   Number of opened ring  Initial number of ring × 100   ( % ) It can be seen that Ni could accelerate the depolymerisation and ring-opening process of cellulose. The depolymerisation and ring-opening occur at around 75 ps in the absence of Ni catalyst, and almost all β-1,4 linkages are cracked after 125 ps. While the start points of β-1,4 linkage cleavage and ring-opening are around at 50 ps in the presence of Ni catalyst, these processes were completed after 100 ps. The results show that decomposition of cellulose can occur at a lower temperature, which helps reduce the cost of biomass SCWG. The two cases show a similar onset time and evolution trend in depolymerisation and ring-opening processes, demonstrating that the ring-opening takes place immediately after cellulose is depolymerised into monomers. This is because the cleavage of β-1,4 linkage would lead to the structural instability of the corresponding monomer, resulting in the ring-opening of d-glucopyranose.The ring-opening of the d-glucopyranose monomer can be achieved by the cleavage of the C–C or C–O bond. Fig. 2(c) shows the cleavage percentage of different types of bonds. Around 64% of rings were opened by the cleavage of the C–O bond in the absence of Ni, while this figure increases to 70% when Ni was added. This result suggests that more rings of d-glucopyranose tend to be opened via the cleavage of the C–O bond under the effect of Ni catalyst.Both high temperature and catalyst would promote bond breaking. The energy of atoms increases with increasing temperature, and the bonds between the atoms become more unstable and eventually break. Catalysts make this process more efficient by lowering the activation energy. If an atom forms a chemical bond with Ni atoms, the cleavage of other chemical bonds connected to this atom will take place. The cracking of such bonds is considered as catalytic cleavage and the others are considered as thermal cleavage. The thermal cleavage and catalytic cleavage of bonds during depolymerisation and ring-opening process in catalytic SCWG (case S5) are shown in Fig. 3 . It can be seen that 87% and 88% of bond breakings take place via thermal cleavage during depolymerisation and ring-opening process, respectively. Thermal cleavage of bond plays a dominant role in the bond-breaking process. Therefore, depolymerisation and ring-opening rate in noncatalytic and catalytic SCWG are similar after 100 ps, when the temperature of the system reaches a certain value, as shown in Fig. 2.As the skeleton of organic matters, the dissociation kinetics of C–C and C–O bonds play a vital role in cellulose decomposition. The decomposition reactions of d-glucopyranose were considered to be first-order reactions [39]. Initial and equilibrium numbers of C–C and C–O bonds can be used to calculate the activation energy of the corresponding bond [40,41]. The reaction rate constant, K, is determined by the following equation [40]: (4) ln N 0 − ln N t e q =   K t e q where N 0 and N teq are the numbers of C–C or C–O bonds at initial and equilibrium stages. The reaction rates are analysed by the Arrhenius equation: (5) K = A exp ( − E a R T ) where R is the universal gas constant. The activation energy (Ea) and the pre-exponential factor (A) in Eq. (5) are calculated by linear fitting. Fig. 4 shows the change in the activation energy of C–C and C–O bonds in the absence and presence of a catalyst. The activation energy of the C–C bond is 25.33 kJ/mol without catalyst, and this figure decreases to 24.02 kJ/mol when Ni catalyst is added. Activation energies for C–O bonds without and with Ni are calculated as 24.95 and 22.97 kJ/mol, respectively. It can be seen that the activation energy of C–O bonds is lower than that of C–C bonds. Thereby, the ring-opening of d-glucopyranose monomers is prone to take place via the cleavage of C–O bonds. Moreover, the activation energy reduction of C–O bonds (1.98 kJ/mol) under the effect of catalyst is more significant than that of C–C bonds (1.31 kJ/mol), demonstrating Ni is more efficient in the cleavage of the C–O bond than the C–C bond. Consequently, the proportion of C–O cleavage increases when Ni is added during the ring-opening process, as shown in Fig. 2(c). It has been reported that cellulose is easier to gasify than lignin in the presence of Ni catalyst [42], which can be explained by the structural differences between cellulose and lignin because there are more C–O bonds in cellulose molecules than in lignin. Fig. 5 shows the time evolution of the total number of H2, CO, and CO2 molecules at different temperatures during SCWG of cellulose in the absence and presence of Ni catalyst. Temperature is one of the most dominant parameters that affect the gaseous product yield, especially when the reaction occurs without a catalyst [1]. Generally, the gaseous produce yield increases with an increase in reaction temperature as high temperatures favour the scission of C–C and C–O bonds. The simulation results show that H2 yield increases with increasing temperature, which is consistent with the experimental results [43]. The free radical reactions in water are believed to be temperature-dependent. When the conditions are above the critical point of water, free radical reactions dominate over ionic reactions [3]. Therefore, water splitting at higher temperature generates more H free radicals. In addition to water splitting reaction, biomass dehydrogenation reaction would also be enhanced at high temperatures [44]. The increase of H free radical number leads to the increase in H2 yield, which will be discussed subsequently.Ni catalyst will significantly increase the yield of H2, as shown in Fig. 5. The H2 generation pathways were analysed to explore the effect of Ni on the H2 production mechanism. The generation of H2 is mainly through the following three pathways: ① R-H + H • → R • + H 2 ② H 2 O + H • → H 2 + OH • ③ H • +H • → H 2 where pathway ①: hydrogen transfer reactions, where H radical interacts with the H atoms in cellulose to produce H2 [17,45]; pathway ②: H radical interacts with water to produce H2 and OH at elevated temperature [17,46]; pathway ③: H radical termination reaction, where two H radicals interact with each to produce H2 [17,45]. R is the abbreviation for any other groups.The occurrence frequency and proportion of pathways ①-③ during SCWG of cellulose in the absence and presence of Ni catalyst are presented in Fig. 6 (a) and Fig. 6(b), respectively. The results suggest that pathway ① plays a dominant role in the H2 generation in the absence of Ni catalyst, especially at relatively low temperatures. It can be seen that around 68% of H2 was produced via pathway ① in 1800 K. Only a few H2 molecules were generated through pathway ③ without Ni catalyst. However, pathway ③ becomes the main H2 generation path with the addition of Ni catalyst. More than 70% of H2 molecules were generated via pathway ③, and this figure reached 82% in 1800 K. This is because a large number of H free radicals would be generated by the water splitting and cellulose dehydrogenation reactions on the Ni surface. Increasing the concentration of H free radical enhances the occurrence frequency of pathway ③. Meanwhile, the frequency of all three pathways increases with increasing temperature, and the increase of pathway ② is the most significant. Therefore, the proportion of pathway ② increases with increasing reaction temperature, as shown in Fig. 6(b).The yield of H2 is closely related to the number of H free radicals. The generation paths of H free radicals were analysed to investigate the effect of Ni on H2 yield. The H radicals could be produced from the water splitting and cellulose dehydrogenation reactions, as listed in the following pathways: ④ H 2 O ↔ OH • + H • ⑤ R-O-H ↔ R-O • + H • ⑥ R-C-H ↔ R-C • + H • where pathway ④: free radical reaction of water, where H2O splits into H • and O • H or the radicals recombine into H2O at elevated temperature [3]; pathway ⑤ and pathway ⑥: dehydrogenation reaction of cellulose, where the H atom connects to or disconnects from oxygen and carbon atoms respectively [41]. Due to the high activity of free radicals, the reaction pathways ④-⑥ are reversible. Fig. 7 (a) shows the frequency differences of forward and reverse reaction of pathway ④-⑥ in cases S2 and S5. It should be noted that the frequency of reverse reaction could be higher than that of the forward reaction, which is because the reactants involved in the reverse reaction could be produced from other reactions. For example, O • H could be generated from pathway ④, while it can also be produced from the cleavage of the C–O bond in cellulose (e.g., R - OH → R • + O • H ). The cleavage of β-1,4 linkage and C–O bond in d-glucopyranose would produce R - O • that involved in the reverse reaction of pathway ⑤. The increase in these reactant concentrations would shift the reactions into the reverse direction. The results show that the H free radicals mainly come from the dehydrogenation reaction of H atoms connected to C atoms, e.g., pathway ⑥. The H radicals generated from pathway ⑥ increase with the addition of Ni, which can be attributed to the promotion effect of Ni on C–H bond cleavage [47].Although the water splitting reaction would produce H radicals, the O • H generated from water and cellulose would consume H radicals to form water simultaneously. The frequency of pathway ④ reverse reaction is higher than that of the forward reaction, leading to an increase in water molecule number. This is due to the structural features of cellulose, which has a large number of hydroxy groups. The increase in O • H concentration shifts the reaction of pathway ④ to the reverse direction, especially with the addition of Ni catalyst, since Ni could promote the scission of C–O bonds to produce more O • H .Experimental results suggested that one of the roles of water in SCWG is being a source of hydrogen and free radicals [3]. However, quantitative information on such effect is difficult to obtain through experimental approaches. In ReaxFF MD simulation, the role of water in producing hydrogen and free radicals can be identified by tracing the evolution of the original water (the water that was added to the system in the initial stage). Fig. 7(b) indicates the total water molecule number and the source of O atom in water in cases S2 and S5. Although it was observed that the frequencies of forward and reverse reaction of pathway ④ are high, especially on Ni surface, only a small part of the original water split into H • and O • H in the end, as indicated by the red line in Fig. 7(b). The results indicate that only a limited number of H radicals produced from pathway ④ forward reaction could be the source of H2 generation. Moreover, Ni could promote the splitting reaction of water, and the reduction of original water in the presence of Ni is more significant. The increased water in two cases is ascribed to the combination of H radicals and hydroxy groups dissociated from cellulose since the oxygen atom in increased water mainly comes from cellulose, as indicated by the violet line in Fig. 7(b). The increase of total water molecule number in Ni catalytic SCWG is more significant than in noncatalytic SCWG as Ni could promote the cleavage of C–O bond to produce more hydroxy groups. It can be deduced that water plays a limited role in providing H free radicals to produce H2, while the hydrogen atoms in cellulose are the primary source of H2 generation.A schematic diagram of H2 generation pathways during the noncatalytic and Ni-catalytic SCWG of cellulose is shown in Fig. 8 . In the absence of a catalyst, H free radicals would be generated via dehydrogenation of cellulose, and a small amount of water would also be split into H free radicals and O • H . H radicals generate H2 through pathways ①-③, where pathway ① plays a leading role, followed by pathway ②. In the presence of Ni catalyst, the decomposed molecular fragments and water would be adsorbed on the Ni surface, where the scission of the C–H and O–H bonds is enhanced to produce a large number of H radicals. H radicals undergo radical termination reactions to produce H2 (pathway ③), which dominates the H2 generation. Meanwhile, the generated O • H would consume H radicals to form water. In the absence of catalyst, mainly comes from pathway ②. Nevertheless, cellulose also produces some O • H on the Ni surface via the cleavage of C–O bonds when Ni is added. The O • H generated through serval pathways would consume a relatively large amount of H radicals, leading to an increase in H2O molecule number. It can be found that the generation of H2O is an H radical consumption process. The concentration of H radicals would increase if the number of O • H in the reacting system can be suppressed, thereby increasing the yield of H2.It can be seen in Fig. 5 that the yield of CO is enhanced at elevated temperature in the absence of a catalyst, which is consistent with the experimental results [48]. The elevated temperature would promote the cleavage of C–C bonds, thereby enhancing the yield of gaseous products. CO yield decreased significantly when the catalyst was added. Yoshida et al. [42] attributed the reduction of CO to the enhancement of water-gas shift reaction ( CO + H 2 O → CO 2 + H 2 ), and the disproportionation of carbon monoxide adsorbed on the catalyst surface ( 2 CO → CO 2 + C ). However, this study found that the molecular fragments produced by cellulose dissociation would be adsorbed on the catalyst surface [30]. The C–O bonds would be cracked under the catalytic effect of Ni, as shown in Fig. 9 (a). Oxygen atoms that might be used to generate CO are prone to be detached from organic fragments to produce water by interacting with H radicals. It is demonstrated that the deoxygenation and dehydroxylation of organic fragments on Ni surface are the main reason for CO reduction.There was a slight change in the yield of CO2 when the temperature increased, as shown in Fig. 5, which is ascribed to the structural features of cellulose. The carbon atom connected with two oxygen atoms is the primary source of CO2, as illustrated in Fig. 9(b). There are a small number of β-1,4 linkage in cellulose; thus, the yield of CO2 is lower than H2 and CO [49]. Consequently, the effects of temperature and catalyst on CO2 production are weaker than those on H2 and CO. Fig. 10 presents the yields of H2, CO, and CO2 under different C/W at 2000 K. The H2 yield increases slightly with the increasing number of water molecules, which is consistent with experimental results [23]. Increasing the water molecule number favours the forward reaction of pathway ④, leading to an increase in H radical number, as shown in Fig. 11 (a). In the presence of Ni catalyst, a large number of H radicals would be generated via water splitting and cellulose dehydrogenation reactions on the Ni surface. Then H radicals are continuously consumed to produce H2 and H2O. A high concentration of H radicals contributes to the formation of H2.Typically, water and small organic fragments would be adsorbed on metal catalyst [50], as shown in Fig. 11(b). It was observed that the addition of water occupies a part of the active sites on the catalyst surface, which weakens the adsorption capacity of the catalyst to organic fragments [23]. When most of the active sites of the Ni catalyst are occupied by water or hydroxyl group, the small dissociative fragments outside Ni surface will generate CO via the cleavage of C–C bonds. Therefore, the yield of CO increases with the addition of water, as shown in Fig. 10. The influence of C/W on CO2 production is negligible since CO2 mainly comes from the carbon that is connected to the two oxygen atoms, and the production of CO2 is relatively low as stated before.Deactivation of catalysts is unavoidable in the catalytic SCWG reaction process. Carbon/coke deposition on catalyst surface is regarded as one main problem for the deactivation of Ni-based catalysts [51]. To investigate the carbon deposition and permeation behaviour in SCWG of cellulose, the carbon deposition and permeation on Ni surface under different temperatures and C/W were analysed. The block of NiNP was divided into three zones (inside the spherical shell) according to different radii, as illustrated in Fig. 12 (a). Carbon deposition rate is determined by the number of carbon atoms in different zones. Fig. 12(b) shows carbon migration on the catalytic surface and in catalyst pores. There are a small number of carbon atoms in Zone 3 at 100 ps, and then the carbon atoms permeate into the inside of NiNP at 250 ps. The evolution of carbon number in different zones under varying temperatures is presented in Fig. 12(d). The results show that carbon atoms infiltrate into Ni over time, and there is no noticeable difference in the total number of carbon when reaching an equilibrium state. However, the difference in the deposition rates under varying temperatures is appreciable. A small number of carbon atoms can be detected in Zone 2 at 250 ps when the temperature is 1800 K. With the increase in temperature, carbon would reach Zone 2 at an earlier time, at around 150 ps. The time instants for carbon reaching Zone 1 at 1800 K, 2000 K, and 2200 K are around 375 ps, 180 ps, and 130 ps, respectively. The results indicate that the permeation rate of carbon increases with increasing temperature. Nevertheless, the number of carbon molecules at the equilibrium state in different zones is roughly the same, which suggests that temperature has a negligible impact on the degree of carbon permeation.To uncover the carbon permeation mechanism on NiNP at different temperatures, the atomic order of Ni atoms was analysed. Steinhardt's bond orientational order parameters Q l (where l can take an integer value between 0 and infinity) [52] were used to explore the local atomic environment. These order parameters are mathematically defined based on certain rotationally invariant combinations of spherical harmonics calculated between atoms and their nearest neighbours, providing information about local atom environments. Q l has been used for various purposes, such as the structure identification of solid and liquid systems [53]. Commonly Q6 is used in the identification of cubic lattice structure. All the particles in a perfect ordered structure have the same value of Q6. As a results, Q6 values can be used to determine whether an ordered structure is beginning to turn into a disordered structure. Therefore, Q6 was adopted to characterise the atomic order of NiNP, which was built as a face-centred cubic lattice structure in this study. The magnitude of Q6 is large when the Ni atoms are ordered and small when the Ni atoms are disordered. Fig. 12(c) shows the sectional view of NiNP atomic Q6 values at different times. The internal atoms of NiNP are in an ordered state in the initial stage and then in transition to a disordered state over time. The evolution of averaged atomic Q6 of Ni atoms in different zones is presented in Fig. 12(e). The equilibrium Q6 values decrease with increasing temperature. The position of atoms in the outermost shell (e.g., Zone 3) shifted first, and then the order degree of internal atoms decreased over time as heat was transferred to the interior region. The Q6 values in Zone 2 and Zone 1 decreased more slowly under lower temperatures. For example, the Q6 value in Zone 1 at 2200 K decreases significantly at around 125 ps, while the time instant for Q6 dramatical reduction at 1800 K is about 310 ps. The decrease in Q6 value represents that the crystal structure of NiNP is destroyed, and there is a relatively large displacement between Ni atoms. It is easier for carbon atoms to infiltrate into the NiNP when the displacement between Ni atoms becomes larger. Fig. 13 shows the evolution of carbon numbers in different zones under varying C/W. The time that carbon reaches Zone 2 and Zone 1 is roughly the same under different C/W conditions, demonstrating that C/W has a negligible influence on carbon permeation on NiNP surface and in NiNP pores. Nevertheless, high C/W would inhibit the carbon deposition number on NiNP. The equilibrium carbon numbers in total and in different zones decrease with the increase in water molecule number. Wu and Liu [51] also reported that the increase of the steam to carbon ratio could favour carbon elimination during bio-oil gasification. The carbon elimination from the catalyst surface can be ascribed to the addition of water occupying the active sites on the catalyst surface, preventing the dissociative carbon atoms from attaching to NiNP.In this study, Nickel catalysed gasification of cellulose in supercritical water is investigated by using reactive MD simulation. The depolymerisation and ring-opening process of cellulose, effects of Ni and C/W on gaseous product yield, and carbon deposition behaviour on Ni catalyst were investigated. This study provides detailed information on Ni-catalysed cellulose SCWG at an atomic level.Calculated activation energies show that Ni can decrease the activation energy of C–C and C–O bond cleavage, promoting cellulose depolymerisation and ring-opening process. Cellulose could be gasified at a lower temperature with the addition of Ni. The activation energy reduction of C–O is more significant than that of C–C bonds under the effect of Ni.The H2, CO, and CO2 yields increase with increasing temperature. H2 yield increases significantly in the presence of Ni due to the large number of hydrogen free radicals generated by the cleavage of C–H and O–H bonds on the surface of NiNP. H radicals can not only interact with each other to produce H2 but also interact with H atoms on water and cellulose to generate H2. The O • H generated would consume H radicals, leading to an increase in H2O number. The concentration of H radicals would increase if the number of O • H in the reacting system can be suppressed, thereby increasing the yield of H2. Simulation results show that water plays a limited role in providing H free radicals to produce H2, the hydrogen atoms in cellulose are the primary source of H2 generation. The cellulose cracking fragments would be adsorbed on the NiNP surface, where these fragments undergo deoxygenation and dehydroxylation reactions, leading to a reduction of CO and CO2 yields. The addition of water will occupy the active sites on Ni surface, reducing the probability of molecular fragments attaching to the Ni surface. The small dissociative fragments outside Ni surface tend to generate more CO.The carbon deposition on the NiNP surface results in the deactivation of the catalyst. Due to the movement of Ni atoms at high temperature, the adsorbed carbon would infiltrate into the NiNP. Results suggest that carbon permeation rate increases with increasing temperature as the relative displacement of Ni atoms would be increased under higher temperatures. The increase in water mass fraction can favour carbon elimination from the catalyst surface because water would occupy the active sites on the NiNP surface, resulting in the failure of carbon adsorption. This study elucidated the detailed mechanism of Ni-catalysed cellulose SCWG from the molecular point of view, providing a basis for further biomass utilisation and cost reduction.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.Supercomputing time on ARCHER is provided by the “UK Consortium on Mesoscale Engineering Sciences (UKCOMES)” under the UK Engineering and Physical Sciences Research Council Grant No. EP/R029598/1. This work made use of computational support by CoSeC, the Computational Science Centre for Research Communities, through UKCOMES.

Reactive force field (ReaxFF) molecular dynamic simulation was performed to elucidate the mechanism of Ni-catalysed supercritical water gasification of cellulose considering the effects of temperature and cellulose to water ratio. Simulations showed that Ni could decrease the activation energy of C–C and C–O bond cleavage, promoting the depolymerisation and ring-opening process of cellulose. The yields of gaseous products increase with the increasing temperature. H2 yield mainly depends on H free radical number, which can be generated from cellulose dehydrogenation and water splitting reactions. These two reactions were promoted on Ni surface, leading to an increase in H2 yield. In the presence of Ni catalyst, water plays a limited role in providing H free radicals to produce H2, while the hydrogen atoms in cellulose are the primary source of H2 generation. Meanwhile, reducing the concentration of O • H could enhance H2 production as the combination of O • H and H • is a H radical consumption process. Small organic fragments would be absorbed on the Ni surface, where they undergo deoxygenation via the cleavage of C–O bonds, resulting in a decrease in CO and CO2 yields. The increase in water mass fraction would promote the H2 yield as more H radical would be produced due to water splitting reaction. Moreover, the addition of water would occupy the Ni active sites and prevent the adsorption of organic fragments. These dissociative fragments are prone to produce more CO. The carbon deposition on the Ni surface results in the deactivation of the catalyst. Simulation results suggested that carbon deposition and permeation increase with increasing temperature. In contrast, the increase in water mass fraction can favour carbon elimination from the catalyst surface.

Cu is the only elemental metal electrocatalyst that can reduce CO2 to a wide range of hydrocarbons and/or oxygenates, but a specific product pathway usually shows low current density and Faradic efficiency [1–5]. In recent years, extensive research efforts have been devoted to addressing the above challenges. Different strategies focus on the modulation of properties of catalysts through both intrinsic routes including morphology design of crystal surfaces (crystal surface orientation, steps, edges, roughness, and particle size, etc.) and phase engineering, or extrinsic controls, such as oxidation state tuning, defect engineering, doping, composition variation, and so on [6–12]. For instance, the product distribution on Cu was shown both experimentally and theoretically to depend critically on the characteristics of Cu surfaces [7,13,14]. Montoya et al. reported that Cu(100) exhibits a lower CO dimerization activation barrier than Cu(111) [15]. Bagger et al. found that Cu(100) × (110) step is the specific ethanol-producing site by experimental and geometric analysis [7]. Regarding defect effects, the grain boundary density engineering in Cu could yield higher activity and selectivity for multi-carbon oxygenates [16]. In addition, a new Cu phase with 4H atomic arrangement exhibited enhanced overall activity and ethylene selectivity in CO2RR compared to the conventional FCC Cu [12]. Taking advantage of dopant-modulated oxidation state of Cu, Zhou et al. demonstrated enhanced selectivity of C2 products by doping boron into Cu [17]. More generally, Nørskov et al. suggested that alloying Cu-based catalysts by dopants with different oxygen affinity can be adopted to break the scaling relationship among different reaction intermediates to lower the overpotential [4]. Thereafter, plenty of Cu-based intermetallic compounds have been proposed, such as CuAl nanoalloy [11], AuCu nanoparticles [18], and polymer-supported CuPd nanoalloys [19].Recently, a new type of catalyst for CO2RR, single-atom alloys (SAAs), serving as model systems for understanding fundamental catalytic properties, has attracted a lot of attention [20]. SAAs are a class of single-atom catalysts, with small amounts of isolated metal atoms dispersed across the surface of a metal matrix. The isolated metal atoms typically have different catalytic activities and/or selectivities compared with host metals on CO2RR. The unique geometry of SAAs has the advantages in both increasing noble metal utilization ratio and breaking the limits of scaling relationships among different intermediates [20–23]. Up to date, SAAs has been successfully applied to various electrocatalytic reactions, including PtPd SAAs for the oxygen reduction reaction [24], AuRu and PtPd SAAs for hydrogen evolution reaction (HER) [24,25]. However, there have been only few studies on SAAs for CO2RR, let alone the extensive investigation on different combinations of metal species in SAAs and their effects on CO2RR.Although CO2RR catalysts have experienced rapid development with improved activity and selectivity in past decades, the large amount of data obtained in these studies combined with the complex reaction processes and the sensitivity to local environment of CO2RR make it challenging to develop a unified understanding of underlying mechanisms and design catalysts purposefully. First-principles simulations without empirical parameters are useful for facilitating the design of catalysts for CO2RR [26–32]. Based on theoretical calculations of free energies and microkinetic model, the thermodynamic descriptors of catalysts, i.e., intermediates adsorption energies, can be used to predict their properties, taking advantage of Sabatier-type trade-off between the adsorption of reactants and desorption of products [10,11,33–36]. Generally, the CO adsorption energy is a widely used descriptor to assess the activity of catalysts for CO2RR instead of extensive thermodynamic and kinetic simulations, which greatly accelerates the efficiency of catalyst design [10,33]. After identifying the optimal active sites, the surfaces of catalysts could be rationally engineered to maximize the presence of such sites, in particular when combined with advanced ML technique [10,11,37–41].In this work, combining high-throughput density functional theory calculations and ML techniques, we have screened optimal active sites of 43 Cu-based SAAs on different Cu surfaces with total 2669 configurations for promising CO2RR catalysis activity, and developed a general model to understand critical factors influencing the catalytic properties. With proper feature sets selected, an ML model based on gradient boosting regression (GBR) was trained to predict CO adsorption energies on Cu-based SAAs with low root mean square error (RMSE). What's more, a cross-group learning scheme was adopted to extend the capacity of ML model to make accurate predictions of CO adsorption energies on Cu-based SAAs with alloying elements unseen to the training set. Finally, by considering key constraints on CO adsorption energy and selectivity over HER, ZnCu, AgCu, GaCu, GeCu and PCu SAAs are identified to exhibit highly active sites for CO2RR.We consider Cu facets with different miller indexes which exhibit distinct catalytic activities and product selectivity on CO2RR. Structurally, different surfaces can be characterized by a set of first-nearest-neighbor coordination number (CN) of surface atoms. Based on the analysis of radial distribution function for all considered Cu-based SAAs, the first and second peaks were identified at 2.57 and 3.63 Å, respectively, as shown in Fig. 1 a. Hereby, 3 Å was adopted as the cutoff radius for first-nearest neighbors. Accordingly, ten distinct Cu surfaces were constructed, with their first-nearest-neighbor CNs listed in Table S1. It can be seen that the first-nearest-neighbor CNs for ten surfaces are limited to the range from 6 to 10. Therefore, Cu(100), (111), (110), (210), and (411), covering all first-nearest-neighbor CNs (6–10), were selected as SAA host surfaces, as shown in Fig. 1b and f. By choosing 43 different elements as the alloying atoms, including 26 subgroup (d-block) elements and 17 main group ⅢA-ⅥA (p-block) elements (details in Fig. S1). 731 different Cu-based SAA surfaces were constructed in total.Different surface sites with the specific combination of local CNs and alloying element are then searched for optimal CO adsorption energies, which closely relates to high-performance catalytic properties of Cu-based SAAs. To generate a sufficiently large database of CO adsorption energies on Cu-based SAAs for ML, various adsorption sites around single alloying atoms were considered. Based on the adsorption energies and structural characteristics of optimized configurations, possible duplicates in the dataset were removed to avoid potential overfitting. Finally, 2669 CO adsorption configurations on specific sites were obtained from 3150 initial structures.To characterize the different catalytic sites of Cu-based SAAs, 12 features, including intrinsic elemental characteristics, such as alloying atom mass (m), alloying atom Bohr radii (R), Pauling electronegativity (χ), valence electron number (N v ), the number of p electron of p-block element or d electron of d-block element (N p/d ), electron affinity of the alloying atom (E ea ), ionization potential (IP), and energy level center of valence electrons of specific elements (E c ); structural characteristics, such as generalized CN of adsorption sites (GCN), the interplanar distance of Cu crystal surfaces (d hkl ), the distance between adsorption site and alloying atom (D sa ), and CO bond number on SAAs (CBN) were selected. It is worth noting that both (210) and (111) surfaces contain the first-nearest-neighbor CN of 9, while (411) surface has the same first-nearest-neighbor CNs of 8 and 7 with (100) and (110). To further distinguish the Cu-base SAA surfaces with the same first-nearest-neighbor CNs, GCN incorporating the information of every first-nearest-neighbor atom's CNs, which has been successfully applied in describing various electrocatalytic reactions on metal surfaces [42,43], is introduced as a more accurate description for local adsorption environment (Details in methods). Meanwhile, to characterize different CO adsorption configurations (on-top, bridge, hollow, etc.) on Cu-based SAAs, the bond number between CO and surface M atoms (M represents Cu or single alloying atom), i.e. CBNs, is adopted. The representation of active sites with different CBNs requires corresponding number of GCNs of surface atoms, constituting a vector. For instance, the adsorption on top of M (CBN = 1) corresponds to 1 GCN, while the bridge site adsorption of CO (CBN = 2) corresponds to 2 GCNs, and some hollow sites result in 3 or more GCNs. As a result, our GCN dataset exhibits variable dimensions. To address this problem, we have adopted a padding method, by which all GCN vectors are described with the same and maximum dimensions for all possible configurations, while those with smaller dimensions are appended by empty values set to zero. Similarly, D sa also has different dimensions, thus, the same padding procedure as in dealing with GCN is applied, but with the empty values set to −1. For GCN and D sa vectors, their components are arranged in ascending order, following by corresponding padding values. For instance, the smallest D sa , i.e. D sa1 , is the first component, and D sa1  = 0 corresponds to adsorption configurations with CO on alloying atom. By covering both surface structural information and chemical nature of catalysts, parts of these features have been successfully used to describe the binding energies of molecules and radicals on intermetallic and single-atom catalysts [44–46].Before ML model training, feature correlation was analyzed by Pearson correlation coefficient [47]. As shown in Fig. 2 a, N p/d , N V and E c show high linear correlation, implying that one of them could in principle replace another to reduce the feature set without losing key information. Thus, N p/d and E c were neglected for feature importance analysis.GBR, support vector regression (SVR), and random forest regression (RFR) were used as the regression algorithms to train the models for CO adsorption energies on Cu-based SAAs. In order to supervise the bias-variance tradeoff and prevent overfitting, 64%: 16%: 20% of the DFT database were randomly divided as the training, validation, and testing sets, respectively. These three sets were separately used for model training, selection of hyper-parameters of model and test of model generalization on unseen data. R 2 and RMSE were chosen to evaluate the accuracy of predicted energies. We used 10-fold cross-validation to select the best hyper-parameters. Afterwards, the model was tested on a test set which was split in advance to make sure that no severe overfitting was observed. With the same ten features of Cu-based SAAs, all three models exhibit high accuracies. The GBR model (testing set: R 2 = 0.956, RSME = 0.094 eV, Fig. S2a) is slightly better than SVR (testing set: R 2 = 0.910, RSME = 0.133 eV, Fig. S2b) and RFR (testing set: R 2 = 0.928, RSME = 0.120 eV, Fig. S2c). It should be mentioned that configurations with significant reconstruction due to large difference in radius between alloying atoms and Cu were removed from training. These systems include ACu (A = C or O) and a small portion of YCu and NCu SAAs (Supplementary information for details, Fig. S3). What's more, 63 non-adsorption configurations with positive adsorption energies were excluded (Supplementary information for details, Fig. S4). For all three ML models, though test error is higher than training error, they remain in a relatively low level, which are comparable with those reported previously [10,37,48]. Moreover, the model complexity was supervised as shown in Fig. S5. When the number of estimators was set to 50, an optimal bias-variance tradeoff of the model was reached. Based on GBR method, mean impact value (MIV) [49] analysis illustrates that the selected features have different importance levels in describing CO adsorption on Cu-based SAA surfaces, as shown in Fig. 2b. The top two important features, i.e., N v and GCN 1 , contain both the electronic and structural properties of alloying atoms, which indicates our selected feature group is rational for Cu-based SAAs. Based on the importance analysis above, features were optimized by removing the few least important ones, including CBN and high-order components of GCN and D sa , i.e. GCN 3 , GCN 4 , GCN 5 , D sa3 , D sa4 , D sa5 . The new model based on GBR with nine features maintains similar accuracy (R 2 = 0.953 and RMSE = 0.097 eV) as that learned from all features, as shown in Fig. 2c. In addition, predicting energy based on the data set extracted from the initial structures is significant and challenging. By using GBR method, the models based on the data sets from relaxed configurations and unrelaxed counterparts have been trained, respectively. It shows the model based on the data set from relaxed configurations (testing set: R 2 = 0.910, RMSE = 0.097, Fig. S6a) is more accurate in CO adsorption energy prediction than the latter (testing set: R 2 = 0.850, RMSE = 0.169 eV, Fig. S6b), which is expected. Obtaining the high-precision ML model based on the original structure without DFT calculations of the database is a key issue, which deserves a further study.A well-performed ML should not only be able to determine an effective regression method within the scope of training set and finally generate a high-precision model, but also have the capacity to predict the CO adsorption energies on Cu-based SAAs with unknown alloying elements, which is later referred as target, using the model trained on the source dataset. Herein, in order to test the model generalization between Cu-based SAAs with different alloying elements, the whole dataset is grouped into 39 subsets according to alloying element, excluding N, Y, C and O alloyed SAAs. Starting from initial source dataset of CO adsorption energies on one specific SAA (corresponding to one type of alloying element), and gradually expanding the source dataset by increasing the number of different types of SAAs (x), energies in target dataset consisting of CO adsorption energies on one SAA were predicted. However, the model doesn't show a trend towards increased accuracy with the increase of the number of alloying elements in the source dataset. The RMSEs as a function of the number of different types of SAAs in training group were irregular, as shown in Fig. S7. These results suggest that the model generalization is sensitive to data. This observation is further proved by predicting every single type of SAA using the model trained on all other 38 types of SAAs. As shown in Fig. S8, the generalization of our model to different types of SAAs is obviously distinct. For example, our model can predict the energies with RMSE smaller than 0.1 eV for SAAs including As, Cd, Ga, Ge, Hf, In, P, Sb and Sn, while for elements like B, Al, Fe, Rh, Sc, W, it shows RMSE larger than 0.25 eV.In order to understand the underlying cause, the nine-dimensional features were reduced to a two-dimensional space by Isomap [50], as displayed in Fig. S9. It shows that the distribution of features for different SAAs exhibits a consistent pattern in the feature space. Accordingly, from a probabilistic point of view, the marginal probability distribution of both source and target, P(X) with X representing feature, could be regarded as the same. Although the ML model trained on source set couldn't be directly used to make accurate predictions on target SAAs with the alloying elements not included in the source, the information contained in the source is believed to be beneficial and could still be utilized. Now referring to the learning scheme as ‘cross-group’ learning: re-fit the former trained model by including 10 additional instances from target set. As is shown in Fig. 2d, after re-fitting the model on the additional 10 instances, the RMSEs of predicted CO adsorption energies on target SAAs decrease notably to 0.2 eV or smaller. The results indicate excellent flexibility of our ML model for predicting CO adsorption energies on SAAs with new elements. Accurate predictions could be guaranteed with a supplement of at most 10 instances into the source training set, which implies that in this case over 90% of computation cost could be saved for new SAAs by cross-group learning scheme.In order to identify specific active sites for CO2RR, 2669 sites on Cu-based SAAs surfaces involving 43 alloying elements were systemically studied. Following the Sabatier principle, too weak binding of the intermediates on the surface leads to their quick desorption and thus weak charge transfer between intermediates and catalysts, while too strong binding results in catalyst poisoning. Accordingly, optimal adsorption energies are required to maximize catalytic activity. For CO2RR, CO is involved in many key reaction steps, and its adsorption energy on catalysts has been proved as an effective descriptor with the optimal CO adsorption energy of ∼0.67 eV using revised Perdew–Burke–Ernzerhof (PBE) functional [10,33]. The distribution of calculated CO adsorption energies on Cu-based SAAs was analyzed for p- and d-block alloying elements, respectively. Fig. 3 a shows a dominant single-peak distribution of adsorption energies on Cu-based p-block-element SAAs, and 35.5% of adsorption energies are within 0.1 eV of the optimal value of −0.67 eV. In contrast, the CO adsorption energies on Cu-based d-block-element SAAs are more diversely distributed as shown in Fig. 3b. Besides 33.0% optimal CO adsorption sites, d-block-element SAAs generally show stronger binding for CO with broader adsorption energy distribution in the range of −2.30 eV < ΔECO < −0.77 eV. The wide distribution of adsorption energies in d-block-element SAAs suggests great tunability of alloying different d-block elements for different catalytic reactions. The enhanced binding strength results from stronger charge transfer and hybridization between Lewis acid CO and transition metals.Based on the feature importance analysis in Fig. 2b, the five most important features of Cu-based SAAs were selected for further discussion, including N v , GCN, D sa , IP, and R. Compared to N v associated with atomic characteristics, the transferred charge (Q t ) of alloying atom can better reflect the bonding properties and the key effects on surrounding Cu. Therefore, the charge transfer between various alloying atoms and Cu is explored to understand the fundamental mechanism. It can be found that Q t corresponding to sites within optimal CO adsorption energy range distributes widely in every segment, as shown in dark purple and dark brown in Fig. 3c and d. Interestingly, both the negative and positive Q t could give rise to optimal adsorption energies. It suggests that charge transfer between alloying atoms and surrounding Cu atoms could activate surface sites, no matter whether it leads to positively or negatively charged active sites. Differently, previous reports showed that usually the negative Q t induced modulation of oxidation state in Cu2O, Cu2S or Cu through doping and alloying can improve CO2RR activity and selectivity [17,51–54]. Here, many alloying atoms of optimal sites exhibit positive Q t values, such as Si, Ga, Ge, In, Sn, Sb, Bi in p-block-element SAAs, and Cr, Mn, Fe, Co, Ni, Zn, Mo, Tc, Cd, W in d-block-element SAAs, indicating possible different mechanisms. The distribution of N v , R, and IP related to the sites with optimal energies were nearly uniform (Fig. S10).Compared to elemental and electronic properties, surface structural features exhibit different distribution patterns. For GCN 1 , the range of [5,6] is the optimal coordination environment for both p-block and d-block elements as displayed in Fig. 3e and f, indicating under-coordinated sites on Cu-based SAAs are highly active. Comparing the distribution of optimal adsorption energies on different surfaces of SAAs shown in Fig. S11, the high-index surfaces, (210) and (411), have more active unsaturated sites than those of (100), (111) and (110), showing higher catalytic activity for CO2RR. Furthermore, Dsa reflects indirectly the effects of alloying atoms for CO adsorption on nearby Cu. Generally, the larger the D sa , the weaker the effects of alloying atoms on CO adsorption. Fig. 3g and h show the partitions of D sa1 according to adsorption sites on p- and d-block-element SAAs, respectively, where 0, ( 0 , 3.5 ] , and ( 3.5 , 4.3 ] Å correspond to cases with CO bonding to alloying atoms, first-nearest neighbor, and second-nearest neighbors of alloying atoms, respectively. For p-block-element alloying, CO mainly bonds to the first-nearest-neighbor Cu atoms (70.55%), which is attributed to the stronger electronegativity of p-block elements compared with Cu, leading to the modulation of Cu oxidation state. Differently, for d-block-element alloying, besides first-nearest-neighbor Cu atoms (54.79%), CO bonding to alloying atoms also takes an important part (34.67%). The strong tendency to top-site adsorption originates from the strengthened adsorption as illustrated in Fig. 3b. Notably, active sites with optimal CO adsorption energies on both p-block- and d-block-element SAAs are mainly the first-nearest neighbors, as shown in Fig. 3i and j.The above analysis indicates that the modulation of either oxidation or reduction states on Cu atoms could improve the catalytic activity for CO2RR. To understand these effects, we further investigate the electronic structures of representative CO adsorption configurations with three different charge states, including AsCu(411), ZnCu(411), and PdCu(111). Q t of As in AsCu(411) and Zn in ZnCu(411) are −0.32 and + 0.22 | e | , respectively. In these two cases, CO prefers to adsorb on the first-nearest-neighbor Cu sites. From the partial density of states (PDOS) plot in Fig. 4 a and b, both the increased occupation of Cu 3d states in the range of [−7.8, −6.3] eV (indicated by red arrows) and their enhanced hybridization with C 2p states contribute to the improved adsorption strength compared with pure Cu cases. While for PdCu(111), Q t is −0.39 | e | . The strong CO adsorption strength results from effective overlap between Pd d states and C 2p orbitals, as shown in Fig. 4c. This could be the main reason for more negative adsorption energies in d-block-element SAAs. CO adsorption energies on PdCu (111), AsCu (411) and ZnCu (411) SAAs are 0.26, 0.07 and 0.03 eV lower than that of Cu counterpart, respectively (Table S2). In addition, the charge density difference and Bader charge analysis show that CO on these three different surfaces keeps as an electron acceptor. The interaction between CO molecule and local charged atoms were further explored by analyzing the projected density of states of the catalysts before and after CO adsorption, shown in Fig. S12. For all considered examples, the bonding d-σ and antibonding d-π∗ orbitals form after CO adsorption. With Fermi level crossing d-π∗ orbital, the excess electrons from surface atoms d orbitals back to antibonding π∗ orbital help stabilize the adsorption configurations. Therefore, the increased charges in CO can be used to roughly qualitatively estimate the contribution of the back-donating effect. Table S3 shows that |ΔQ(CO)| for ZnCu and PdCu were larger than those of Cu counterparts, consistent with the enhanced CO adsorption. It is noted the excess electrons on CO was transferred from negative Cu around positive Zn or negative Pd, highlighting the importance of the negative charged metal atom center for enhancing CO2RR activity, which has been proposed for single-atom catalysts for oxygen evolution reaction [55]. In contrast, for AsCu, CO adsorbs on positive Cu site around As, leading to decreased |ΔQ(CO)|. In this case, the slightly increased CO adsorption energy might result from the enhanced d-σ interaction. Furthermore, the charge transfer between Cu and alloying atoms could render CO molecules on alloying atoms and adjacent Cu oppositely charged, inducing enhanced Coulomb attraction interaction between them which promotes the activity and selectivity of C2 products [54].The active sites on Cu-based SAAs can then be screened against the competing hydrogen evolution reaction (HER). For HER, H adsorption energies ΔEH, the key indicator for HER activity, were calculated for the screened configurations with optimal CO adsorption strength, with the screening criteria ΔEH ∉ [−0.37,-0.17] eV [10]. Figs. S13a and S13b are two-dimensional distribution of CO adsorption energies versus adsorption configuration and element on p- and d-block-element SAA surfaces within (−0.77, −0.57) eV, respectively. Out of total 1102 and 1567 CO adsorption energies on p-block- and d-block-element SAAs, there are 391 and 517 sites showing optimal values, respectively. The proportion of optimal adsorption sites to total sites for a specific Cu-based SAAs could reflect its activity, which gives rise to the following sequences: P > Ga > As > Ge > Sb > Sn > In > Si for p-block-element SAAs, and Pd > Cd > Ag > Au > W > Zn > Tc > Re > Mo > Pt > Ru for d-block-element SAAs. The optimal CO adsorption sites were then screened by H adsorption energy out of [−0.37, −0.17] eV, as shown in Figs. S14 and S15. Consequently, target SAAs with promising high-activity and selectivity over HER include PCu, AgCu, GaCu, SbCu, ZnCu, SnCu, GeCu, InCu, WCu, MoCu and SiCu. Here, due to the high cost or toxicity, Pd, Au, Pt, Cd, Ru, Re, Tc and As were excluded, despite their promising CO2RR activity.In order to check the feasibility of Cu-based SAAs in the experiments, their thermodynamic stabilities were first estimated by formation energies shown in Fig. S16. In addition, we also compared the energy differences between SAAs and dopant dimmer configurations with the same concentrate for one of the most promising surfaces, Cu (411), as listed in Table S4. ΔE is calculated as ΔE = ESAA - EDimer, where ESAA and EDimer are the energies of SAAs and the dopant dimmer counterparts with the same concentrate. It shows that the energies of MoCu and WCu with dopant dimmers are much lower than those of SAAs, indicating Mo or W atoms in Cu based SAAs tend to form clusters. For the candidates with a small energy difference (0 < ΔE < 0.2 eV), Ab initio molecular dynamics (AIMD) simulations were further performed at 300K for 10 ps. Our results show that Sb atoms tend to adsorb on the surface rather than doping, as shown in Fig. S17. After excluding SAAs with too high energy (MoCu and WCu) and poor thermal stability (SbCu), eight promising catalyst materials were screened out, including PCu, AgCu, GaCu, ZnCu, SnCu, GeCu, InCu, and SiCu. Experimentally, dilute Zn, Ga, Ag, and Pt alloyed Cu have been synthesized [53,56,57]. Notably, some of the predicted results have been confirmed by recent CO2RR experiments, verifying the reliability of our work. For instance, CuZn nanoparticles with low Zn concentration showed enhanced CH4 Faraday efficiency compared to pure Cu [53]. CuAg with the surface radio 7:1 demonstrated a higher activity for yielding a single liquid product, acetaldehyde, in contrast to pure Cu which forms seven different products at the same potential [57]. Following these reports, the recommended synthesis routes include, but not limited to, surface replacement reaction in the solution and selective dealloying of bimetal materials [58].Based on the analysis of feature distributions, the low-coordination high-index crystal surfaces exhibit more active sites than those of low-index surfaces. Therefore, Cu (411) surface was selected to evaluate the limiting potential of SAAs for CO2 reduction to CO by considering reaction pathways R1 and R2 as following: (R1) ∗ + CO2 + (H+ + e−) → ∗COOH (R2) ∗COOH + (H+ + e−) → ∗CO + H2O Based on the implicit solvent models, the detailed Gibbs free energies diagrams for the reduction of CO2 to intermediate product ∗CO on eight suggested SAAs are calculated in Fig. 5 a. Compared with pure Cu(411), all SAA surfaces show enhanced ∗COOH and ∗CO binding, leading to improved catalytic performance. In particular, the limiting potential decreases from 1.53 V for pure Cu (411) surface to 1.21 and 1.18 V for Si and P SAAs. Moreover, their selectivity over HER was also explored by calculating the thermodynamic diagram shown in Fig. 5b. Here, the ∗H adsorption models were constructed according to most stable ∗CO configurations. The obtained results in Fig. 5b show that all |G(∗H)| on the selected SAAs were larger than that on pure Cu (411), indicating alloying further improves the selectivity of SAAs over HER.The active sites on Cu-based SAAs catalysts of CO2RR have been extensively explored based on 2669 CO adsorption configurations by DFT calculations. A simple ML model with high accuracy was generated for predicting CO adsorption energies. Our trained ML model is further applied to accurately predict CO adsorption energies on Cu-based SAAs with alloying element unseen to the training set by a cross-group learning scheme. Extensive ML analysis demonstrates valence electron number of alloy atoms and GCN are two key features to characterize the optimal active sites of Cu-based SAAs for CO2RR. By screening of optimal CO adsorption sites with high selectivity over HER, PCu, AgCu, GaCu, ZnCu, SnCu, GeCu, InCu, and SiCu SAAs, were recognized as promising catalysts for CO2RR. This work provides a feasible strategy to design Cu-based and other SAAs catalysts materials for improved CO2RR.Five Cu miller–index facets with different coordination characteristics, including (100), (111), (110), (210), and (411), were constructed. Because catalysis mainly occurs near the surface, substitutional sites for alloying atoms were located in the surface and subsurface with specific coordination numbers. In this sense, our SAA models can be viewed as single atom surface alloy, a simplified model of the SAAs. The surface models were constructed with each cell vector of surfaces at least 7.2 Å, and in c direction at least four layers of Cu were chosen. For the description of geometric structure, GCN of a specific site x is introduced as G C N ( x ) = ∑ y = 1 n x C N ( y ) n y C N m a x − 1 , where CN of first-nearest-neighbor atom y is weighted by n y C N m a x − 1 , with ny and CNmax representing the number of first-nearest-neighbor atoms y with the same CN(y) and the maximum CN in the bulk, respectively. For FCC crystals, CNmax is 12 [42,43].Radial distribution function, g(r), was used for atomic pairwise distance analysis, in order to shed light on the reasonable selection of cutoff distance for first nearest neighbors. By dividing the space volume into shells dr, it is possible to compute the number of atoms dn(r) at a distance between r and r + dr from a given atom: d n ( r ) = N V · g ( r ) · 4 π r 2 · d r where N and V represent total number of atoms and the whole volume of the system.First-principles calculations were performed within the framework of DFT, as implemented in the Vienna ab initio simulation package [59]. Projector-augmented wave potential was used with a plane-wave cutoff energy of 400 eV [60]. Exchange and correlation were described by the generalized gradient approximation in the scheme of revised PBE [61,62]. Spin-polarized calculations were also performed for magnetic systems. The Brillouin-zone integration was done using a Monkhorst–Pack grid of k-point sampling [63]. K-spacing was set to 0.3 for all structures to allow the smallest spacing between k-points in units of 0.3 Å−1. A vacuum distance larger than 20 Å was employed to avoid interactions between neighboring images, and the bottom two layers were fixed during relaxation. Adopting the conjugated gradient method, geometrical optimizations were carried out with the convergence threshold set at 1 × 10−4 eV atom−1 in energy and 0.05 eV Å−1 in force. AIMD simulations were performed at 300 K, in an NVT ensemble using the Nos e ′ -Hoover heat method for 10 ps with a time step of 1.0 fs. CO and H adsorption energies were calculated from Δ E = E a d − E S A A − E C O / H , where Ead and ESAA are the energies of SAA surfaces with and without adsorbates (CO/H), ECO and EH are the energies of H2 and CO, respectively. The formation energies of SAAs was calculated by E f = E S A A − E C u − μ A + μ C u , where ESAA and ECu are the energies of SAA surfaces and Cu counterpart, μ A and μ C u are the chemical potentials of alloying element and Cu.Gradient boosting regression (GBR), support vector regression (SVR), and random forest regression (RFR) were carried out using scikit-learn package [64]. The correlation of features was evaluated by Pearson correlation coefficient (p) [47], which is determined by p = ∑ i ( f i − f ¯ ) ( F i − F ¯ ) ∑ i ( f i − f ¯ ) 2 ∑ i ( F i − F ¯ ) 2 Here, p is correlation coefficient between feature f i and F i within the range of [–1,1]. The feature importance was analyzed by MIV, determined as [49]. m ( i ) = y ˆ 1.1 ( i ) − y ˆ 0.9 ( i ) ∑ j = 1 n ( y ˆ 1.1 ( j ) − y ˆ 0.9 ( j ) ) Here, y ˆ 1.1 ( i ) and y ˆ 0.9 ( i ) represent the prediction values of the model after the corresponding feature is multiplied by 1.1 and 0.9, respectively. The summation loops over all n features. Intuitively, the analysis depicts the sensitivity of the model prediction to a certain feature. If the prediction varies dramatically as the certain feature fluctuates, the corresponding feature is considered to be important.The accuracy of models was identified by RMSE and the coefficient of determination (R2), which are defined as R M S E = 1 N ∑ i n ( Y i − y ˆ i ) 2 R 2 = 1 − ∑ i ( Y i − y ˆ i ) 2 ∑ i ( Y i − Y ¯ ) 2 Y i and y ˆ i are the adsorption energies by DFT calculations and model prediction, respectively, and Y ¯ is the average energy of DFT calculations.Isomap, one representative of isometric mapping methods, is used for nonlinear dimensionality reduction. It seeks a lower-dimensional embedding which maintains geodesic distances between all points [50].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 (Grant Nos. 62006219 and 62001266) Guangdong Innovative and Entrepre-neurial Research Team Program (grant No. 2017ZT07C341), the Bureau of Industry and Information Technology of Shenzhen for the 2017 Graphene Manufacturing Innovation Center Project (No. 201901171523), the China Postdoctoral Science Foundation (No. 2020M680506) and Guangdong Basic and Applied Basic Research Foundation (No. 2020A1515110338). D. Wang and R. Cao contributed equally.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.2021.10.003.

Various strategies, including controls of morphology, oxidation state, defect, and doping, have been developed to improve the performance of Cu-based catalysts for CO2 reduction reaction (CO2RR), generating a large amount of data. However, a unified understanding of underlying mechanism for further optimization is still lacking. In this work, combining first-principles calculations and machine learning (ML) techniques, we elucidate critical factors influencing the catalytic properties, taking Cu-based single atom alloys (SAAs) as examples. Our method relies on high-throughput calculations of 2669 CO adsorption configurations on 43 types of Cu-based SAAs with various surfaces. Extensive ML analyses reveal that low generalized coordination numbers and valence electron number are key features to determine catalytic performance. Applying our ML model with cross-group learning scheme, we demonstrate the model generalizes well between Cu-based SAAs with different alloying elements. Further, electronic structure calculations suggest surface negative center could enhance CO adsorption by back donating electrons to antibonding orbitals of CO. Finally, several SAAs, including PCu, AgCu, GaCu, ZnCu, SnCu, GeCu, InCu, and SiCu, are identified as promising CO2RR catalysts. Our work provides a paradigm for the rational design and fast screening of SAAs for various electrocatalytic reactions.

Magnetic ferrite nanoparticles in the spinel phase have been considered as an influential class of materials, which are employed in various high-frequency device applications [1]. The cubic spinel structure has the chemical formula of AB2X4, where the anions X are occupied by O atom as metal oxides forming the cubic close-packed lattice, tetrahedral interstices fill the A site as the ‘network formers’ and octahedral interstices occupy the B site as the ‘modifiers’, called the Ferro-spinel and semiconductor in nature [2–5]. Most spinel ferrites belong to the space group of Fd3m (No. 227, Z = 8), which provide the highest symmetrical face-centered cubic (FCC) spinel structure. A spinel unit supercell's crystal is formed by 8 A-sites and 16 B-sites cations. Based on the distribution of divalent metal ions and trivalent ferric ions over A and B-sites, spinel ferrites are of three classes; normal spinel, inverse spinel, and mixed spinel [6].Magnetically soft spinel ferrites are used in a large spectrum of biomedical and industrial applications including medical treatments, magnetic resonance imaging, antenna fabrication, computer memories, energy storage in supercapacitors, high-density information storage, high-frequency transformers, hyperthermia treatment, multi-layered chip inductors, water purification methods, sensing of nucleic acid, separation of DNA and RNA, gene therapy and delivery, ferrofluids and so on [7–13]. The advancement of electronic devices is now moved to integrated circuits-based technologies, where the use of highly efficient transistors is increasing gradually in accordance with the Moore’s law, thus requiring the nano-level engineering and fabrication. Therefore, in contrast to bulk materials, researchers are now focusing on nanocrystalline ferrites’ for utilizing them in the latest nano-technological devices. The physical and chemical characteristics of ferrite nanomaterials mostly depend on their scale size, shape, or morphology. The structural parameters such as crystal size and lattice parameters are somehow linked to the electrical and magnetic properties of ferrite nanoparticles. Therefore, the controlling of several factors such as the particle size, surface-to-volume ratio, and magnetic anisotropy eventually improves the electronic properties of magnetic nanoparticles in the spinel phase, owing to their transition from bulk to nano-shape.Researchers are continuously paying efforts to employ an easy and efficient method for yielding the nanocrystalline ferrites to tailor their structural, dielectric, electric, and magnetic properties under favorable environmental conditions. Various techniques have been used to synthesize nanostructured ferrite materials such as the sol–gel auto combustion, co-precipitation, high-energy milling, hydrothermal synthesis, precursor method, mechano-chemical route, and microwave hydrothermal [7–9,14–17]. Among these, the sol–gel route appears to be a prominent method for preparing highly crystalline ferrite nanoparticles, as it is an eco-friendly, less expensive, and effective approach to maintaining a good stoichiometry during the synthesis process. The sol–gel is a wet chemical method, which is widely used due to its potential advantages such as enhanced control over homogeneity, elemental composition, and powder morphology with a uniform narrow particle size distribution at relatively low temperature [7,16–18].Researchers previously attempted sporadically to study the effects of doping on the structural, electrical, morphological, photocatalytic, and magneto-dielectric properties of Ni-Cu ferrite NPs. Doping is an effective method to ameliorate the applications to a broad range by achieving excellent optoelectronic properties. Investigations are still continued with various dopants or substitutions in A and B-sites to tailor the physical, structural and electromagnetic properties of Ni-based ferrite nanoparticles [19–28]. Munir et al. [29] conducted an experiment with a noble nanocomposite CuFe2O4/Bi2O3 by introducing Bi2O3 nano-petals into the porous CuFe2O4 and observed a significant increase in the photocatalytic activity in the effect of photo-degradation activity. The investigated nanocomposite showed an excellent magnetic separation at room temperature for the reduced recombination and improved separation of electron-hole pairs. Carbon coated highly active magnetically recyclable hollow nanocatalysts were synthesized by Shokouhimehr et al. [30], where the authors projected that the prepared nanocomposite can be used as a general platform for loading other noble metal catalyst nanoparticles, resulting in the high yields (up to 99 percent) in selective nitroarenes reduction and Suzuki cross-coupling reactions. Furthermore, magnetic properties revealed that the catalysts could be easily separated using a suitable magnetic field and recycled five times in a row. Moreover, Rahman et al. [31] thoroughly investigated the photocatalytic efficiency and recycling stability of rGO-supported cerium substituted nickel ferrite nanoparticles under visible light illumination. According to their findings, NiCeyFe2-yO4/rGO (NCFOG) nanocomposite outperformed NiCeyFe2-yO4 nanoparticles by two times in photocatalytic efficiency and recycling stability, which is attributed to the formation of NCFOG heterojunction that enables the separation of photo-induced charge carriers with maintaining a strong redox ability. Recently, M. Arifuzzaman et al. [17] studied Cu substituted Ni-Cd ferrite NPs and reported the decrease of average crystallite size and saturation magnetization of Ni0.7-xCuxCd0.3Fe2O4 with Cu-substitution up to x = 0.2. Besides, V. A. Bharati et al. [32] reported the influence of doping of both Al3+ and Cr3+ on the structural, morphological, magnetic, and MÖssbauer properties of Ni ferrite NPs and justified their suitability in HF device applications.In [33], K. Bashir et al. revealed the electrical and dielectric properties of Ni-Cu ferrite NPs with doping of Cr3+, which made them the potential for HF applications and photocatalytic activity. Le-Zhong Li et al. [34] examined the Al3+ substituted Ni-Zn-Co ferrites and observed a decrease in saturation magnetization at x > 0.10. They reported the metal–semiconductor transition behavior of Ni-Zn-Co ferrites as an effect of varying temperature and the increase of dc resistivity with Al content. The structural and magneto-optical properties of Ni ferrite NPs were reported in [35], where the authors calculated the electronic bandgap of 1.5 eV and observed a decrease in saturation magnetization and Tc with Al3+ content. In addition, a density functional theory (DFT) was used in estimating the electronic structure of CuO NPs with the optimized geometric crystal calculation, which showed the variation of energy band gap with Al content [36]. The effect of doping materials on the characteristics of different spinel ferrite nanoparticles is available in the literature, Zn ferrite [37,38], Ga ferrite [39], Co ferrite [40–42], Fe ferrite [43], Mg ferrite [25,44], and Ni-Zn ferrite [45,46].However, as per the literature survey, no study has been found yet on the structural, dielectric, and electrical properties of Al3+ substituted nanocrystalline Ni-Cu spinel ferrites. Therefore, it is important to perceive the role of Al3+ substitution on the structural, physical, and dielectric characteristics of Ni-Cu ferrite NPs. Henceforth, the present study aims to explore the influence of Al3+ substitution on the structural, dielectric, and electrical properties of the synthesized Ni0.70Cu0.30AlxFe2-xO4 (x = 0.00 to 0.10 with a step of 0.02) through the sol–gel process.To synthesize the studied nanocrystalline Ni-Cu spinel ferrites, analytical-grade reagents of nickel (II) nitrate Ni(NO3)2·6H2O (98%), copper (II) nitrate Cu(NO3)2·3H2O (95–103%), ferric (III) nitrate Fe(NO3)3·9H2O (98%), and aluminum (III) nitrate Al(NO3)3·9H2O (98%) were used in this experiment, which were purchased from the Research-Lab Fine Chem.Derivatives of Ni0.70Cu0.30AlxFe2-xO4 (0 ≤ x ≤ 0.1) nanoparticles with a step of 0.02 were synthesized by the sol–gel process. In this process, raw materials of Ni(NO3)2·6H2O, Cu(NO3)2·3H2O, Fe(NO3)3·9H2O, and Al(NO3)3·9H2O dissolved in ethanol and mixed properly with a magnetic stirrer to make the homogeneous solution. The pH of the mixture was kept at 7 using the liquid NH4OH solution and the sol was continued to heat up to a temperature of 70 °C and stopped when it was turned into a dry gel. In an electric oven, the dried gel was heated at 200 °C for 5 h, during which a self-ignition occurred and the compositions gradually became fluffy-loose powder. To obtain the resulting ingredients in a highly crystalline form, the derived powder was further annealed at 700 °C for another 5 h to eliminate any impurity in the samples. The powder was then grinded in a hand-milling process in a mortar to make it more homogeneous. A hydraulic press of 65 MPa was then applied to the samples for 2 min to condense and turned into disk shapes.The prepared samples were 12 mm in diameter and 2.3 mm in thickness. Powder samples were finally used for dielectric and electrical measurements.The structural parameters of the yielded nanocrystalline ferrites were determinedthrough the powder x-ray diffractometer (XRD) analysis using the model PW3040, with CuKα radiation of λ = 1.5418 Å. The lattice parameter, crystal size (D), and displacement density were retrieved by using the XRD data. The theoretical density (ρth), micro-strain (εms), lattice strain (εls), and stacking faults in the crystal structure were also determined. The lattice parameter ( a ) and crystallite size (D) were measured by the following relations [20]: (1) a = d hkl h 2 + k 2 + l 2 (2) D = 0.9 λ β hkl c o s θ where, λ, βhkl, θ, and dhkl, respectively, indicate the wavelength of the X-ray, the full width at half maximum (FWHM) at the most prominent peak (311), the Bragg’s angle, and the distance between adjacent planes. The cubic spinel phase of the prepared samples was reconfirmed through Fourier transform infrared spectroscopy (FTIR). The morphology of the studied materials has been interrogated by the Field Emission Scanning Electron Microscopy(FESEM) (JEOL-JSM 7600F model).The Wynne- Kerr Impedance Analyzer (model: 6500B) was used to determine the complex dielectric (ε*), complex electric modulus (M*), complex impedance (Z*), and AC resistivity (ρAC) of the investigated Ni-Cu-Al ferrite nanoparticles.XRD patterns of Al3+ substituted Ni-Cu ferrites annealed at 700 °C, are illustrated in. Fig. 1 , where the peaks are resulted due to diffractions from the planes of (111), (220), (311), (222), (400), (422), (511), and (440). The peaks are found well-defined with a homogeneous distribution of nanoparticles, which verify their high crystalline nature with no impurity. The existence of such peaks indicates the cubic single-phase formation of the spinel materials [32–34]. The peak diffracted from the plane (311) is found to have a high intensity, which was used to determine the average crystallite size (see Table 1 ) of the materials using Debye-Scherer’s equation. The lattice constant (a0 ) values of the samples are calculated by the Nelson-Riley technique and unit cell volumes (V) of the compositions are listed in Table 1. A decreasing trend in the variation of lattice constant and cell volume with Al3+ content is observed as shown in Fig. 2 , which is due to the replacement of larger ionic (0.67 Å) cations by cations with smaller radius (0.51 Å). As Fe3+ is replaced by Al3+ in the investigated ferrites, the unit cell becomes shrinkage, as a result, both a 0 and V decrease linearly with Al3+ content, which is well satisfied by the Vegard's law [47,48]. As appeared in Table 1 and Fig. 2, the average crystallite size decreases with Al3+ content, which might be because of the ionic radius difference between Al3+ and Fe3+. The redistribution of cations in A and B sites ultimately causes the increase in stress and strain of the samples. Lattice spacing is determined by the following equation: (3) d = n λ 2 s i n θ where d is the inter-spacing distance between crystal planes and the value of n is taken as 1, which represents the order of diffraction.The sharp diffraction peaks from XRD confirm the higher crystallinity of the analyzed ferrite nanoparticles. The percentage crystallinity of the prepared nanoparticles is measured by the following equation [49,50]: (4) % C r y s t a l l i n i t y = A r e a u n d e r t h e c r y s t a l l i n e p e a k s A r e a o f t h e a l l p e a k s × 100 The theoretical density (ρth ) is calculated by the following relation [51]: (5) ρ th = 8 M w N a a 0 3 where Mw and Na indicate the molecular weight of the samples and Avogadro’s number, respectively. The experimental density (ρex ) is calculated by the following equation: (6) ρ ex = M π r 2 l where M, r, and l represent the mass, radius, and height of the synthesized samples in tabloid shape, respectively.The experimental density and estimated density of the samples annealed at 700 °C are listed in Table 1. The porosity is found to increase as presented in Table 1, which is due to the discontinuity of the grain size, resulting in the decrease in density. The percentage porosity of the samples is calculated by the following relation: (7) P ( % ) = ρ th - ρ ex ρ th × 100 % The porosity is the contribution of inter-granular and intra-granular as shown by the following equation: (8) P % = P inter + P intra The total displacement length per unit volume of the crystal structure is referred as the dislocation density (δ) and the way to reduce it to annealing the samples at high temperatures, which in turn increases their grain size [36]. This annealing is also considered the regulator of the strength and flexibility of the crystal structure. The visible parallel lines and random lines in the crystal may indicate the displacements, which means these lines may result due to the displacement. Displacement density and particle size follow an inverse relationship with giving an error called the linearity error. The dislocation density is calculated by the following equation: (9) δ = 1 D 2 The length due to the deformation of an object is closely related to the pressure applied, known as the lattice strain (εls). The defects caused by imperfections in the crystal structure compel atoms to deviate slightly from their normal position [52]. These structural flaws include interstitial and/or impurity atoms that cause lattice strain, which can be determined by the following relation: (10) ε ls = β 4 t a n θ where θ represents the angle of diffraction and β indicates the full width at half maximum. The stacking faults are induced by the atomic planes of the crystal due to interruption of the layered arrangement in a normal lattice structure. The stacking fault [SF] is determined by the following equation: (11) S F = 2 π 2 45 √ ( 3 t a n θ ) Various defects in the crystal structure such as displacement, plastic deformation, point defects, and domain boundary defects are considered as key factors of deformation in the structure. The deformation is assumed to occur in one per million parts of the materials, which is defined as the micro strain (εms). A notable feature of the micro strain is that it maximizes the peak and the following equations are introduced to comprehend it [53]: (12) ε ms = β c o s θ 4 The ionic radii of A and B sublattices are calculated by the following relations [9,54]: (13) r A = 3 a 0 ( u - 0.25 ) - r o (14) r B = a 0 ( 0.625 - u ) - r o where ro and u represent the radius of oxygen (1.32 Å) and oxygen parameter having the value of 3 8 , respectively. The distance between the centers of adjacent ions is the hoping length, whose values for A-A sites, B-B sites, and A-B sites are calculated by using the following equations [54,55]: (15) L A - A = a o 3 4 (16) L A - B = a o 11 8 (17) L B - B = a o 2 2 where a₀ represents the lattice constant.To confirm the structure of the spinel phase in all prepared samples, Fourier transform infrared (FTIR) spectroscopy technique was utilized. Fig. 3 depicts the FTIR spectra of nanocrystalline Ni0.70Cu0.30AlxFe2-xO4, measured over the frequency region of 450–4000 cm−1. Two fundamental strong absorption bands v1 and v2 are observed in the effect of the metal–oxygen (M−O) bonds at the tetrahedral and octahedral sites. The entity of the high-frequency v1 band is found in the range of 585–615 cm−1, which is formed by the internal stretching of the M−O bond at tetrahedral sites whereas the low-frequency v2 band appears at around 400 cm−1, which corresponds to that of octahedral site [56]. The formation of spinel structures in the prepared Ni-Cu ferrite nanoparticles is ascertained by the observed bands. The bands observed in this investigation are consistent with previous findings. [57,58]. The absorption peaks, however, are induced by the tetrahedral site of the metal's intrinsic stretching vibration. Moreover, the stretching vibration of both sites is influenced due to changes in the lattice parameter. The tetrahedral stretching frequency band (v1) is found to shift towards the higher frequency region with the Al3+ substitution increases. The cause of the band shifting with Al3+ content as illustrated in Fig. 3, might include the fact of cations redistribution over the tetrahedral and octahedral sites [56,58,59]. Fig. 4 shows the FESEM micrographs of nanocrystalline Ni0.70Cu0.30AlxFe2-xO4 annealed at 700 °C. As depicted in Fig. 4(A-F), the grains are formed of semi-spherical shapes with a uniform and even distribution in multi-domains separated by grain boundaries. The average grain size of the synthesized ferrite nanoparticles is measured by [54]: (18) G a = 1.5 L XN where L, X, respectively, indicate the total length in cm and the magnification of the micrographs, and N is the number of intercepts. Fig. 5 illustrates the EDX analysis of Ni0.70Cu0.30AlxFe2-xO4, which ensures the presence of all expected elements in each sample with appropriate proportions. The sum of elements in each composition provides 100%, which confirms the accuracy of the sol–gel synthesize technique and manifests its novelty. Fig. 6 (A, B) demonstrates the variation in real (ε′) and imaginary (ε′′) parts of the complex dielectric constant of nanocrystalline Ni0.70Cu0.30AlxFe2-xO4 annealed at 700 °C with varying frequency. The dielectric property of ferrites is contingent to different factors such as preparation method, chemical composition, grain size, electronic di-polarity, and so on. The ε′, ε′′, and dielectric loss tangent (tan δE ) are calculated by the following relations: (19) ε ′ = Ct ε o A (20) ε ″ = ε ′ t a n δ E and (21) tan δ E = 1 ω ε o ε ′ ρ where C is the capacitance, ω = 2πf with representing f as the applied field frequency, εo is the free-space permittivity, t is the thickness and A is the area of the contact surface of the tabloids. Fig. 6(A) illustrates that ε′ decreases with frequency up to 105 Hz and thereafter remains almost constant with showing a very low value. On the contrary, the imaginary part (ε′′) reveals higher values at low frequency regime and decreases vigorously with frequency as observed in Fig. 6(B). The observed dielectric dispersion of the investigated materials can be described by the Maxwell–Wagner interfacial theory of polarization supported by Koop’s phenomenological theory [54,55,60]. The grain boundaries are more active at low frequencies, whereas at high frequencies grains are more contributing. At the low frequency regime, the value of ε′ is higher because of the high resistive grains attributed to the space charge polarization [47,48]. The decreasing trend of the real dielectric constant (ε′) with frequency is found in Fig. 6A. This happens as the grains come into action at higher frequencies and the hopping electrons cannot follow the applied electric field after a certain frequency, which causes the polarization to be decreased. As a result, the value of ε ′ appears to be very low, becoming almost constant [61].The sample with x  = 0.1 shows the maximum value of ε ′ because of the redistribution of Fe3+ at both A- and B-sites in Ni0.70Cu0.30AlxFe2-xO4. The substitution of Fe3+ by Al3+ in the compositions results in the transfer of Al3+ to A-sites and replaces some Fe3+ to B-sites, which causes the enhancement of Fe3+ ions in the grain and assembles them in the grain boundary [48,61]. Consequently, the space charge polarization is increased and caused a higher value of dielectric constant. Heat generated by the high flow of electricity in dielectric materials is dissipated and considered as the material’s loss that is characterized as the imaginary part (ε′′) of dielectric constant [54]. From Fig. 6(B), it is observed that the value of ε′′ increases significantly with increasing Al3+ content in ferrites. The decrease of ε′′ with frequency is occurred due to the high resistive effect of the grain boundaries. The electrons reverse their direction of motion frequently at higher frequencies and the hopping electrons can no longer follow the applied electric field after certain frequencies. Therefore, the probability of charge transport at the grain boundary decreases, resulting in the decrease of polarization, which gives the low value of ε′′ [17,55,60,61]. Fig. 7 shows the variation of dielectric loss tangent (tan δE ) of the synthesized samples annealed at 700 °C with varying frequencies. Due to impurities and imperfections, the polarization lags behind the applied voltage, causing tanδE to form there [54,61]. The highest value of tanδE is found under the relaxation condition of ωτ = 1, where ω = 2πfmax , and τ = 1/2P represent the peak frequency and the relaxation time, respectively and both of which are closely related to the hopping or jumping probability. Electron sharing between Fe3+ and Fe2+ requires very little energy and the maximum peak is achieved when the hopping frequency between them is well-matched with the frequency of the applied electric field. Koop’s theory explains in a very simple way how tanδE of the investigated materials decreases with frequency [62,63]. It is noted that at lower conductive grain boundaries, tanδE exhibits the maximum value as more electrons are available to be conductive at the low-frequency region. There is energy loss that occurred during the electrons sharing between Fe3+ and Fe2+, therefore high energy is required [47,64,65].The role of microstructure is important in determining the tanδE . H. Jia et al. showed that the grain boundaries and porosity between polycrystalline crystals affect the ε′ and ε′′ [66]. The inter-relation among porosity, grain boundaries, and dielectric loss is defined by the following relation: (22) tan δ E = 1 - P t a n δ o + C m P n where Cm is the material-dependent constant, P represents the porosity and tanδo is the dielectric loss of material with full densification. Uniform density and lower porosity reduce the ε′ and ε′′, respectively and the intrinsic and extrinsic fault are responsible for the dielectric loss.The variation in ac resistivity (ρac ) of the investigated samples with frequencies (annealed at 700 °C) is depicted in Fig. 8 , which is explained based on the hopping mechanism. The ρac is calculated by the following equation [54]: (23) ρ ac = 1 ε 0 ε ′ ω t a n δ E where ω defines the angular frequency. According to the hopping mechanism, electrons jump from one state to another, which prefer to be distributed over the sites in the lattice. In Fig. 8, it is anticipated that at lower frequencies, the ρac of the investigated ferrites has higher values and depletes with increasing frequency. After a certain frequency, it gets almost saturation showing a very small value. This variation of ρac with frequency can be described by the frequency dependency of grains and grain boundaries. The conductivity mechanism illustrates the particle’s ability to be highly conductive [67,68].The high-resistive grain boundaries are more active at lower frequencies, which impedes the movement of free charges and thus the hopping of electrons between Fe2+ and Fe3+ is less, resulting in the higher values of ρac [54]. To increase the hopping of electrons between Fe2+ and Fe3+, it must be operated at higher frequencies, which plays a critical role in reducing the ρac value. The main reason for the low values of ρac is that the hopping of electrons almost stops after a certain frequency. As shown in Fig. 8, the maximum value of ρac is found for the sample Ni0.70Cu0.30Fe2O4.With the increase of Al3+ concentration in Ni–Cu ferrites, the AC conductivity increases. The sample with Ni0.70Cu0.30Al0.1Fe1.9O4 shows the minimum value at the low-frequency region. The conduction takes predominantly through the highly resistive grain boundaries at low frequencies, whereas it occurs through low resistive grains at high frequencies [65–67].The electric relaxation mechanism in the materials can be explained through the spectroscopy of electric modulus (M*), which is resolved into two components [63] as given in the following: (24) M ∗ = 1 ε ∗ = 1 ε ′ - i ε ″ = ε ′ ε ′ 2 + ε ″ 2 - i ε ″ ε ′ 2 + ε ″ 2 = M ′ + i M ″ where M ′ = ε ′ ε ′ 2 + ε ″ 2 is the real and M ″ = ε ″ ε ′ 2 + ε ″ 2 is the imaginary part of the electric modulus. From the above equations, both the real (M′) and imaginary (M′′) parts of the modulus are found to be frequency-dependent, which plays a key role in investigating the relaxation mechanism of the materials. From Fig. 9 (A), it is perceived that M′ responds very well to higher frequencies exhibiting the highest value for x = 0.00. It indicates the lower value of ε′ at high frequencies. The inadequacy of the restorative force and the release of space charge polarization near the grain boundary helps to attain its saturation. This phenomenon occurs at higher frequencies and at the same time ensures frequency independency in the electrical properties of the materials [47,63,64].To illustrate the peaking behavior, one has to look at the variation of M′′ as shown in Fig. 9(B). The hopping mechanism is used to illustrate the peaking behavior better, as it more accurately explains the transition of the charge carriers. In the figure above, it is clear and understand that charge carriers contributing to the hopping process cover long distances at low frequencies. On the other hand, charge carriers are able to cover short distances at higher frequencies, which indicates the relaxation in the polarization process [69,70].The relaxation of the material is distinguished by the cole–cole plot (M′′ vs M′) of the electric modulus as presented in Fig. 10 . The grain and grain boundary is thought to be responsible for this separation [62,69]. A clear non-Debye type relaxation is found by looking closely at the non-overlapping semicircular pattern in Fig. 10. Nanoparticles annealed at 700 °C show two identical non-overlapping semicircular patterns [47].To study the electrical behavior of the material, the impedance spectroscopy was employed in this study for the synthesized nanoparticles. This is a long-established method to distinguish the impedance contributions of the materials’ grains, grain boundaries and electrodes. The complex impedance (Z*) includes both the resistive and reactive components of the impedance as follow: (25) Z ∗ = Z ′ - j Z ″ where the resistive part is designated as the real part Z' which is the horizontal component of the complex impedance denoted by Z' = |Z*|cosθ and the imaginary part is designated as the reactive (capacitive) part expressed by Z“ = |Z*|sinθ. However, these two components are combined impedance effect of resistance and capacitance due to grain and grain boundary, which are embroiled to dielectric and electric modulus parameters following the relation: (26) t a n δ = ε ″ ε ′ = Z ″ Z ′ = M ″ M ′ The variation in the real part of complex impedance (Z′) of the investigated ferrites annealed at 700 °C is illustrated in Fig. 11 (A) with varying frequencies. The higher values of Z′ of the synthesized ferrites are revealed at lower frequencies with dispersed behavior and drop sharply up to 1 KHz and thereafter it remains almost constant at high frequencies.Besides, the variation in imaginary part (Z′′) of the complex impedance for the synthesized Ni0.7Cu0.3AlxFe2-xO4 nano-ferrites is illustrated in Fig. 11(B). As observed in Fig. 11(B), the materials show higher Z′′ values at the lower frequencies likewise the real part (Z′) and decrease rapidly with increasing frequency (up to 10 KHz), as the conductivity of the ferrites increases. However, at higher frequencies ≥ 100KHz, Z′′ shows frequency-independent behavior with small constant values in the effect of reduction in polarization [54,61,71,72]. Both Z′ and Z′′ in Fig. 11 show the similar trend as the dielectric properties of the materials. For all compositions, the impedance curves are appeared to merge at higher frequencies; indicating the predominance contribution of low resistive grains. Moreover, the space-charge polarization is considered important only when the materials are resolved into grains and grain boundaries [61,68]. The curves tend to converge at higher frequencies owing to a decrease in space charge polarization and this behavior elucidates the increasing tendency of ac conductivity with frequency, confirming the semiconducting behavior of the prepared nanocrystalline spinel ferrites [47,71].The Nyquist impedance plot (known as the cole–cole plot) of the prepared Ni0.7Cu0.3AlxFe2-xO4 annealed at 700 °C is shown in Fig. 12 . This plot is the combined response of the RC circuit-connected resistor and capacitor in parallel, which reveals the contribution of grain and grain boundary resistance. The heterostructure nature of synthesized materials along with the characteristic nature of complex impedance spectra are revealed through the appearance of multiple electrical responses. Those responses are resulted due to grain resistance Rg, grain boundary resistance Rgb, and electrode effects), which can be easily determined by the semicircular arcs that appeared in the cole–cole plots [47,54,72]. By looking at Fig. 12, the two semicircular arcs are clearly visible which are formed with their center placed below the real axis, which manifests the single-phase nature of Al3+ substituted nanocrystalline Ni-Cu materials. The diameter of the semicircle arcs is found to decrease with increasing Al3+ concentration, which is actually caused by the resistance of grain boundaries. However, at lower frequencies, the Rg dominates the appearance of the first semicircle, whereas at higher frequencies, the Rgb dominates the appearance of the second semicircle. The difference in relaxation time is considered the main catalyst behind the separation of semicircle arcs [61].The sol-gel method was used to synthesize a series of highly crystalline nanomaterials of Ni0.7Cu0.30AlxFe2-xO4. The single-phase cubic spinel structure of the investigated materials was confirmed through the XRD study with showing no impurity. The surface morphology was studied through the FESEM measurements, which illustrated the distribution of semi-spherical grains separated by the grain boundaries with a homogenous distribution of particles on the surface. The structural parameters were determined using the XRD and FESEM data. The electrical and dielectric properties were carried out by using the impedance analyzer supported by the modulus and impedance spectroscopy. Both the average crystallite size and the average grain size of the studied materials are found to be in the nano-scale range (55.63–70.74 nm) and (59.00– 65.00 nm), respectively. The dielectric dispersion nature of the materials was revealed through the dielectric study of the materials. The electrical response of the materials was inspected by means of impedance and modulus spectroscopy, which resolved the contribution of grains and grain boundaries as the electrical response of the investigated nanoparticles. The relaxation phenomena in the materials was justified through the cole–cole analysis of both impedance and modulus spectra. A little substitution of Al3+ is found to be influential in the structural, dielectric, and electrical properties of Ni-Cu spinel ferrites prepared by the cost-effective sol gel method.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 the center of excellence of the Department of Mathematics and Physics at North South University (NSU), Dhaka 1229, Bangladesh. This research is funded by the NSU research grant CTRG-20/SEPS/13.

In this study, a series of nanocrystalline Ni0.7Cu0.30AlxFe2-xO4 (x = 0.00 to 0.10 with a step of 0.02) has been synthesized through the sol–gel auto combustion technique. The structural, morphological, dielectric, and electrical properties of the synthesized nanoparticles are analyzed due to the substitution effect of Al3+ content. The structural study has been performed through the XRD and FTIR analyses. The extracted XRD patterns assure the single-phase cubic spinel structure of all samples in high crystalline nature with maintaining their homogeneity. The average crystallite size of the nanoparticles is found in the range of 55.63–70.74 nm, and the average grain size varies from 59.00 to 65.00 nm. FTIR study also confirms the formation of spinel structures in the prepared Ni-Cu ferrite nanoparticles. The surface morphology of the materials has been studied through the FESEM study linked with EDX analysis. The dielectric dispersion of the materials is reflected at lower frequencies up to 10 kHz. The impedance spectroscopy confirms the non-Debye relaxation phenomena of the synthesized nanomaterials. The contribution of grains and grain boundaries is resolved through the modulus study of the materials. The trend in variation of AC resistivity with frequency has been explained by the hopping mechanism.

Calcium (Ca) is the fifth richest element in the Earth's crust. It is one of the cheapest and most biocompatible metals, with high content in the human body. The price of Ca is close to three millionths of the price of noble metal Pt of the same quality (Hill et al., 2016). Like other alkaline earth metals, calcium has, in its outermost S orbital, two valence electrons which are easily given up in chemical reactions. Therefore, calcium is usually bivalent in its compounds and exists in ionic forms. The application of calcium in catalytic reactions could be sustainable, economical and green. However, due to the lack of a d-orbital to enable its oxidation state to change rapidly and reversibly, (a prerequisite for many catalytic cycles) (Harder, 2010; Zhu et al., 2020a, 2020b), calcium metal is generally considered as a stoichiometric reagent with no catalytic performance in heterogeneous catalysts (Gerken et al., 2014; Zhu et al., 2015).Differing from the rare usage of calcium in heterogeneous catalysis, applications of calcium in homogeneous catalysis have made tremendous progress during the past decade (Hill et al., 2016; Harder, 2010). For example, calcium alkoxide and calcium amide complexes are sufficiently reactive to promote many catalytic reactions. In some cases carbanions, such as benzyl calcium complexes or (Me3Si)2HC-stabilized alkyl calcium reagents, are highly effective as well. So far, calcium metal complexes have been reported to play a central role in the catalytic cycles of alkenes polymerization (Begouin and Niggemann, 2013), intramolecular hydroamination of aminoalkenes (Crimmin et al., 2005) and hydrosilylation and alkene hydrogenation (Harder and Brettar, 2006). The rapid development of Ca compounds for homogeneous catalysis is mainly based on the viewpoints that the d0 valence configuration of a Ca2+ center in the calcium metal complexes will give it a certain level of ‘lanthanide mimetic’ characteristics so that a catalytic cycle can be constructed (Hill et al., 2016).Recently, Zhou and coworkers found that alkaline earth metal elements Ca, Sr, and Ba can form stable octacarbonyl compound molecules which meet the 18-electron rule and exhibit typical transition metal bonding characteristics (Wu et al., 2018). This indicates that the heavy alkaline earth metal elements may behave like transition metals in certain heterogeneous catalytic processes. However, there are few reports on the use of alkaline earth metals for heterogeneous catalysis. For example, Xia et al identified through theoretical calculations that alkaline earth metals, placed in a covalent organic framework, can become effective electrocatalysts for oxygen reduction reaction (ORR), which is the major reaction for hydrogen fuel cells and metal-air batteries (Lin et al., 2017). Chen et al. proved experimentally that Mg, atomically dispersed in the graphene framework, has extremely high ORR activity under both alkaline and acidic conditions (Xu et al., 2019). However, due to the lack of more experimental results, there is still insufficient evidence to show that alkaline earth metals have enough active catalytic sites in heterogeneous catalysis. In addition, the catalytic mechanism of alkaline earth metals in heterogeneous catalytic reactions can be an exciting field for renewable hydrogen production.Single atom catalysts (SACs) are an innovative type of heterogeneous catalysts in which each isolated active metal atom is fixed on supporting materials (Wang et al., 2019; Kaiser et al., 2020; Zhuo et al., 2020). Although SACs are classified as the heterogeneous catalysts, the presence of single metal atoms in SACs is very similar to that in homogeneous catalysts (Yang et al., 2017). The surface atoms of the supporting materials can be considered as ligand molecules in homogeneous catalysts, which not only stabilize the active metal atoms but also engage in the catalytic reactions (Wang et al., 2019; Wu et al., 2019). The similarity between SACs and homogeneous catalysts has driven us to explore the use of calcium metals for heterogeneous catalytic hydrogen evolution reaction (HER).In this research, we have found that atomically confined Ca in nitrogen-doped graphene (Ca1-NG) can be an effective heterogeneous catalyst to boost the electrocatalytic hydrogen evolution (EHE) and photocatalytic hydrogen evolution (PHE) reactions. To the best of our knowledge this is the first report that calcium single atoms have been used as catalysts for the HER. The performance of Ca1-NG loaded CdS is comparable to that of noble metal Pt loaded CdS for PHE under the same experimental conditions. Density functional theory (DFT) calculations have shown that the excellent performance of Ca1-NG can be attributed to the optimal adsorption capacity of hydrogen atoms on the Ca-doped active centers.Ca1-NG was prepared using a facile method previously described for the preparation of Co1-NG and Ni1-NG (Zhao et al., 2017; Zhao et al., 2018; Fei et al., 2015). Briefly, a complete mixture of graphene oxide (GO) and CaCl2 was thermochemically treated in an NH3 atmosphere to form the Ca1-NG. During this process GO was reduced to NG (supplemental information Figure S1), and the N dopants were incorporated into the graphene lattice to form a strong interaction with metal atoms (Wang et al., 2019; Zhao et al., 2017).No diffraction peaks of Ca oxides or carbides were detected in the X-ray diffraction (XRD) patterns of Ca1-NG samples (supplemental information, Figure S1). Transmission electron microscopy (TEM) images show that there are no Ca-related nanoparticles in the prepared Ca1-NG samples (Figure 1 A). However, the energy-dispersive X-ray elemental mapping spectroscopy (EDS) indicated that Ca, N, and C elements are distributed evenly on the prepared Ca1-NG (Figure 1D). The aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of Ca1-NG further demonstrated that Ca species were homogeneously dispersed in the substrates. As shown in Figures 1B and 1C, a small number of bright spots with diameters less than 0.2 nm are well dispersed on the substrates. The absence of Ca clusters has been confirmed with careful examination at several randomly picked locations during HAADF-STEM observations. These results indicate that all Ca species are atomically dispersed in the Ca1-NG. The loading content of Ca in Ca1-NG is 0.52 wt.% based on the analysis of inductively coupled plasma optical emission spectrometer (ICP-OES).X-ray photoelectron spectroscopy (XPS) analysis was performed to investigate the chemical composition and valence state of Ca1-NG (Figure 2 and Table S1). The survey spectrum with major C peaks and some smaller peaks of N, O and Ca confirms the presence of C, N, O and Ca in Ca1-NG (Figure 2A). The high-resolution N 1s spectrum shows that Ca1-NG catalyst contains mainly pyridinic N (398.0 eV) as well as a small amount of pyrrolic (399.5 eV), graphitic (400.8 eV), and oxidized N species (402.1 eV) (Figure 2B). The presence of pyridinic N is not only favorable for hydrogen evolution activity of Ca1-NG but also serves as anchoring sites for single metal atoms. Figure 2C shows the high-resolution Ca 2p spectrum of Ca1-NG. According to the National Institute of Standards and Technology XPS database (Naumkin et al., 2012), the 347.2 eV and 350.8 eV binding energy peaks can be attributed to Ca 2p3/2 and Ca 2p1/2, respectively. The absence of metallic Ca 2p3/2 spectrum (344.9 eV) indicates that the scattered Ca species (shown in HAADF-STEM images) in Ca1-NG are Ca2+ cations. The Raman spectra of the resultant catalysts in Figure 2D exhibit a D-band for defected graphite and a G-band for the doubly degenerate zone center E2g mode (Ferrari and Basko, 2013). The intensity ratio of D band to G band (ID/IG) for Ca1-NG (0.98) is close to that of NG (1.03). This result indicates that the dispersion of individual Ca atoms in the NG matrix has little effect on the degree of disorder and structural defects in the NG laminar structure (Zhao et al., 2018).The atomic dispersion of Ca cations in Ca1-NG was further confirmed by the X-ray absorption near-edge structure (XANES) spectroscopy and the extended X-ray absorption fine structure (EXAFS) spectroscopy, which are sensitive to the local environment of metal atoms. Figure 3 A shows the Ca K-edge of XANES curves of Ca1-NG and CaO. Usually a metal foil is used for energy calibration. However, because calcium metal is very active in air, the Ca K-edge XANES spectrum of CaO was used as calibration reference material. As shown in Figure 3A, the adsorption edge position of the Ca1-NG XANES curves is comparable to that of CaO, indicating that Ca metal atoms in Ca1-NG are in cationic states. This agrees well with the results of the XPS spectra (Figure 2C). Further structural information was obtained from Ca K-edge EXAFS analyses (Table S2). Figures 3B and 3C show the Ca K-edge EXAFS K-space and R-space plots, respectively, for the Ca1-NG. It is noted that the EXAFS curve of Ca1-NG is obviously different from that of CaO. The R space plots of Ca1-NG show a sharp peak at approximately 2.1 Å. However, CaO shows two strong bonding features at around 1.9 Å and 3.1 Å, which are attributed to the Ca-O bond and Ca-O-Ca bonds, respectively. The major peak for Ca1-NG at approximately 2.1 Å can be corresponded to the formation of Ca-N bond, which is longer than that of Ca-O bond (1.9 Å) in CaO. Atomic structure simulations indicate that the anchored Ca single atoms are located at the defective sites of NG derived from pyridine-N (Figure 3D and supplemental information Figure S2). The fitting results indicated a CN of 2.8 for Ca-N contribution in Ca1-NG. This result corresponds well to DFT calculations (Figure S17), which indicate that single Ca atoms anchored in pyridinic N defects in graphene are stable (Detailed information can be found in the following DFT calculation section of this research).Experimental results have shown that the prepared Ca1-NG exhibits more enhanced activities for HER under both acidic and weak basic conditions than do other obtained catalysts (Figures 4A and 4B and supplemental information Figure S3). The HER activities of Ca1-NG were evaluated, both in 0.5 M H2SO4 and 1.0 M (NH4)2SO3 solutions using a standard three-electrode electrochemical cell. The commercial 20 wt.% Pt/C and the prepared GO and NG were also evaluated as baseline catalysts. All potentials were referenced to the reversible hydrogen electrode (RHE) and with iR-corrected. As shown in Figure 4A, Ca1-NG shows HER activity in acidic solution with an onset potential (Eonset) of 21 mV and an overpotential of 151 mV to deliver a current density of 10 mA cm−2. For comparison, NG and GO show poor activities toward HER, requiring much greater overpotentials of 297 mV and 338 mV, respectively, to generate the same 10 mA cm−2 current density.The enhanced HER activity of Ca1-NG is further confirmed by the smaller Tafel slope of 76 mV dec−1 for Ca1-NG as compared to 137 mV dec−1 for NG and 147 mV dec−1 for GO (Figure 4B). The small Tafel slope indicates that the rate-determining step of Ca1-NG is either the electrochemical desorption of H or the discharge reaction, following the Volmer-Heyrovsky mechanism (Dong et al., 2018). Although the Tafel slope of Ca1-NG is higher than that for the benchmarked 20 wt.% Pt/C catalyst (36 mV dec−1), it is significantly lower than that of NG without Ca single atoms. This result suggests that the new Ca single atom can be effectively used as the catalytically active site of HER. In addition, Ca1-NG has also shown more favorable HER activity under neutral (or weak basic) conditions of 1.0 M (NH4)2SO3 with pH = 8.0, as the obvious shift of the polarization curve for Ca1-NG catalyst to a lower overpotential (Figure S3). These results indicate that the incorporation of calcium single atoms into N-doped graphene can lead to a profound enhancement of the HER activity for Ca1-NG under both acidic and weak basic conditions.The electrochemical active surface area (ECSA) of prepared catalysts was analyzed by means of Cdl in Figure S4. The results showed that the capacitances of GO, NG, and Ca1-NG are 3.07, 3.68, and 5.22 mF cm−2 in a 0.5 M H2SO4 solution, corresponding to 76.8, 92.0, and 130.5 cm2 ECSAs, respectively. The ECSA of a 20 wt.% commercial Pt/C catalyst was measured using the underpotential deposition hydrogen (UPD-H) adsorption/desorption voltammetry method, which is usually used for the determination of ECSAs for noble-metal electrocatalysts. As shown in Figure S5, the ECSA for Pt/C was determined to be 285.7 cm2. The turnover frequencies (TOFs) of the testing catalysts were calculated to evaluate the intrinsic activities of the catalysts. At overpotential of 100 mV, the TOF values of the GO, NG, and Ca1-NG were 0.125, 0.147, and 1.134 H2 s−1, respectively. These values revealed that Ca1-NG had intrinsic HER activity excelling other catalysts. An equivalent circuit simulation for electrochemical impedance spectroscopy (EIS) tests was carried out from 10−2 Hz–106 Hz (Figure S6). Ca1-NG shows a smaller arc radius compared to those of GO and NG, which means that the electrochemical impedance of Ca1-NG is smaller than those of GO and NG.Electrochemical stability is an important indicator used to evaluate the catalytic performance of catalysts. The result of the i-t curve (Figure S7) shows that the catalytic current remains constant at about 17 mA cm−2 at 200 mV for over 30,000 s. This result indicates the high stability of Ca1-NG catalyst in a 0.5 M H2SO4 solution. The XPS and STEM analyses of Ca1-NG after HER are shown in Figures S8 and S9. The results show that the XPS and STEM characterizations of Ca1-NG do not change significantly after the reaction, proving that the structure of Ca1-NG is stable.On the other hand, the prepared Ca1-NG can significantly enhance the performance of CdS for PHE. The formation of the Ca1-NG loaded CdS composite photocatalysts (Ca1-NG/CdS) was confirmed via TEM images (Figure S10). XRD patterns and UV-visible light absorption spectra show that a small amount of Ca1-NG loading does not affect the crystal structure of CdS but significantly improves the light absorption capacity of the photocatalyst (Figures S11 and S12). The PHE performance of Ca1-NG/CdS under visible light irradiation at 420 nm was evaluated using (NH4)2SO3 as an electron donor. As shown in Figures 4C and 4D, Ca1-NG/CdS exhibits much higher PHE activities compared to bare CdS and NG loaded CdS. The rate of hydrogen evolution for 0.5 wt.% Ca1-NG/CdS (92.0 μmol/h) is 8.1 times greater than that of bare CdS (11.3 μmol/h) and 1.6 times greater than that of 0.5 wt.% NG/CdS photocatalyst (57.7 μmol/h). Moreover, the catalytic performance of 0.5 wt% Ca1-NG/CdS was comparable to that of 0.5 wt.% Pt/CdS (Figure S13), an active photocatalyst for PHE. In order to verify the role of CaO nanoclusters in hydrogen production we loaded CaO onto the surface of NG and successfully prepared CaO-NG/CdS. The hydrogen evolution performance of CaO-NG/CdS is shown in Figure S14. It is noted that CaO nanoclusters have no catalytic effect on the HER. Therefore, we can conclude that it is the Ca single atoms in Ca1-NG/CdS that play a major catalytic role, rather than the CaO nanoclusters. It is noted that the loading content of Ca in Ca1-NG is only 0.52 wt.% (based on ICP-OES analysis). That means that very few Ca atoms, only 26 parts per million mass of CdS, are needed in order to facilitate the PHE reactions. This result also indicates that single calcium atoms in NG play an important role in the improvement of the hydrogen evolution activity, which agrees well with the enhanced EHE performance for Ca1-NG.The effect of Ca1-NG loading concentration was investigated and the results are shown in Figure 4E. The rate of hydrogen evolution increases from 11.3 μmol/h to 92 μmol/h as the Ca1-NG loading on CdS photocatalysts increases from 0.0 to 0.5 wt.%. Further increasing Ca1-NG loading, however, results in a significant drop in the rate of hydrogen evolution. This decline is possibly due to the light blockage effect of Ca1-NG on the surface of CdS. The optimal loading of Ca1-NG on CdS is about 0.5 wt.% under the present reaction conditions. The apparent quantum efficiency of the optimal Ca1-NG/CdS photocatalysts for hydrogen production is 57.5% at 420 nm wavelength. As illustrated in Table S3, this efficiency (57.5%) is one of the greatest ever reported for non-noble-metal cocatalysts.The stability of Ca1-NG/CdS photocatalyst was verified by a three PHE reaction cycles test. As shown in Figure 4F, no significant decrease in the rate of hydrogen evolution was observed during the cyclic test. During the three PHE cycles test a total of 1.264 mmol H2 was produced. The turnover numbers (TONs), which are defined as the total hydrogen atoms evolved per mole of CdS photocatalyst and per mole of Ca anchored in Ca1-NG/CdS, are 73 and 779363, respectively. These large TONs indicate that hydrogen is produced from the photocatalytic reduction of water rather than from the photo-corrosion of either CdS or Ca in the Ca1-NG/CdS photocatalysts. Additionally, XRD, TEM and ICP tests have shown that there are no significant differences for the Ca1-NG/CdS photocatalyst before or after the stability test (Figures S10 and S6 and Table S4), indicating that Ca1-NG/CdS is stable during PHE processes. The stability of Ca1-NG/CdS was further confirmed by a long-term photoelectrocatalytic test in 0.5 M H2SO4. As can be seen in Figure S15, the linear sweep voltammetry (LSV) curves for Ca1-NG/CdS show no clear difference before or after a long-term photoelectrocatalytic hydrogen evolution test. This result also indicates that Ca1-NG/CdS is a stable catalyst for PHE.Photoluminescence (PL) and time-resolved photoluminescence (TRPL) decay spectra measurements were carried out to evaluate the charge carrier trapping and transfer mechanism in Ca1-NG/CdS photocatalyst during photocatalytic reactions (Figures 5A and 5B). The weak peak around 475 nm in PL spectra can be ascribed to the band edge emission of CdS, while the higher broad band at around 550 nm originates from the trap states (Veamatahau et al., 2015; Mathew et al., 2011). Clearly, the PL intensity of Ca1-NG/ CdS is much weaker than that of bare CdS, indicating that the photogenerated electron-hole pair recombination is effectively suppressed after Ca1-NG is loaded onto the surface of CdS. This may result from the effect of co-catalyst trapping photogenerated electrons (Chen et al., 2010). Moreover, the PL intensity of Ca1-NG/CdS was weaker than those of NG/CdS and GO/CdS. This result is consistent with the better PHE performance for Ca1-NG/CdS.The transfer efficiency of photogenerated charge carriers was further confirmed by the TRPL decay spectra (Figure 5B). The decay curves easily approximate a biexponential function. As shown in Table S5, the average lifetime of the PL decay in bare CdS was 2.37 ns. However, after NG and Ca1-NG loading, the PL lifetimes of the NG/CdS and Ca1-NG/CdS photocatalysts were reduced to 1.88 and 0.93 ns, respectively. These results suggest that the presence of Ca1-NG provides a new pathway for the electron transfer from CdS to Ca1-NG, leading to a significant decrease in the PL decay lifetime (Jiang et al., 2017). In addition, the lower PL average decay lifetime of Ca1-NG/CdS compared to that of NG/CdS further confirms that Ca single atoms anchored in NG result in more effective separation of the photogenerated carriers, thereby leading to higher photocatalytic activity.To further understand the role of Ca1-NG cocatalyst in PHE, the transient photocurrent-time curves of Ca1-NG/CdS, NG/CdS, GO/CdS and bare CdS samples underwent several on-off cycles of intermittent irradiation at 420 nm. As shown in Figure 5C, all the samples demonstrated a prompt photocurrent generation during the on and off illumination cycles. These on-off cycles also show high reproducibility. It is noteworthy that Ca1-NG/CdS exhibits greater photocurrent compared to NG/CdS and bare CdS. The photocurrent intensity of Ca1-NG/CdS was almost two times higher than that of NG/CdS, suggesting the positive roles of Ca doping in the acceleration of charge separation, which agrees with the results shown in Figures 5A and 5B.From a charge transfer viewpoint, EIS further shows the positive roles of Ca single atoms in Ca1-NG/CdS for PHE. In this research, EIS was carried out under visible light illumination and using a typical three-electrode setup. A smaller semicircle radius of an EIS curve generally means a lower charge transfer resistance and thus faster interface charge transmission of a photocatalyst (Zheng et al., 2020; Shi et al., 2020; Yao et al., 2019). As shown in Figure 5D, the Nyquist plots of Ca1-NG/CdS have much smaller semicircles than those of NG/CdS and bare CdS, suggesting a more efficient charge separation and transfer within Ca1-NG/CdS and, therefore, a better PHE performance.LSV tests under 420 nm visible light irradiation using 1.0 M (NH4)2SO4 aqueous solution as a photolyte show that Ca1-NG loading can effectively reduce the overpotential of CdS for PHE. As shown in Figure 5E, the overpotential for Ca1-NG/CdS at −10 mA cm−2 is 0.62 V, much lower than those of NG/CdS (−0.76 V) and bare CdS (−0.90 V). (Note that a lower overpotential means a lower required activation energy for the HER (2H+(aq) + 2e− → H2(g)) (Kweon et al., 2020) and is also favorable for photocatalytic H2 production (Shi et al., 2020; Yao et al., 2019; Luo et al., 2015)). Additionally, the conduction band (CB) potentials of Ca1-NG and CdS were estimated to be −0.54 and −0.39 V (vs. NHE) using the Mott-Schottky method (Figure S16). A more negative CB position indicates that photogenerated electrons in CdS under light irradiation can migrate from CdS to Ca1-NG (Figure 5F), which agrees well with the results of PL and TRPL decay spectra.We can conclude, based on these characterization results, that Ca1-NG can serve not only as an electron storage medium to effectively inhibit the recombination of charge carriers, but also as active sites to accelerate the HERs. In addition, single Ca atoms doping in NG plays a key role in the improvement of catalytic performance of Ca1-NG and Ca1-NG/CdS for hydrogen evolution.DFT simulations were carried out to provide an in-depth theoretical understanding of the roles the Ca single-atoms play in the HER and PHE.Our calculations show that the Ca atom is located on the central axis after structural relaxation (Figures S17), which is consistent with experimental observations (Lin et al., 2015). Both the SV+3N + Ca and DV+4N + Ca structures (two most common carbon vacancies: single vacancy (SV, refer Figure S17A) and double vacancy (DV, refer Figure S17B), exist in the NG. Creation of an SV (DV) leads to three (four) carbon atoms having dangling electrons (CN equals 2). Replacing one of these three/four carbon atoms by N results in a pyridinic-N that coexists with an SV/DV. In principle we can replace multiple carbon atoms to form SV + xN (x = 1, 2, 3) and DV + yN (y = 1, 2, 3, 4) structures, and the Ca atom is about 1.77 Å and 1.30 Å above the 2D plane, forming identical N-Ca bonds with lengths of 2.18 Å and 2.26 Å, respectively (slightly longer N-Ca bonds in the DV+4N + Ca structure indicates weaker N-Ca bonding strength. This is because the two valence electrons of Ca split into only 3 Ca-N bonds in the SV+3N + Ca structure, whereas they have to split into 4 Ca-N bonds in the DV+4N + Ca structure, resulting in less electron density forming each Ca-N bond in the latter case). In a previous study, we found that a Ni single atom could also be supported above the SV+3N structure, but it would drop into the double vacancy surrounded by 4 N, making it no longer useful for HER. Here, a Ca single atom could be supported above the plane for both SV+3N and DV+4N structures, as the size of the Ca atom is larger than most transition metal atoms. The calculated adsorption energies for Ca single atoms adsorbed at the centers of SV+3N and DV+4N structures are ΔECa = −4.50 eV and −5.96 eV, respectively. These adsorption energies are much more negative than the Ca crystal cohesive energy of about −1.84 eV (Lee et al., 2009), indicating that the Ca adsorption at the SV+3N and DV+4N centers is extremely stable.Next, we studied H adsorption on these four structures and examined how the Ca single atom affects the H adsorption energy. We considered H adsorption at both Ca-sites and N-sites (we ignore H adsorption at the C-sites because they are not stable and are much less affected by the Ca atom). Because each structure involves several N atoms, we denote them as N1, N2, N3 (and N4), as illustrated in Figure 6 . We consider all possible situations, with several H adsorbed onto a combination of Ca and N atoms, and we name an H adsorption configuration by the H adsorption sites. For example, [N1, N2, Ca∗] denotes a configuration with three H atoms adsorbed onto N1, N2, and Ca, respectively. When discussing ΔGH∗ values of a particular H within a configuration involving several H, we further denote the adsorption site of the discussed H using ∗. For instance, in the former example [N1, N2, Ca∗], the discussed H is on Ca∗ site. Various H adsorption configurations for the SV+3N versus SV+3N + Ca and DV+4N versus DV+4N + Ca structures are shown in Figures 6 and 7 . In particular, for the structures involving Ca, H+ coming from solution above graphene can adsorb onto the Ca-site, and H+ from underneath graphene can adsorb onto the N-site. With the same number m (m > 1) of adsorbed H, the catalytic system can have these H atoms (i) all adsorb onto the N-sites (Figure 6B) or (ii) it can have one adsorbed onto Ca and the remaining m-1 H adsorbed onto N-sites (Figures 6C and 7). Although the energies (i) and (ii) might be slightly different, both structures could exist in solution with sufficient lifetime for catalyzing HER. It is difficult to have structural transition from one to the other since H on Ca-site and H on N-site are spatially separated on different sides of graphene. Therefore we considered both structures.In the case of a single Ni atom supported on SV+3N or DV+4N structures, we find that Ni-N bond could be broken if too many N and Ni sites are adsorbed with H. Here the Ca-N bonds are not broken, even if all the N and Ca sites are adsorbed with an H (Figure 6). This is due to the unique property of Ca, that it can host a large CN (Yoon et al., 2008). This unique property makes the Ca structure extremely stable/robust in the dynamic solution and makes sure all the H adsorption sites can contribute to HER. Figures 6 and 7 show clearly that with a Ca atom adsorption we get not only an extra Ca-site for H adsorption, but also more than three times as many possible processes for H adsorption. In addition, many of these H adsorption configurations are associated with small |ΔGH∗| values, as highlighted in bold in Figures 6 and 7.All H adsorption processes in the four structures and corresponding ΔGH∗ values are illustrated in Scheme 1 , where ΔGH∗ is reflected by the G difference between the initial configuration and final configuration. For the structures without Ca (black curves in Scheme 1), most |ΔGH∗| values are very large. There is only one small |ΔGH∗| (0.34eV), when an H is adsorbed to N2 of the SV+3N [N1] configuration, as highlighted in red. On the contrary, the two structures with Ca (blue curves in Scheme 1) both involve many |ΔGH∗| values close to zero, as highlighted by red or orange. In particular, the red processes are especially useful for HER as they not only involve small |ΔG H∗| values but also start from configurations that are highly likely to exist in the solution, because the starting configurations are either the initial configuration without any H or a configuration requiring a small or even negative ΔG H∗ from the initial configuration. The SV+3N and DV+4N structures involve only one red H adsorption process, while the SV+3N + Ca and DV+4N + Ca structures involve 4 red H adsorption processes. In addition, all the red lines H adsorption are on the Ca-site, indicating that the Ca atom plays an essential role in providing many suitable H adsorption configurations to catalyze HER.Before explaining the detailed mechanism of Ca atoms in HER we first examine the effect of H coverage in the structures without Ca. For both SV+3N and DV+4N structures higher H coverage induces larger ΔGH∗ values (Figure 6A), consistent with the general trend that H binding becomes less stable when more H are adsorbed in the vicinity. When H coverage is low (Figure 6A black boxes), the ΔGH∗ values are very negative (−2.32 eV, −1.53 eV, and −1.34 eV). (The delta G_H value for the first H∗ in the SV+3N defect is about 0.8eV more negative than that in the DV+4N defect because the former configuration also involves more interaction between H∗ and N2, N3 (more details are given in SI), indicating a very strong H binding. Adding one more H to either structure dramatically increases ΔGH∗ to quite positive values (0.34 eV and 0.78 eV), indicating a significant reduction in the binding strength of additional H. The abrupt reduction of H binding strength can be understood from their atomic structures. For the three configurations with low H coverages (Figure 6A black boxes) the H atom(s) is located inside the vacancy hole and the whole structure is well within a 2D plane (see side view), where H and N form a sp2-like bond. In addition, the in-plane H interacts with other N atom(s) via a quasi H-N bond as they are spatially close enough for the charge densities to sufficiently overlap (Figure 8 A), which further enhances the binding strength of H (Especially in the [N1] configuration of SV+3N, as H forms quasi H-N bonds with two other N atoms). However, if more H atoms were added in, they would be too squeezed within the small vacancy hole. Hence they become out-of-plane due to Pauli repulsion (Figure 6A, below black boxes). Some N atoms also move out-of-plane. This changes the H-N bonding from sp2-like toward more sp3-like and also reduces the charge density interaction between H and other N atoms. Hence the H binding strength is significantly reduced.When a Ca atom is deposited onto SV+3N or DV+4N, the Ca atom strongly binds to all the N atoms. This increases the CN of each H adsorbed N from three to four (Figure 6B) and changes the N-H bonding nature to sp3-like. In addition, H is pushed out-of-plane substantially, losing quasi-bonding interaction with other N atoms (Figure 8B). This significantly reduces the H binding strength on N-sites. As a result, the corresponding ΔGH∗ values become quite positive, but with |ΔGH∗| closer to 0 than in the structures without Ca (Figure 6B boxes vs Figure 6A boxes). This is one effect of Ca single atoms, namely reducing the H binding strength on N-sites to better values for HER by changing the H-N bond nature to more sp3-like and reducing the charge density interaction between H and other N atoms.In fact, the Ca atom too significantly reduces the H binding strength at the N-site, making ΔGH∗ a bit too positive for HER (Figure 6B boxes). The influence of Ca could be slightly weakened by further adsorbing an H on top of Ca, which adjusts ΔGH∗ (of H adsorption on N) to less positive values, making the systems more suitable for HER (Figure 6C blue boxes). For example, the [N 1 ∗, Ca] configuration of SV+3N + Ca exhibits ΔGH∗ values of 0.06 eV, superior for catalyzing HER. Here the influence of Ca on the graphene structure is reduced because the H-Ca bond weakens Ca-N interactions, as can be seen by the increase of the Ca to graphene-plane distance (Figure 6C vs 6B). This makes the whole graphene structure deviate less from a 2D planar layer, hence allowing most of the N-H bonds to become less sp3-like and more sp2-like (Figure 6C vs 6B). The returning of the H toward the 2D planar layer also enhances the charge density interactions between H and other N atom(s) (Figure 8C vs 8B). Hence the overall H binding to N-sites is strengthened and becomes more suitable for HER.For the four structures outside the boxes in Figure 6A, the H bindings are very weak because these H atoms are already repulsed out-of-plane by other H, even without a Ca deposition. Depositing a Ca atom makes slight differences to the H binding strength. Furthermore, adsorbing an H onto Ca also makes little change to the H binding strength on N-sites.Besides the two effects discussed above, the Ca atom itself also serves as an H adsorption site with exceptional ΔGH∗ values. Our calculations predict that an H atom adsorbs on top of Ca, with ΔGH∗ values of 0.46 eV and 1.50 eV for SV+3N + Ca and DV+4N + Ca, respectively (Figure 7). The latter case has a weaker H-Ca bond because its Ca CN is higher. Although the two ΔGH∗ values are too positive for HER, they can be reduced to very good values by adsorbing H atoms onto the N-sites (Figure 7). For example, for the [ N 1 , Ca∗ ] and [ N 1 , N 2 , Ca∗ ] configurations of SV+3N + Ca, and the [ N 1 , N 3 , Ca ∗] configurations of DV+4N + Ca, ΔG H∗ are reduced to exceptional values for HER: 0.19, 0.22, and 0.16 eV, respectively. However, this excludes the [N 1 , N 2 , N 3 , Ca∗] configuration of SV+3N + Ca, where ΔGH∗ is slightly increased compared to [Ca∗] configuration.Adsorbing H on the N-sites can reduce the ΔGH∗ values of H adsorption on Ca sites, because the H-N bond weakens the N-Ca bonds, as can be seen by the N-Ca bond length increase (Figure 7 rows 2-5 compared to row 1). The weakening of N-Ca bonds is also reflected in the charge density plots in Figure 8. When there is no H adsorbed on any N we clearly see charge density overlap between Ca and the three/four N atoms. After adsorbing one or more H onto the N sites, most overlapping between Ca and N diminishes substantially. As a result, more valence electrons of Ca are involved to form a stronger Ca-H bond. Again, the strengthening of the Ca-H bond is reflected in both decreased Ca-H bond length and increased charge density overlap between Ca and H (Figure 7 row 2-5 compared to row 1).In summary, we have explained three effects of Ca single atoms. First, the Ca atom makes H binding on N sites less stable by changing the H-N bonding nature more toward sp3-like and reducing the charge density interaction between H and other N atoms. Secondly, the H-N binding is over-weakened by a Ca single atom. With an extra H adsorbed on top of Ca the H-N binding can be strengthened. Thirdly, the Ca atom itself serves as an H adsorption site, with the adsorption strength adjustable by H adsorbed onto N. The latter two effects both result in many H adsorption processes with perfect ΔGH∗ values. In particular, without Ca single atoms there are 7 unique processes of H adsorption and most of them have ΔGH∗ values that are either too negative or too positive. Depositing a single Ca atom generates 23 unique processes of H adsorption and many of them are better than the 7 processes in former situations for HER. Therefore, we conclude that the Ca single atom significantly enhances the HER activity of N-doped graphene.Atomically confined calcium in NG (Ca1-NG) was successfully synthesized as an efficient catalyst for electrocatalytic and photocatalytic hydrogen evolution. HADDF-STEM images and X-ray absorption spectroscopy analyses confirm the uniformly dispersed single Ca atoms on the NG substrate. Ca K-edge EXAFS fitting curves and DFT calculations indicate the Ca single-atoms are anchored in the pyridinic-N defects in graphene to form a Ca-N3 structure. DFT calculations suggest that Ca atoms are trapped in SV+3N and DV+4N centers and Ca clustering is prevented. The high catalytic activity of Ca1-NG for HER and PHE results from the Ca single-atoms in NG, which leads to multiple H adsorption configurations with very favorable ΔGH∗ values for HER. This research has pointed to a new approach for the development of high performance HER catalysts using non-transition metals.Here we have revealed that atomically confined Ca in NG (Ca1-NG) can effectively boost the electrocatalytic and photocatalytic HERs (EHE and PHE). Catalyst characterizations have shown that Ca single atoms anchored in NG can efficiently enhance the HER performance, improve the interfacial charge transfer, and suppress the photo-generated charge recombination. However, one limitation of this study is that the loading concentration of single-atom Ca prepared by the current method is low. We will further improve the single-atom preparation method to increase the loading in our future work. REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals, peptides, and recombinant proteins cadmium sulfide Sinopharm Chemical Reagent (Shanghai, China) CAS:1306-23-6 anhydrous calcium chloride Sinopharm Chemical Reagent (Shanghai, China) CAS:10043-52-4 concentrated sulfuric acid Sinopharm Chemical Reagent (Shanghai, China) CAS:7664-93-9 potassium permanganate Sinopharm Chemical Reagent (Shanghai, China) CAS:7722-64-7 graphite powder Macklin Reagent Co., Ltd CAS:7782-42-5 sodium nitrate Macklin Reagent Co., Ltd CAS:7631-99-4 30% hydrogen peroxide Macklin Reagent Co., Ltd CAS:7722-84-1 ammonium sulfite monohydrate Aladdin Reagent Co., Ltd CAS:7783-11-1 Software and algorithms Vienna Ab-initio Simulation Package (VASP) Tongji University http://software.tongji.edu.cn/Home/IndexPage Other JEOL JEM-2100F/HR transmission electron microscope JEOL (BEIJING) CO., LTD. http://www.jeol.com.cn/product/detail/617 JEOL JEM-ARM200F microscope JEOL (Beijing) Co., Ltd. http://www.jeol.com.cn/product/detail/402 BRUKER-D8 X-ray diffractometer Bruker (Beijing) Scientific Technology Co. Ltd. https://www.bruker.com/zh/products-and-solutions/diffractometers-and-scattering-systems/x-ray-diffractometers/d8-advance-family/d8-advance-eco.html Lab RAM high-resolution (HR) evolution Raman spectrometer HORIBA Jobin Yvon https://www.horiba.com/cn/scientific/markets-industries/display-technologies/ ESCALAB250 spectrometer Thermofisher Scientific(China)Co.,Ltd. https://www.thermofisher.cn/order/catalog/product/SID-10148252?SID=srch-hj-ESCALAB250%20spectrometer#/SID-10148252?SID=srch-hj-ESCALAB250%20spectrometer Inductively coupled plasma optical emission spectrometer (ICP-OES) Optima 8000 PerkinElmer Management (Shanghai) Co., Ltd https://www.perkinelmer.com.cn/searchresult?searchName=Optima%25208000&_csrf=f3b614e8-0109-41be-a294-dfb37e7310da Fluorescence Detector (RF-10A, Shimadzu, Japan) Shimadzu (Japan) Co., Ltd. https://www.shimadzu.com.cn/an/gc/index.html Edinburgh FLS9800 Edinburgh Instruments Ltd. https://www.selectscience.net/companies/edinburgh-instruments-ltd/?compID=7445 Nicolet iS10 (Thermo Fisher, USA) infrared spectrometer Thermofisher Scientific(China)Co.,Ltd. https://www.thermofisher.cn/cn/zh/home.html beamline XAFCA Singapore Synchrotron Light Source (SSLS) https://lightsources.org/cms/?pid=1000130 Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Weifeng Yao ([email protected]).This study did not generate new unique reagents.This study did not generate any unique datasets or code.Graphene oxide (GO) was synthesized using the traditional Hummer method. In detail, 2.000 g of graphite powder, 1.000 g of NaNO3 and 46 ml of H2SO4 were added into a beaker soaked in an ice bath and well stirred. Under stirring, then 6.000 g of KMnO4 powder was slowly added to the above mixture for 10 minutes. The mixture was then heated to 35°C for 30 minutes. Next, after adding 92 mL of deionized water the mixture was heated to 98°C. At the same time 60 mL 30% H2O2 was slowly added to the mixture to prevent graphene oxidation. Finally, the mixture was centrifuged and washed repeatedly with deionized water. A golden yellow suspension was obtained by dispersing the obtained precipitate in water and then filtering. GO was obtained by freeze-drying the golden yellow suspension.Single calcium atom anchored nitrogen doped graphene (Ca1-NG) was synthesized via an impregnation method, followed by a calcination process under NH3 atmosphere. The synthesis details are as follows: 100.0 mg of GO and 1.0 mg of CaCl2 were dispersed into 50 mL deionized water. The mixture was sonicated for 4 hours to form a uniformly dispersed suspension. Then liquid nitrogen was added into the suspension to form a solid mixture, followed by freeze-drying for 24 hours. The resulting product was named Ca-GO. Finally, Ca-GO powder was calcinated under NH3 at 750°C for 1 hour to synthesize Ca1-NG. The method for the preparation of CaO-NG was adopted from a similar method reported, except that in this research it was calcined in air at 750°C for 1h before calcining under NH3. The prepared Ca1-NG (or CaO-NG) was then coupled with CdS using an impregnation method. Briefly, certain amounts of Ca1-NG (or CaO-NG) and CdS were added into an ethanol solution. Then the mixture was stirred at room temperature until the ethanol had completely evaporated. The obtained dark yellow powder was Ca1-NG/CdS (or CaO-NG/CdS).Photocatalytic hydrogen evolution activity was measured at 420 nm wavelength. 5.0 mg prepared catalysts were dispersed in 10 mL 1.0 M aqueous (NH4)2SO3 solution. The solution was degassed with N2 for 1 h to remove dissolved oxygen before being irradiated with a single-wavelength (420 nm) light-emitting diode (LED) monochromatic lamp (CEL-LED 100). The H2 evolution volume was analyzed via an online gas chromatograph (Techcomp Limited Co., GC7890II) equipped with a thermal conductivity detector. Ultra-pure nitrogen was used as a carrier gas.The apparent quantum efficiency (AQE) was measured and calculated according to the following equation: A Q E ( % ) = n u m b e r o f r e a c t e d e l e c t r o n s n u m b e r o f i n c i d e n t p h o t o n s × 100 % = number of evolved H 2 molecules x  2 number of incident photons × 100 % = 2 × n H 2 I 0 × t × 100 % where n H 2 is the mole numbers for hydrogen evolution from t = 0 to time t. and I 0 is the Einstein of incident photons per second measured at λ = 420 nm.Electrochemical properties of catalysts were measured using a CHI 660E electrochemical workstation in a standard three-electrode cell. 5.0 mg catalysts were dispersed in a solution consisting of 500 μL water, 500 μL ethanol and 80 μL 5.0 wt.% Nafion solution. The above mixture was then sonicated for 1 h to form a homogeneous suspension. A working electrode was prepared by dropping 5 μL of the suspension onto the surface of a glassy carbon electrode (GCE), which was then dried in air. The electrode surface area is 0.07 cm2 with 0.265 mg cm-2 catalyst loading density. A saturated calomel electrode and a Pt foil were used as the reference electrode and the counter electrode, respectively. Linear-sweep voltammograms (LSV) were carried out at a scan rate of 2 mV S-1 in two electrolytes: one was a 0.5 M H2SO4 aqueous solution, and the other was a 1.0 M (NH4)2SO3 aqueous solution.The turnover frequency (TOF) values were calculated according to the Equation below: TOF = j x A (2F x n) Where, j is the current density obtained at overpotential of 100 mV, A is the surface area of the electrode, F is the Faraday efficiency (96,485 mol-1), and n is the mole numbers of catalysts deposited onto electrodes.Cyclic voltammetry (CV) measurements were performed with scanning rates from 20 to 100 mV s-1 and potential ranges from 0.00 - 0.10 V (vs. RHE) in a 0.5 M H2SO4 solution. Double-layer capacitances (Cdl) were estimated based on current density variation as a linear function of scan rate. Δj = (ja - jc)/2 was obtained at 50 mV vs. RHE. The electrochemically active surface area (ECSA) was determined by the double layer capacitance (Cdl). The following equation was used to calculate ECSA: ECSA (cm2) = Cdl/Cs The specific capacitance (Cs) of a flat surface is usually in the range of 20 ∼ 60 μF cm-2. We assumed Cs was 40 μF cm-2 in the calculation of the ECSA.The ECSA of 20 wt.% Pt/C was calculated using the under-potential deposition hydrogen (UPD-H) adsorption/desorption voltammetry based on the following equation: E C S A ( cm 2 ) = 0.5 × S H / v 0.21 ( m C · cm − 2 ) Where SH was the integral area of the adsorption/desorption region for H atoms (0.05 V–0.40 V), which was marked red in Figure S5, v is the scan rate.In this research, we also estimated the CB potentials of Ca1-NG and CdS using the Mott−Schottky method. As shown in Figure S16, the slopes of the Mott−Schottky plots for CdS and Ca1-NG are greater than 0.00, suggesting that CdS and Ca1-NG are both n-type semiconductors. Their flat band potentials (Efb) are determined to be −0.58 V and −0.43 V (vs. SCE) for CdS and Ca1-NG, respectively. In general, the CB edge potential (ECB) is more negative by about −0.10 or −0.20 V than the Efb for the n-type semiconductors. Therefore, the ECB for CdS and for Ca1-NG are −0.78 V and −0.63 V (vs. SCE), that is −0.54 V and −0.39 V (vs. NHE) (normal hydrogen electrode). This result indicates that under light irradiation, photogenerated electrons in CdS can migrate from CdS to Ca1-NG at the heterojunction interfaces between Ca1-NG and CdS.Photoelectrochemical properties of catalysts were measured using a CHI 660E electrochemical workstation in a typical three-electrode system. The working electrode was prepared by dropping 50 μL of photocatalyst suspension onto the surface of a fluorine-doped tin oxide (FTO) conducting glass support with an area of 1.0 × 1.0 cm2 and then dried in air. An Ag/AgCl and a Pt foil were used as the reference electrode and the counter electrode, respectively. 0.1 M Na2SO4 aqueous solution was used as the electrolyte, which was purged with N2 to remove dissolved O2. The light source was a single-wavelength (420 nm) LED monochromatic lamp, which was identical to the light source for photocatalytic H2 evolution.XPS analyses were performed using an ESCALAB250 spectrometer equipped with a monochromatized Al Kα (1486.6 eV) source. The survey spectra were recorded in a 0.5 eV incremental with a pass energy of 140 eV. Detailed scans spectra were recorded in a 0.1 eV incremental with a pass energy of 140 eV. The elemental spectra were all corrected with respect to C1s peaks at 284.8 eV.To verify the above EXAFS results a least-squares curve fitting analysis was carried out for the first coordination shell spreading from 1.5 to 2.5 Å. All backscattering paths were calculated based on the structures provided by ab initio simulations. The energy shift (ΔE) was constrained for scatters at the same level. The path length R, coordination number (CN), and Debye–Waller factors σ2 were left as free parameters. The fit was completed in R space with k range of 3.5–12.6 Å−1 and k2 weight.All structures are calculated using density functional theory (DFT) implemented in the Vienna Ab-initio Simulation Package (VASP) (Kresse and Furthmüller, 1996). The exchange-correlation interaction is described by generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional (Perdew et al., 1996). The Ca_sv pseudopotential is used. The vdW interaction is considered by using the DFT-D3 method (Grimme et al., 2010) and spin-polarization effect is included. The electron wavefunctions are expanded using plane waves with an energy cutoff of 400 eV. Slab model is used for all calculations with a fixed cell thickness of 15 Å to ensure sufficient vacuum space. All structures are relaxed until all final residual forces on the atoms are smaller than 0.005 eV/Å. They are built from a graphene unit cell with lattice constant of 2.467 Å, as relaxed using the above parameters with a k-point mesh of 12 × 12×1. A supercell of 4 × 4×1 and k-mesh of 3 × 3×1 are employed for all structures.ΔG H∗ includes three parts: the difference in electronic energy ΔE H, the difference in zero point energy ΔE ZPE, and the difference in entropy TΔS H (Equation 1) ΔG H∗ = ΔE H∗ + ΔE ZPE – TΔS H∗. All the differences are between H in the adsorbed phase (H∗) and in the gas phase (H2). The vibrational frequency in H2 is much higher than in H∗ phase, so ΔS H mainly results from the H2 molecule, namely, TΔS H ∼ 0.5×TS H2 ∼ 0.205 eV at the standard condition (300 K, 1 bar) (Nørskov et al., 2005). The difference in zero point energy is usually very small. For example, ΔE ZPE is around 0.02 eV for H adsorbed onto the double-coordinated N of graphitic-C3N4 (Gao et al., 2015) and around 0.035eV for H adsorbed onto Cu (111) surface (Nørskov et al., 2005). Here, we use these two values for H adsorbed on pyridinic-N and Ca single atom, respectively. In particular, we use (Equation 2) ΔG H∗ = ΔE H∗ + 0.23 eV for H adsorbed on pyridinic-N (Equation 3) ΔG H∗ = ΔE H∗ + 0.24 eV for H adsorbed on Ca The major contribution to ΔG H∗ is the H adsorption energy, calculated as (Equation 4) ΔE H∗ = E(catalyst + mH) – E(catalyst+(m-1)H) – 0.5×E(H2), where E(catalyst + mH) and E(catalyst+(m-1)H) refer to the total energies of the catalytic system with and without the adsorbed H that we are studying; E(H2) is the total energy of a gas phase H2 molecule. These three structures are all with the same supercell size and sufficiently relaxed. When more than one H is adsorbed onto the structure we consider the adsorption of H atoms one by one. In other words, when we consider the m th H atom, we use the structure with m-1 H atoms as the reference system.Similarly to defining the H adsorption energy in Equation (4), we define the Ca adsorption energy as (Equation 5) ΔE Ca = E NG+Ca – E NG - E isolated_Ca_atom where E NG+Ca and E NG refer to the total energies of the N-doped graphene (NG) with or without Ca adsorption, and E isolated_Ca_atom is the total energy of an isolated single Ca atom.This work was financially supported by the Natural Science Foundation of Shanghai (19ZR1420200), Science and Technology Commission of Shanghai Municipality (19DZ2271100), and Shanghai Committee of Science and Technology (17DZ2282800). The authors thank Prof. Song Hong from Beijing University of Chemical Technology for his help on the electron microscopy characterization at the atomic level. S.L. acknowledges the postdoc fellowship provided by Agency for Science Technology and Research (A∗STAR) of Singapore. The computations in this paper were performed on the Odyssey cluster supported by the FAS Division of Science, Research Computing Group at Harvard University. S.L. also thanks Prof. Efthimios Kaxiras for helpful discussions.W.Y. designed the research. J.S., Q.Z., and Q.W. performed the syntheses, most of the structural characterizations, electrochemical and photocatalytic tests. S.L. and W.C. performed DFT simulations. The paper was co-written by W.Y., S.L., and C.H. The research was supervised by W.Y. and Q.X. All authors discussed the results and comments on the manuscript.The authors declare no competing interests.Supplemental information can be found online at https://doi.org/10.1016/j.isci.2021.102728. Document S1. Figures S1–S17 and Tables S1–S5 The following reference appears in the Supplemental information: Chen et al., 2019; Gopannagari et al., 2017; Hu et al., 2019; Irfan et al., 2019; Li et al., 2019a; Li et al., 2019b; Liu et al., 2020; Ran et al., 2017; Rangappa et al., 2020; Ruan et al., 2020; Sun et al., 2020; Wang et al., 2020; Ye et al., 2019; Zhang and Jin, 2019; Zhang et al., 2019; Zhang et al., 2020.

Calcium is one of the most abundant and cheapest elements on earth. However, due to the lack of d-orbitals for chemical adsorption, it is generally considered as a stoichiometric reagent with no catalytic activities in heterogeneous catalysis. In this research, we have revealed that atomically confined Ca in nitrogen-doped graphene (Ca1-NG) can be an effective heterogeneous catalyst to boost both electrocatalytic and photocatalytic hydrogen evolution reactions (HER). Ca single atoms anchored in NG can efficiently enhance the HER performance due to the improvement of the interfacial charge transfer rate and suppression of the photo-generated charge recombination. Density functional theory calculations show that the high catalytic activity of Ca1-NG results from the Ca single atoms in NG, which leads to multiple H adsorption configurations with favorable ΔGH∗ values for HER. This research can be valuable for the designing of environmentally friendly, economical and efficient catalysts for renewable hydrogen production.

Nowadays, the increasing environmental pollution and emissions of greenhouse gases stemming from the fast depletion of fossil resources imposed an increasing attention for developing efficient transformations of abundant and renewable lignin-based biomass resources. In this regard, lignin derivatives can be converted into high value-added chemicals and biofuels through the catalytic hydrodeoxygenation (HDO) process to partly replace at least non-renewable fossil resources [1–3]. For instance, anisole is often considered as a representative model substrate of lignin derivatives. For the catalytic HDO of anisole, due to their high catalytic activity, numerous metal oxides supported noble metal heterogeneous catalysts (e.g., Ru [4,5], Re [6], Pd [7], and Pt [8]) were widely explored. In addition, some transition metals and metal phosphides (e.g., Ni [9], Mo [10], CoMo [11], and Ni2P [12]) were investigated. Despite the high selectivities to deoxygenated products obtained, HDO processes are usually conducted under harsh reaction conditions, namely: high reaction temperatures (T > 250 °C), high hydrogen pressure (3–6 MPa), and high metal loading amounts. Additionally, there exist some disadvantages of high catalyst cost and difficulty in catalyst reusability in some cases. Therefore, despite numerous research efforts, it is a huge challenge to develop highly efficient heterogeneous catalysts for the HDO of lignin derivatives.On the other hand, reducible Co3O4 is often used as a heterogeneous catalytic material [13,14]. Specifically, surface defects on reducible Co3O4 may benefit the activation of oxygen-containing functional groups, thereby promoting the catalytic performance of Co3O4-based catalysts. Recently, metal oxide-based macroporous materials have been applied in the fields of adsorption and catalysis (e.g., environmentally benign oxidation of volatile organic compounds [15,16]), because of their higher porosity and mass transfer rates, and larger concentration of surface defects, in comparison to bulk materials. However, there have been no reports on the development of macroporous Co3O4 supported noble metal catalysts for catalytic hydrogenation applications.In this communication, we fabricated a new three-dimensional (3D) ordered macroporous Co3O4-supported Ru catalyst (Ru/OM-Co3O4). Subsequently, he latter catalytic structure was investigated in the HDO of anisole under mild reaction conditions (i.e., 250 °C and 0.5 MPa hydrogen pressure). For comparison, the HDO reaction was also performed over Ru supported on other 3D ordered macroporous NiO and Al2O3 structures (OM-NiO and OM-Al2O3). The results showed that the present Ru/OM-Co3O4 catalyst could attain a much higher cyclohexane selectivity of 92.4% at a complete anisole conversion, compared with the commercial Co3O4, OM-NiO and OM-Al2O3 supported catalysts. The high catalytic efficiency of Ru/OM-Co3O4 was associated with both the beneficial activation of oxygen-containing groups in anisole at the surface oxygen vacancies of the OM-Co3O4 support, and the presence of highly dispersed Ru NPs, as well as the full exposure of active reaction/adsorption sites and favorable mass transfer related to the ordered macroporous structure of OM-Co3O4 support.Polymethyl methacrylate (PMMA) template beads were prepared by emulsion polymerization [17]. OM-Co3O4 support was fabricated by a sacrificial hard template method, and the resulting OM-Co3O4 supported Ru sample having a Ru loading of about 1.1 wt% was synthesized by the liquid-phase reduction process using sodium borohydride as reductant (see details in the Electronic Supporting Information, ESI). Other 3D ordered macroporous NiO and Al2O3 (denoted as OM-NiO or OM-Al2O3) and resulting supported Ru catalyst samples were synthesized according to identical procedures to those for OM-Co3O4 and Ru/OM-Co3O4 samples.Samples were characterized by X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS), H2 temperature programmed desorption (H2-TPD), H2 temperature programmed reduction (H2-TPR), and Raman spectroscopy (see details in the Electronic Supporting Information, ESI).Details of catalytic HDO tests are included in the Electronic Supporting Information (ESI).As presented in Fig. S1 (ESI), XRD patterns for Ru/OM-Co3O4 sample exhibit several diffraction peaks, which match well those of the cubic Co3O4 spinel phase (JCPDS 42–1467). No diffractions corresponding to metallic Ru0 phase were observed, mainly owing to the small size (< 4 nm) of Ru0 particles and the low Ru loading used (~ 1.1 wt%) determined by ICP-AES analysis. The results reflect the good dispersion of Ru species on the surface of OM-Co3O4 support. As shown in Fig. S2 (ESI), Ru/OM-Co3O4 displays a 3D honeycomb-like ordered macroporous structure, in which hollow spheres are interconnected together through walls. TEM images of Ru/OM-Co3O4 (Fig. 1 ) depict that the large quantities of small Ru NPs with an average diameter of ~2.53 nm are uniformly attached on the surface of a nearly uniform OM-Co3O4 support, thereby forming the close interface between them. This is well consistent with the powder XRD results. Meanwhile, one can discern the lattice fringes of the (101) and (511) crystal facets of Ru0 and Co3O4 phases with facet spacing of 0.206 and 0.155 nm, respectively. In this case, both the ordered macroporous framework of OM-Co3O4 and the high dispersion of Ru NPs may favor full exposure of the adsorption and reaction active sites. In contrast, in addition to a few aggregates of particles, larger Ru particles with the size of ~15–20 nm are found to be distributed over the surface of the commercial Co3O4- supported Ru sample (Fig. S3, ESI). The above results illustrate that OM-Co3O4 support shows a promotional effect on the improvement of the dispersion of Ru and the formation of smaller Ru NPs.The structural defects can be easily generated through calcination or reduction treatments during the synthesis of supported catalysts, and further promote their catalytic performance of catalysts [18,19]. Therefore, XPS characterization was performed to identify surface electronic states of metal and oxygen species on Ru-based samples (Fig. 2 ). In the XPS of Ru 3d5/2 region for the Ru/OM-Co3O4 and Ru/Co3O4 samples, a peak with a binding energy at ~280.3 eV is observed, minoring the presence of metallic Ru0 species (Fig. S4, ESI). In the deconvoluted Co 2p region, Co 2p3/2 and Co 2p1/2 core levels appear at 777–785 and 792.5–801 eV, respectively, indicative of the presence of Co2+ and Co3+ species [20]. Notably, the surface fraction of Co2+ in the total Co species on the Ru/OM-Co3O4 (0.51) is larger than that on Ru/Co3O4 (0.42), reflecting the formation of more defective Co2+ sites. Meanwhile, XPS of the O 1 s region depicts the existence of three kinds of oxygen species at ~529.6, 531.3 and 532.8 eV, respectively, which correspond to lattice oxygen (OI), oxygen species adsorbed on defects (e.g., oxygen vacancies) or hydroxyl species (OII), and surface carbonate ions (OIII) [21]. Noticeably, the surface OII/(OI + OII + OIII) fraction on the Ru/OM-Co3O4 (0.48) is higher than that on the Ru/Co3O4 (0.41), likely suggestive of the generation of more oxygen vacancies.Raman spectra provide insight into the defective crystal structures. As illustrated in Fig. S5 (ESI), compared with those for Ru/Co3O4, five characteristic Raman peaks (F2g 1, E2g, F2g 2, F2g 3, and A1g) of Co3O4 phase for the Ru/OM-Co3O4 solid all shift to low frequencies at the 532-nm laser wavelength, despite the reduced peak intensities. These results demonstrate the presence of lattice distortion/strain of Co3O4 spinel phase, and thus the formation of more Co2+-Ov-Co2+ like structural defects (Ov: oxygen vacancies) in the vicinity of Co2+ species on the Ru/OM-Co3O4 [22,23], mainly thanks to the multiple calcination processes conducted during the synthesis of OM-Co3O4. Fig. 3 shows the variation of anisole conversion and product distribution with reaction time after HDO reaction at 250 °C and 0.5 MPa over the Ru/OM-Co3O4 catalyst. The main deoxygenated products are benzene (BEN) and cyclohexane (CHA), along with the formation of small amounts of methoxycyclohexane (MCHA) and cyclohexanol (CHOL) by-products, and trace amounts of cyclohexanone (CNON) and phenol. With prolonged reaction time, the benzene selectivity gradually decreases, whereas the cyclohexane selectivity progressively increases. Besides a small amount of benzene with a yield of 4.2%, a large amount of cyclohexane with a high yield of 92.4% is obtained at complete conversion after 5 h of reaction. Over the pure OM-Co3O4 support, almost no conversion of anisole was obtained. As shown in Table 1 (entry 1), Ru/Co3O4 exhibits a low catalytic activity for the HDO reaction with a much lower conversion, ca. 38.3% and a lower selectivity to deoxygenated products (ca. 66.3%) after 1 h. Notably, compared to Ru/Co3O4, the catalytic HDO performance of Ru/OM-Co3O4 is significantly improved, along with high conversion (90.2%) and selectivity to deoxygenated products (87.7%) (Table 1, entry 2). Compared with Ru/OM-Co3O4, the other two Ru/OM-NiO and Ru/OM-Al2O3 reference catalysts deliver lower conversions and selectivity to deoxygenated products (Table 1, entries 3 and 4).Since the activity and selectivity to deoxygenation products over the Ru/Co3O4 are much inferior to those over the Ru/OM-Co3O4 catalyst in the HDO process, one can confirm that surface Ru species and OM-Co3O4 support should play important roles in controlling the HDO process of anisole. In the present Ru/OM-Co3O4 catalyst, OM-Co3O4 support with a higher surface area of 24.3 m2/g can effectively serve as a support for achieving higher dispersion of small-sized Ru NPs, compared with the commercial Co3O4 with smaller specific surface area, ca. 4.5 m2/g. Further, TEM observations reveal the formation of highly dispersed and small-sized Ru NPs on the OM-Co3O4, which greatly facilitate the accessibility of active metallic Ru sites to substrates, and thus the dissociation of molecular hydrogen. Meanwhile, XPS and Raman results demonstrate the presence of more defects on Ru/OM-Co3O4, in comparison to Ru/Co3O4. Furthermore, the present Co3O4-supported Ru catalysts should be activated by H2 at high temperature before testing, and correspondingly the Co3O4 support could be reduced partly, as evidenced by the XPS analysis of the used catalysts (Fig. S6, ESI), thus leading to the increased surface Co2+/(Co2++Co3+) ratio (0.56 for Ru/OM-Co3O4 and 0.47 for Ru/Co3O4), and the OII fraction (0.65 for Ru/OM-Co3O4 and 0.5 for Ru/Co3O4) after the HDO reaction. Such abundant surface defective structures probably lead to the easier activation of methoxy group in anisole through the interaction between defective Co2+ species and oxygen atom of the methoxy group, and thus direct deoxygenation process to form benzene [5]. Subsequently, a further ring‑hydrogenation of benzene can produce cyclohexane.In this work, we further carried out H2-TPD experiments to determine the ability of H2 dissociation and the occurrence of hydrogen spillover on Ru-based catalyst samples. As presented in Fig. S7 (ESI), in the case of Ru/OM-Co3O4 sample, two desorption peaks located at ~92 and 300 °C correspond to desorption of hydrogen from Ru particles and highly dispersed Ru species strongly interacting with the support, respectively. The desorption peak at 535 °C is assigned to spillover hydrogen adsorbed on the support [24,25]. In contrast, no hydrogen spillover occurs on the Ru/Co3O4, besides a remarkably reduced desorption originating from the absence of highly dispersed Ru species. What's more, H2-TPR traces display that compared to pure OM-Co3O4 (386 and 501 °C, respectively), Ru/OM-Co3O4 exhibits lower reduction temperatures for Co3+ to Co2+ and Co2+ to Co0 species (360 and 489 °C, respectively, Fig. S8, ESI), confirming the occurrence of a much stronger hydrogen spillover from highly dispersed Ru0 species on the Co3O4 surface of the Ru/OM-Co3O4 catalyst.Therefore, thanks to the highly dispersive character of Ru species on the OM-Co3O4 support, and its abundant defective structure, both H2 dissociation and hydrogen spillover take place more easily on the Ru/OM-Co3O4 than on the Ru/Co3O4, thereby, significantly improving the catalytic HDO performance of Ru/OM-Co3O4. Also, the unique macroporous structure of OM-Co3O4 support provide a large number of open channels, which may likely favor the exposure of active reaction/adsorption sites and the facile diffusion of reactants and products, thereby promoting the catalytic HDO performance of Ru/OM-Co3O4 to some extent. It can be concluded that the higher catalytic performance of Ru/OM-Co3O4 should be closely associated with the surface cooperation between highly dispersed Ru NPs and defective sites present in the OM-Co3O4 support, as well as the unique macroporous framework of OM-Co3O4 support.The influence of reaction temperature and hydrogen pressure on the HDO of anisole was also investigated over the Ru/OM-Co3O4 catalytic system. As presented in Fig. S9 (ESI), with the elevated reaction temperature from 200 to 275 °C, the anisole conversion gradually increases from 42.5 to 98.6%, while the selectivity to deoxygenated products (benzene and cyclohexane) progressively increases from 44.7 to 94.7%. These results demonstrate that the high reaction temperature can promote the HDO of anisole. Fig. S10 (ESI) shows that benzene selectivity becomes quite low (<1.0%) above 1.0 MPa of hydrogen pressure. This is because the high hydrogen pressure favors the hydrogenation of benzene ring to form methoxycyclohexane and cyclohexane, thus inhibiting the direct cleavage of the methoxy group to form benzene. As summarized in Table S1 (ESI), compared to Ru-based catalysts previously reported, the present Ru/OM-Co3O4 catalytic system possesses better or at least comparative catalytic performance in the anisole HDO under mild reaction conditions.The stability of heterogeneous catalysts is one of the key indexes for their practical application. As displayed in Fig. 4 , the selectivity to each product is almost unchanged, and the conversion is only decreased by ~1.3% after five successive HDO tests using the Ru/OM-Co3O4 catalytic system. Further, SEM and TEM images of the used catalyst (Fig. S11, ESI) reveal that the macroporous structure of OM-Co3O4 is kept unchanged, and no structural collapse occurs. It is indicated that Ru/OM-Co3O4 catalyst has good structural stability and reusability, mainly thanks to the strong interactions developed between Ru NPs and OM-Co3O4 support.In summary, we synthesized a new supported Ru catalyst on the OM-Co3O4 carrier, and utilized the OM-Co3O4 with appropriate surface defects to enhance the dispersion of small-sized Ru nanoparticles and create strong metal-support interactions. The Ru/OM-Co3O4 could afford a 92.4% yield of cyclohexane in the anisole HDO under mild reaction conditions (0.5 MPa hydrogen pressure and 250 °C), indicative of high activity and selectivity to deoxygenated products (benzene and cyclohexane). The formation of highly dispersed Ru species and more surface defects could favor the adsorption and activation of reactants on the catalyst surface. A significant diffusion behavior of reactants and products originating from the unique macroporous framework structure of OM-Co3O4 is very likely to account for the high catalytic efficiency of Ru/OM-Co3O4. It is expected that the present reported approach using OM-Co3O4 as catalyst support is novel and reproducible, and would be a promising approach for designing other high-performance supported catalysts applied in several other heterogeneous catalytic processes.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 thank the National Natural Science Foundation of China (21776017;21991102; U19B6002) for financial support. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2021.106302.

A three-dimensional ordered macroporous Co3O4 (OM-Co3O4) supported Ru catalyst was developed for the efficient hydrodeoxygenation (HDO) of anisole. It is revealed that small-sized Ru nanoparticles evenly distributed over the surface of OM-Co3O4 with large quantities of oxygen vacancies could strongly capture Ru0 species, thereby resulting in strong Ru-Co3O4 interactions. Compared with commercial Co3O4 supported Ru catalyst, Ru/OM-Co3O4 displays a better catalytic HDO performance, with a high cyclohexane yield of 92.4% at 250 °C and 0.5 MPa hydrogen pressure after 5 h on stream. Such a significant efficiency of Ru/OM-Co3O4 is mainly attributed to both high dispersion of Ru0 species and an enhanced formation of surface defects, as well as the unique macroporous framework of OM-Co3O4 support.

The oligomerization of linear terminal alkenes is one of the significant problems in the linear α-olefins (LAOs) production in both industry and academia [1–6], together with hydrogenation of alkanes [7]. Linear alpha olefins, normally produced by ethylene oligomerization processes [8–11], and by Fischer–Tropsch synthesis followed by purification [12], found applications as co-monomers in high density and linear low density polyethylene (HDPE and LLDPE) production, and as detergents, plasticizers, surfactants, and lubricants.[13,14]. Traditional technology for LAO synthesis is based on a full-range production process in which ethylene oligomerizes to achieve a broad range of the products. It is a non-selective approach and cannot match the constantly growing market demand. Accordingly, the change of a statistical ethylene oligomerization process into selective approach appears highly demanded [13–15]. The selective oligomerization of ethylene has recently attracted considerable attention [16,17]. In this regard, the catalytically selective trimerization of ethylene to 1-hexene has been extensively studied [18–23]. Amongst all the systems, the catalysts based on chromium metal has attracted more attention in the recent years [23,24]. This metal is the main center of Phillips’ Cr-pyrrolide catalysts [25]. BP's (o-OMe)PNP catalysts [26], Sasol's PNP/SNS trimerization catalysts [27,28], and PNP tetramerization catalysts [29]. Catalysts based on other transition metals such as Zr, Ti, V, Ta, or Ni have been less studied [30,31].Hessenʼs group in 2001 for the first time reported that the change of R from methyl to phenyl in [(η 5‐C5H4C(Me)2RTiCl3]/MAO, switches the reaction from ethylene polymerization into ethylene trimerization and facilitates formation of 1-hexene as the major product. The hemilabile behavior of cyclopentadienyl ligand with the arene group is the main reason of this significant change in catalyst behavior [32]. Deckers et al. in 2002 synthesized a new family of highly active catalysts for the trimerization of ethylene based on (arene-cyclopentadienyl) titanium complexes [(ƞ 5-C5H3R-(bridge)-ArTiCl3] activated by MAO co‐catalyst. Selectivity to produce 1-hexene not only depends on the presence of the arene pendant group but also the bridge nature between cyclopentadienyl (Cp) and arene. In the absence of arene, polyethylene was the main product [33,34]. Huang's group synthesized a half‐sandwich titanium complex containing pendant thienyl group and used it in ethylene trimerization. In the reported results, they affirmed the important role of thiophene in ethylene trimerization [35]. Cp-based ligands have been widely studied as important ligands in organometallic chemistry and most studies on Cp modification can be focused on the type of the bearing pendant group on it. In this regard, in 2004, Huang et al. used half-sandwich titanium complexes with the pendant ethereal group activated by MAO for ethylene trimerization [36]. In 2013, Zhang et al. synthesized half-sandwich indenyl-based titanium complexes [Ind-(bridge)-Ar]TiCl3 bearing pendant arene group on the indenyl ring and examined the selective ethylene trimerization in the presence of MAO co-catalyst [37–39]. In the following of previous works, Zhang et al. synthesized another series of half-sandwich indenyl-based titanium complexes with the thienyl group (Cp(Ind)-bridge-thienyl]TiCl3, which showed high selectivity in ethylene trimerization and its conversion to 1-hexene [23]. In 2015, Varga et al. synthesized and characterized two titanium‐based heterogeneous catalysts using different methods including grafting through a covalent Ti-O-Si bond as well as through a pendant flexible tether from the Cp ligand [40]. The catalyst synthesized by the second method did not show any activity in ethylene trimerization because the active species were too close to the support surface [40]. In 2015, Duchateau et al. synthesized different types of phenoxy-imine titanium catalyst according to the Fujita method and examined homogeneous and heterogeneous types [41]. Despite the high activity and selectivity of the synthesized titanium catalysts, very little polyethylene was produced as a by-product. In this work, an attempt was also made to stabilize the catalyst and prevent the formation of the polymer, while maintaining the desired catalyst activity and selectivity. In this line, MAO co-catalyst and phenoxy-imine titanium catalysts were stabilized using a two-step process on silica carrier [41]. In 2019, Mohamadnia et al. successfully synthesized and identified three titanium-based catalysts {[ƞ 5-C9H6-C(R)]-C4H3S}TiCl3 with different bridges (cyclohexane, cyclopentane and dimethyl) active in ethylene trimerization [29]. Factors affecting the catalyst activity in the production of 1-hexene including catalyst concentration, ethylene pressure, and reaction temperature were optimized [29].In the following of our research on the selective trimerization of ethylene, due to the high importance of α-olefins in the petrochemical industry, here a series of indenyl half‐sandwich titanium complexes, namely [Ind-C(R)-phenyl]TiCl3 was synthesized for possible application in ethylene trimerization process. The main aim is to fulfill the process at mild operating conditions such as low pressure and temperature to reduce operator-related risks, and to reach high economic efficiency by reducing catalyst consumption. In this regard, the effect of the arene bridge and indenyl ring, temperature, ethylene pressure, and MAO and catalyst concentrations on the catalytic efficiency was investigated. Furthermore, the effect of ligand type on the 1-hexene selectivity was also examined using DFT simulations by considering the energy path for each catalyst (C1-C4, see Scheme 1 ), during ethylene oligomerization process [42–44].All manipulations of water- and/or air-sensitive compounds were performed using standard Schlenk and glove-box techniques under deoxygenated argon or nitrogen. The modified MAO (MMAO) co‐catalyst (7 wt% in toluene), titanium tetrachloride (TiCl4), n‐butyllithium (n‐BuLi; 2.5 M in n‐hexane), phenyllithium (1.5 M in dibutyl ether), and molecular sieve were obtained from Aldrich (Germany). Indene, pyrrolidine, 4-tert-butylcyclohexanone, cyclohexanone, cycloheptanone, acetone, ethanol, n‐hexane, diethyl ether, toluene, Na2CO3, MgSO4, NaCl, Na2SO4, sodium, and NaOH were purchased from Merck (Germany). Ethylene was provided by Bandar Imam Petrochemical Company (Iran) and purified by passing through NaOH, activated silica gel, and molecular sieve (3 Å) columns, respectively. Methanol, n‐hexane, toluene and diethyl ether were dried and vacuum-distilled using calcium hydride (CaH2) and sodium metal consecutively before use.The 1H NMR and 13C NMR spectra have been recorded by the Bruker 400 MHz Ultra shield NMR instrument (Germany) at room temperature. The progress of the catalyst synthesis and trimerization reactions was followed by thin-layer chromatography (TLC) and gas chromatography GC system (Varian CP 3800), respectively. The inductively coupled plasma analysis (ICP), model 3410 ARL made in Switzerland, was used to determine the metal components of the catalyst. The UV-visible spectrophotometer (Pharmacia Biotech Ultrospec 4000) was used to further examine the spectral characteristics of synthetic complexes. Elemental analysis was performed using a Vario EL III CHNS elemental analyzer.The different fulvene precursors F1–F4 were synthesized with a slight change according to the method proposed by Stone and Little (Scheme 2 ) [45]. For this purpose, freshly distilled indene (5 mmol, 0.58 mL) and freshly distilled pyrrolidine (3 mmol, 0.25 mL) were dissolved in 2 mL of methanol under argon atmosphere at ambient temperature. Then, different ketones)2 mmol(, such as cyclohexanone, cycloheptanone, 4-t-butyl cyclohexanone, and acetone, were added dropwise to the stirred solution and the reaction mixture was stirred for 12 h. At the end of the reaction, acetic acid (3 mmol, 0.18 mL) was added to neutralize the residual base, and dilution was performed with diethyl ether (10 mL). To separate the remaining indene, and other unreacted materials, extraction was performed with deionized water (3 × 10 mL) followed by brine (2 × 10 mL). Finally, the water remaining in the organic phase was dried by anhydrous MgSO4. Synthesized fulvenes were purified using column chromatography by silica gel (petroleum ether as eluent). The pure fulvenes were characterized using 1H NMR, 13C NMR, and FT‐IR spectroscopies.Fulvene (C15H16, F1) was obtained as white crystals in 70% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 1.70–1.79 (2H, m, C9H6-C(cyclo‐C5 H 10)), 1.79–1.9 (4H, m, C9H6-C(cyclo‐C5 H 10)), 2.75 (2H, t, C9H6-C(cyclo‐C5 H 10)), 3.05 (2H, t, C9H6-C(cyclo‐C5 H 10)), 6.79 (1H, d, C9 H 6-C(cyclo‐C5H10)), 6.94 (1H, d, C9 H 6-C(cyclo‐C5H10)), 7.17–7.27 (2H, m, C9 H 6-C(cyclo‐C5H10)), 7.37 (1H, d, C9 H 6-C(cyclo‐C5H10)), 7.9 (1H, d, C9 H 6-C(cyclo‐C5H10)). 13C NMR (100 MHz, CDCl3, δ, ppm): 26.40, 28.10, 28.76, 32.29, 34.48 (CH2), 121.05, 123.73, 124.64, 126.01, 127.30, 128.25 (CH), 133.65, 135.87, 144.56, 152.46 (C q). FT‐IR (KBr, υmax, cm−1): 3010, 3014 and 3064 (sp2 C-H), 2848 and 2921 (sp3 C-H), 1780–1930 (overtone of aromatic ring), 1619 (C=C), 1443 (CH2), 723 and 740 (=C-H) (Figs. S1–S3).Fulvene (C12H12, F2) was obtained as a yellow oil in 70% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 2.31 (3H, s, C9H6-C(CH 3)2), 2.54 (3H, s, C9H6-C(CH 3)2), 6.84 (1H, d, C9 H 6-C(CH3)2), 6.91 (1H, d, C9 H 6-C(CH3)2), 7.25–7.32 (2H, m, C9 H 6-C(CH3)2), 7.41 (1H, d, C9 H 6-C(CH3)2), 7.82 (1H, d, C9 H 6-C(CH3)2). 13C NMR (100 MHz, CDCl3, δ, ppm): 22.84, 25.00 (CH3), 121.03, 123.52, 124.71, 126.02, 127.61, 128.35 (CH), 135.74, 136.70, 143.38, 143.97 (Cq). FT‐IR (KBr, υmax, cm−1): 3600 (adsorbed water), 3030–3090 (sp2 C-H), 2913 and 2854 (sp3 C-H), 1930–1780 (overtone of aromatic ring), 1630 (C=C), 1450 (CH2), 727 and 750 (=C-H) (Figs. S4–S6).Fulvene (C19H24, F3) was obtained as yellow oil in 75% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 0.9 (9H, s, C9H6-C(4-tBu-cyclo-C5H9)), 1.4 (3H, m, C9H6-C(4-tBu-cyclo‐C5 H 9)), 2.2 (2H, t, C9H6-C(4-tBu-cyclo‐C5 H 9)), 2.4 (2H, m, C9H6-C(4-tBu-cyclo‐C5 H 9)), 3.2 (1H, d, C9H6-C(4-tBu-cyclo‐C5 H 9)), 3.8 (1H, d, C9H6-C(4-tBu-cyclo‐C5 H 9)), 6.86 (1H, d, C9 H 6-C(4-tBu-cyclo‐C5H9)), 6.96 (1H, d, C9 H 6-C(4-tBu-cyclo‐C5H9)), 7.24-7.30 (2H, m, C9 H 6-C(4-tBu-cyclo‐C5H9)), 7.40 (1H, d, C9 H 6-C(4-tBu-cyclo‐C5H9)), 7.95 (1H, d, C9 H 6-C(4-tBu-cyclo‐C5H9)). 13C NMR (100 MHz, CDCl3, δ, ppm): 27.6 (CH3), 28.1, 29.3, 31.5, 33.8 (CH2), 47.5, 121.4, 124, 124.8, 126.1, 127.5, 128.3 (CH), 32.8, 133.6, 136, 144.8, 152.1 (Cq). FT‐IR (KBr, υmax, cm−1): 3040–3090 (sp2 C-H), 2870 and 2956 (sp3 C-H), 1940–1780 (overtone of aromatic ring), 1627 (C=C), 1446 (CH2), 727 and 748 (=C-H) (Figs. S7–S9).Fulvene (C16H18, F4) was obtained as yellow oil in 75% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 1.68 (4H, m, C9H6-C(cyclo‐C6 H 12)), 1.86 (2H, m, C9H6-C(cyclo‐C6 H 12)), 1.97 (2H, m, C9H6-C(cyclo‐C6 H 12)), 2.93 (2H, t, C9H6-C(cyclo‐C6 H 12)), 3.17 (2H, t, C9H6-C(cyclo‐C6 H 12)), 6.87 (1H, d, C9 H 6-C(cyclo‐C6H12)), 6.95 (1H, d, C9 H 6-C(cyclo‐C6H12)), 7.25–7.32 (2H, m, C9 H 6-C(cyclo‐C6H12)), 7.41 (1H, d, C9 H 6-C(cyclo‐C6H12)), 7.81 (1H, d, C9 H 6-C(cyclo‐C6H12)). 13C NMR (100 MHz, CDCl3, δ, ppm): 26.2, 28.6, 28.8, 29.5, 34.5, 34.9 (CH2), 121.2, 123.6, 124.7, 124.9, 126, 127.2 (CH), 128.3, 136.1, 144.1, 154 (C q). FT‐IR (KBr, υmax, cm−1): 3010–3090 (sp2 C-H), 2852 and 2921 (sp3 C-H), 1790-1940 (overtone of aromatic ring), 1627 (C=C), 1448 (CH2), 721 and 750 (=C-H) (Figs. S10–S12).Indenyl‐based ligands were prepared according to a slightly modified literature method (Scheme 2) [26]. Solution of synthetic fulvenes derivatives (0.5 mmol) in diethyl ether (3 mL) was added dropwise to the phenyllithium solution in dibuthyl ether (2 mmol, 1.3 mL, 1.9 M) in 5 mL of dry diethyl ether under argon atmosphere at -40 °C. The mixture was stirred at room temperature for 12 h. After one day, the reaction mixture was hydrolyzed by 10 mL of cold water. The aqueous layer was extracted with light petroleum ether (three times), and the organic layer was dried with anhydrous MgSO4. The solvent was removed under vacuum. The obtained L1–L4 ligands were purified using column chromatography via petroleum ether as eluent.Ligand (C21H22, L1) was obtained as a white solid in 91% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 1.4–1.55 (1H, m, [C9H7-C(cyclo‐C5 H 10)]-C6H5), 1.51–1.66 (5H, m, [C9H7-C(cyclo‐C5 H 10)]-C6H5), 2.23–2.25 (2H, m, [C9H7-C(cyclo‐C5 H 10)]-C6H5), 2.39–2.42 (2H, m, [C9H7-C(cyclo‐C5 H 10)]-C6H5), 3.44 (2H, d, [C9 H 7-C(cyclo‐C5H10)]-C6H5), 6.56 (1H, t, [C9 H 7-C(cyclo‐C5H10)]-C6H5), 7.02–7.1 (3H, m, [C9 H 7-C(cyclo‐C5H10)]-C6 H 5), 7.15–7.21 (1H, m, [C9H7-C(cyclo‐C5H10)]-C6 H 5), 7.28–7.31 (2H, m, [C9H7-C(cyclo‐C5H10)]-C6 H 5), 7.4–7.49 (3H, m, [C9 H 7-C(cyclo-C5H10)]-C6H5). 13C NMR (100 MHz, CDCl3, δ, ppm): 23.03, 26.61, 36.46, 37.48 (CH2), 44.55 [C9H7- C(cyclo-C5H10)]-C6H5, 122.35, 123.64, 123.86, 125.35, 125.67, 127.06, 128.13, 129.61 (CH), 143.93, 145.22, 147.38, 149.96 (Cq). FT‐IR (KBr, υmax, cm−1): 2974 (sp2 C-H), 2933 and 2921 (sp3 C-H), 1610 (C=C), 1461 (CH2bending), 700–800 (=C-H) (Figs. S13–S15).Ligand (C18H18, L2) was obtained as a yellow oil in 90% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 1.75 (6H, s, [C9H7-C(CH 3)2]-C6H5), 3.47 (2H, d, [C9 H 7-C(CH3)2]-C6H5), 6.54 (1H, t, [C9 H 7-C(CH3)2]-C6H5), 6.73–6.78 (1H, d, [C9 H 7-C(CH3)2]-C6H5), 7.01–7.08 (1H, t, [C9 H 7-C(CH3)2]-C6 H 5), 7.10–7.17 (1H, t, [C9 H 7-C(CH3)2]-C6H5), 7.20–7.26 (1H, t, [C9H7-C(CH3)2]-C6 H 5), 7.27–7.34 (2H, t, [C9H7-C(CH3)2]-C6 H 5), 7.35–7.42 (2H, d, [C9H7-C(CH3)2]-C6 H 5), 7.48–7.52 (1H, d, [C9 H 7-C(CH3)2]-C6H5). 13C NMR (100 MHz, CDCl3, δ, ppm): 29.65 (CH3)2, 37.8 (CH2), 40.5 [C9H7-C(CH3)2]-C6H5, 122.34, 123.70, 123.96, 125.47, 125.82, 126.21, 127.42, 128.29 (CH), 143.79, 145.28, 148.15, 152.35 (Cq). FT‐IR (KBr, υmax, cm−1): 3072 (sp2 C-H), 2800–3000 (sp3 C-H), 1677 (C=C), 1465 (CH2bending), 780 (=C-H) (Figs. S16–S18).Ligand (C25H30, L3) was obtained as a white solid in 92% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 0.8 (9H, s, [C9H7-C(4-tBu-cyclo‐C5H9)]-C6H5), 1.1–1.2 (1H, m, [C9H7-C(4-tBu-cyclo‐C5 H 9)]-C6H5), 1.3–1.53 (2H, m, [C9H7-C(4-tBu-cyclo‐C5 H 9)]-C6H5), 1.6–1.80 (2H, t, [C9H7-C(4-tBu-cyclo‐C5 H 9)]-C6H5), 1.86–2.07 (2H, t, [C9H7-C(4-tBu-cyclo‐C5 H 9)]-C6H5), 2.58–2.85 (2H, m, [C9H7-C(4-tBu-cyclo‐C5 H 9)]-C6H5), 3.5 (2H, d, [C9 H 7-C(4-tBu-cyclo‐C6H9)]-C6H5)), 6.63–6.79 (1H, t, [C9 H 7-C(4-tBu-cyclo-C5H9)]-C6H5), 6.9–7.2 (4H, m, [C9 H 7-C(4-tBu-cyclo‐C6H9)]-C6 H 5), 7.2–7.34 (2H, m, [C9H7-C(4-tBu-cyclo‐C6H9)]-C6 H 5), 7.3-7.5 (3H, m, [C9 H 7-C(4-tBu-cyclo-C5H9)]-C6H5). 13C NMR (100 MHz, CDCl3, δ, ppm): 24.1 (CH2), 27.5 [C9H7-C(4-tBu-cyclo-C5H9)]-C6H5, 32.5, 36.9 (CH2), 37.7 (CH2), 44.5, 48 [C9H7-C(4-tBu-cyclo-C5H9)]-C6H5, 122.5, 123.5, 123.9, 125.3, 125.7, 126.2, 128.2, 131.5 (CH), 144.1, 145.1, 147.5, 148.9 (Cq). FT‐IR (KBr, υmax, cm−1): 3010–3085 (sp2 C-H), 2974 (sp3 C-H), 1606 (C=C),1458 (CH2bending), 700–800 (=C-H) (Figs. S19–S21).Ligand (C22H24, L4) was obtained as a white solid in 90% yield. 1H NMR (400 MHz, CDCl3, δ, ppm): 1.57–1.92 (8H, m, [C9H7-C(cyclo‐C6 H 12)]-C6H5), 2.24–2.50 (4H, m, [C9H7-C(cyclo‐C6 H 12)]-C6H5), 3.48 (2H, d, [C9 H 7-C(cyclo‐C6H12)]-C6 H 5), 6.51–6.55 (1H, t, [C9 H 7-C(cyclo‐C6H12)]-C6H5), 6.75–6.79 (1H, d, [C9 H 7-C(cyclo‐C6H12)]-C6H5), 6.98–7.01 (1H, t, [C9 H 7-C(cyclo‐C6H12)]-C6H5), 7.07–7.10 (1H, t, [C9 H 7-C(cyclo‐C6H12)]-C6H5), 7.15–7.19 (1H, t, [C9H7-C(cyclo‐C6H12)]-C6 H 5), 7.25–7.31 (2H, t, [C9H7-C(cyclo‐C6H12)]-C6 H 5), 7.36–7.38 (2H, d, [C9H7-C(cyclo-C6H12)]-C6 H 5), 7.43–7.54 (1H, d, [C9 H 7-C(cyclo‐C6H12)]-C6H5). 13C NMR (100 MHz, CDCl3, δ, ppm): 24.07, 31.63, 37.36, 39.11 (CH2), 47.80 [C9H7-C(cyclo‐C6H12)]-C6H5, 122.5, 123.58, 123.87, 125.37, 125.57, 126.69, 127.49, 128.22 (CH), 144.09, 145.27, 148.77, 151.83 (Cq). FT‐IR (KBr, υmax, cm−1): 3025 (CHaromatic, sp2 C-H), 2971 and 2861 (CHaliphatic, sp3 C-H), 1610 (C=C), 1467 (CH2bending), 792 (=C-H) (Figs. S22–S24).For the synthesis of the desired catalysts, ligands L1–L4 (0.5 mmol) in dry petroleum ether (5 mL), was added dropwise to a solution of the 0.5 mmol n‐BuLi (0.2 mL, 2.5 M) solution in petroleum ether at -70 °C under an argon atmosphere. After mixing for four hours at the same temperature, white milky salt was obtained. Then, 0.5 mmol (0.05 mL) of TiCl4 was added to the reaction mixture at -78 °C. As soon as the TiCl4 was added, a dark red solution was obtained. The liver red solution was maintained at room temperature for 12 h. The unreacted solvent and TiCl4 were then removed by vacuum and the residue were dissolved in 10 mL of dry petroleum ether and then centrifuged. After cooling the solution containing the catalyst to -20 °C, the catalysts were obtained (Scheme 2). Catalyst C1 was obtained as dark red crystals in 70% yield. Elemental analysis (%): calculated for C21H21TiCl3 (found): C 58.99 (59.30), H 4.95 (5.12), Ti 11.19 (10.53), Cl 24.87 (—). Catalyst C2 was obtained as dark red solid in 65 % yield. Elemental analysis (%): calculated for C18H17TiCl3 (found): C 55.79 (54.90), H 4.42 (4.40), Ti 12.35 (11.95), Cl 27.44 (—). Catalyst C3 was obtained as dark red solid in 68% yield. Elemental analysis (%): calculated for C25H29TiCl3 (found): C 62.08 (62.22), H 6.04 (6.11), Ti 9.90 (9.12), Cl 21.99 (—). Catalyst C4 was obtained as dark red solid in 73% yield. Elemental analysis (%): calculated for C22H23TiCl3 (found): C 59.83 (59.92), H 5.25 (5.22), Ti 10.84 (9.93), Cl 24.08 (—).The ethylene trimerization reactions were performed using synthetic titanium catalysts in a pressurized steel reactor equipped with a mechanical stirrer. The pressure and temperature inside the reactor and the mixer speed were controlled by the reactor's digital displays. In this regard, the reactor was first purged with dried pure argon at 120 °C for 2 h. It was reached the desired temperature and the solvent and MMAO were injected subsequently. After 10 min, the solution containing catalyst/solvent was injected and trimerization was started by charging the reactor with ethylene monomer. The reaction temperature and ethylene pressure were fixed constant throughout the process. After 30 min, the reactor was cooled to -10 °C. Then, the liquid phase including, 1-C6 and probable by‐products were collected and analyzed using GC instrument. The produced polyethylene by‐product was also washed with acidified ethanol (3% HCl), and dried under vacuum at 60 °C to a constant weight.DFT static calculations were performed at B3LYP level [46] with the Gaussian16 package [47]. The electronic configuration of the system was described with the standard split valence basis set with a polarization function for all the atoms (def2SVP keyword in Gaussian) of Ahlrichs and co-workers [48]. Geometry optimizations were performed without symmetry constrain, and analytical frequency calculations performed the characterization of the local stationary points. These frequencies were used to calculate unscaled zero-point energies as well as thermal corrections and entropy effects at 298 K and 1 atm. The transition states were located using the synchronous transit-guided quasi-Newton (QST3) approach and the extrema have been checked by analytical frequency calculations. All transition states have associated only one imaginary frequency. Solvent effects were estimated in single point energy calculations on the gas phase optimized structures based on the polarizable continuous solvation model (PCM) [49], as implemented in Gaussian16, using toluene (Tol) as a solvent. Energies were obtained using the B3LYP functional [46], in conjunction with the triple-ζ basis set cc-pVTZ for all the atoms [50], together with the Grimme D3 correction term [51] to the electronic energy. The reported free energies in this work include energies obtained at the B3LYP/cc-pVTZ level of theory corrected with zero-point energies, thermal corrections and entropy effects evaluated at 298 K, achieved at the B3LYP/def2SVP level, without translational entropy corrections [52].In the structure of synthetic catalysts, [C9H6-C(R)-C6H5]TiCl3, due to the presence of halide groups and bulky indenyl moiety, the coordination number of Ti is IV [53,54]. In general, for this group of complexes, tetrahedral and square‐planar structures are observed. Due to the presence of bulky ligands in the structure and electrostatic repulsion between them, the complex tends to have a deviated tetrahedral structure. For the latter geometry, the steric hindrance has the minimum value [55], and in agreement the relative stability of the complex increases. According to the valence bond theory, orbitals of the valence layer of the central atom in the tetrahedral structure are hybridized either as sp3 or as d3s [29]. In our synthesized titanium-based catalysts as well as TiCl4, the electron arrangement of the central atom is d0. As shown in Fig. 1, due to the d0 spin of Ti atom in the synthesized catalysts, d→d transitions were not observed for the studied complexes. Actually, the peaks observed at 260–280 nm were related to π→π (intra‐ligand) transitions. In addition to this type of transition, the ligand‐to‐metal transition at about 240 nm was also detected. However, due to the increase in the length of the resonance system, the intra‐ligand transition in all catalysts has shifted to higher wavelengths (Figs. S25–S27).The ethylene trimerization reaction using the as-synthesized titanium‐based catalysts activated by MMAO afforded 1‐hexene and some by-products. The key role of co‐catalyst, between other functions such as removal of oxygen and water and eliminating environmental pollution, is to reduce the oxidation state of titanium metal and to facilitate the production of cationic species [56]. In general, co-catalyst facilitates alkyl abstraction from the catalyst pioneer to yield an anionic co‐catalyst species [RX−] and a cationic metal species [LnM+], which together represent the active catalytic system as an ion pair with [LnM+][RX−] formula (Scheme 3 ) [54]. The MMAO‐activated system was highly active and selective in ethylene trimerization reaction [57]. In fact, an analysis of the liquid fraction by GC disclosed that ethylene trimerization via catalyst/MMAO under different position achieved 1‐hexene with high selectivity.The results of ethylene oligomerization using the C2 catalyst are shown in Table 1 . First, the effect of reaction temperature on catalyst activity was investigated (entries 3, 5-7). To do this, the reaction was performed at four temperatures of 20, 40, 60, and 80 °C with 1.5 μmol of C2 catalyst solution, and 8 bar ethylene pressure. From Fig. 2, it is clear that for reaction temperature there is an optimal value of 60 °C, in which catalyst activity and 1-C6 selectivity have their maximum amount of 2593 kg 1-C6/mol-Ti h and 78.77%, respectively. It was reported that the pendant ring in the complex structure can coordinate easily to the electron deficient cationic Ti species due to the higher catalyst flexibility at elevated temperatures which subsequently causes a substantial decrease in the catalyst activity [33,58,59].Due to the negligible productivity difference in the productivity values at the reaction temperatures of 40 °C and 60 °C (2389 and 2593 kg 1-C6/mol-Ti h, respectively), the temperature of 40 °C was chosen for the next studies. In the next step, effect of ethylene pressure was taken into account. According to Fig. 3, increasing the ethylene pressure from 3 to 8 bar, the activity of the C2 catalyst increased due to the enhanced solubility of ethylene gas at higher pressures. However, after P = 8 bar, the reaction switched to ethylene polymerization, so that at P = 12 bar, the activity decreased to 337, and 465 kg 1-C6/mol-Ti h, at the T = 20 and 40 °C (entries 4 and 10, Table 1), respectively. Therefore, the next experiments were conducted at P = 8 bar as the optimum ethylene pressure value.Worth mentioning, effect of ethylene pressure on C2 catalyst activity and 1-C6 selectivity was considered at T = 20 °C, as well (entries 1–4, Table 1 and Fig. 4). Notably, in this condition the same trend as it at T=40°C was observed. Indeed, the highest catalyst activity and 1-C6 selectivity of 1179 kg 1-C6.mol Ti−1.h−1 and 74% were obtained at P = 8 bar (entry 3) after which the process was switched to the polymerization reaction with catalyst activity of 1294 kg PE/mol-Ti h.In the following, effect of Al/Ti molar ratio on the C2 catalyst activity and selectivity was investigated, Fig. 5. Obviously, by increasing Al/Ti molar ratio up to 2000, catalyst activity increases due to the formation of more catalytic active sites. As the Al/Ti ratio raised above 2000, the activity decreased due to the poisoning of catalytic sites. Therefore, Al/Ti=2000 molar ratio was selected as the desired value, due to high activity of C2 catalyst toward 1-C6 formation.According to the results, C2, pressure of 8 bar, T = 40 °C and catalyst dosage of 1.5 μmol were selected as the optimum reaction conditions for achieving high catalytic activity and 1‐hexene selectivity. Finally, the effect of bridge type on the catalytic efficiency was elucidate. According to the previous studies conducted in these catalyst systems, there is no coordination between dangling arene (phenyl) and Ti metal at the primary precatalyst. However, after activation with the MAO co‐catalyst, it happens easily which facilitates the formation of 1-C6 product [26,60]. The bridge between the indenyl and phenyl moieties has a significant effect on the direction of phenyl ring relative to the titanium metal center and the size of C(Ind)C(bridge)C(phenyl) angle. When the catalyst is activated by the co‐catalyst, the dangling phenyl will go toward the cationic metal with a low oxidation number. Therefore, the type of bridge and its size have a direct effect on the coordination of arene and Ti. In the C1 with the bridge C(cyclo-C5H10), it has stronger coordination than the same complex with the CMe2 bridge. Catalyst C4 has the most space constraints since the pendant arene moiety has a very strong coordination with Ti. Consequently, this can prevent ethylene from approaching the metal active site. That is, the bridge has a dual effect on catalyst efficiency. While the cycloheptane bridge leads to the largest spatial hindrance and angular pressures, and thus low activity, the catalyst with a cyclohexane bridge with stable chair structure, high activity and selectivity. Table 2 shows the trimerization results for C1–C4 catalysts (see Scheme 1 for the detail of the catalysts).According to Table 2, it is noteworthy that, regardless of the polymer produced in the trimerization, the selectivity for 1-hexene in all catalysts was higher than 90%. However, considering the unwanted polymer by-product, the selectivity of 1-hexene for all four catalysts decreased. 1-hexene selectivity for catalyst C4 was obtained due to the production of 1.7 g of polyethylene in the reaction of only 30% (Fig. 6). In the last part of our study, to compare quantitatively the effect of ligand type on the energy values of ethylene trimerization reaction path, and shed light on the structural parameters, DFT calculations were carried out. In this regard, the main responsible and effective steps in the reaction pathways were considered (Scheme 4 ).The catalyst activation steps by the MAO co-catalyst (Scheme 3) were not considered and M0 is our starting active catalyst (see Scheme 4), in which the titanium has an oxidation state II. Thus, it has a severe electron deficiency and tends to coordinate rapidly with the two ethylene molecules leading to M1 (note that in the energy profile the insertion of only one ethylene molecule (M1’) has also been considered) [61], from which a five-membered metallacycle is formed (M2), switching the corresponding oxidation of the metal center from Ti(II) to Ti(IV). It is possible that as a result of the ring-opening reaction 1-butene will be formed through two different reaction pathways. The first mechanism is the β-hydrogen transfer to Ti, forming a hydride, and the subsequent elimination, the second mechanism is the transfer of intramolecular β-hydrogen and the formation of M3, which means a reduction of the metal center again to Ti(II), and the olefin is bonded to titanium. Since 1-butene is not formed, instead the reaction will lead to the formation of M4 and the third ethylene molecule will be coordinated. After ethylene insertion (M5), the reaction likely continues with a ring-opening (M6) to finally produce 1-C6 (experimentally observed). However, a new ethylene molecule may bond to titanium and M7 may be formed, leading to the release of 1-octene or even higher olefins such as 1-decene. However, the latter species was not observed experimentally. After considering the general path shown in Scheme 4, the structure of the catalysts was optimized. Since 1-hexene was observed experimentally as the main product and 1-octene as the by-product, only the energy of these steps was investigated for the four catalysts included in Fig. 5. The energy diagram for ethylene oligomerization was then obtained for the four catalysts. In summary, the catalysts can evolve via four main stages of release of 1-butene, the formation of a seven-membered ring, the release of 1-hexene, and the formation of a nine-membered ring.The reaction mechanism proposed in Scheme 4 has been studied for the four different catalysts C1–C4. In general, results in Fig. 7 are similar for the different candidates. In all cases, the formation of the seven-membered metallacycle (M5) through an insertion of a third molecule of ethylene (T4) is more kinetically favorable than the release of a 1-butene molecule through the intramolecular β-hydride transfer (T2B). This difference is more significant in systems C2 (9.5 kcal•mol−1) and C4 (8.3 kcal•mol−1) than in systems C1 (2.7 kcal•mol−1) and C3 (3.2 kcal•mol−1). These observations match with the experimental evidence reported above, since it has not been observed any release of 1-butene molecules. After the formation of the seven-membered ring (M5) the mechanism proposes both the insertion of a fourth ethylene molecule subsequently forming a nine-membered metallacycle (M8) or, the release of a 1-hexane molecule through, again, an intramolecular β-hydride elimination, leading to M6, being the latter, the winner not thermodynamically, but kinetically speaking. The transition state involving the intramolecular migration of the β-hydrogen (T5) (see Fig. 8a) is thus lower in energy than the transition state involving the formation of the nine-membered metallacycle (T7) (see Fig. 8b). Matter of fact, in all cases the expected product is the 1-hexene while 1-octane becomes only a by-product. The catalysts showing better selectivity towards 1-hexene release is the system C1 as the difference in energy is around 2.9 kcal•mol−1, being C4 the one showing less selectivity as this difference is only 0.8 kcal•mol−1, with C2 and C3 in between, with values of 1.6 and 2.2 kcal•mol−1, respectively. Again, these observations match with the experimental data reported before.According to the summarized data included in Table 3 (all the optimized geometries and more relevant distances are collected in Table S1 in the supporting information), the energy barrier of ΔETS for the formation of the seven-membered ring is lower than for the release of 1-butene. On the other hand, the energy barrier of ΔETS to release 1-hexene is lower than for the formation of a nine-membered ring. Because the amount of ΔE2 is less than that of ΔE1, the reaction tends to form 1-hexene. Calculations for other catalysts are shown in Table 3, indicating that the most feasible path is the formation of 1-hexene. The C4 catalyst has a lower selectivity than other catalysts. Anyway, still the next kinetic barrier leading to M9 is higher, and thus the 1-octene formation is even more disfavored than the 1-hexene one, for the four catalysts C1-C4.To evaluate the sterics among the series C1–C4 the %VBur of the rate determining intermediate M5 was evaluated [62,63]. The values are 26.9, 25.4, 27.0 and 26.3%, respectively. Even though the cyclohexane based systems C1 and C3 have a more hindered metal center [64], the difference is scarce and not enough to describe any trend (see Tables S28–S31 for further details) [65–67]. Nor can emphasis be placed on weak interactions such as non-covalent ones [68], since for the four catalysts the H-bonds are almost identical.High selectivity and productivity of 1‐hexene production was obtained using the four titanium‐based catalysts C1–C4, in good comparison with recent work [69]. Ethylene trimerization was performed using these synthetic catalysts with changes in MAO concentration, reaction temperature, and ethylene pressure. Using the C2 catalyst, the concentration of 1.5 μm for the C2 catalyst, the selectivity, and activity of the three catalysts C1, C3, and C4 were also examined in this optimal concentration. The activity of all four catalysts increased with increasing temperature to 60 °C, indicating the thermal resistance of synthetic catalysts at high temperatures. Increasing ethylene pressure always increases the solubility of ethylene and increases the activity, which was also true for the synthesized titanium-based catalysts but, the C2 catalyst moves from ethylene trimerization to ethylene polymerization at an ethylene pressure of 12 bar at 20 and 40 °C.The authors state that “There are no conflicts to declare”.This work was supported by the Iran National Science Foundation (INSF) through the grant number of 98020308. A.P. is a Serra Húnter Fellow and ICREA Academia Prize 2019, and thanks the Spanish MINECO for project PGC2018-097722-B-I00 and CTQ2017-85341-P. G.P. gratefully acknowledges the support of Institut de Química Computacional i Catàlisi (IQCC) and the computer resources and technical support provided by the Barcelona Supercomputing Center (BSC).Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mcat.2021.111636. Image, application 1

Different types of [Ind-C(R)-Phenyl]TiCl3 catalysts based on pendant arene containing indenyl (Ind) ligand bearing various types of bridges (R=cyclo‐C5H10 (C1), (CH3)2 (C2), 4-tBu-cyclo‐C5H9 (C3), and cyclo‐C6H12 (C4)) have been synthesized, and used in the ethylene trimerization to 1-hexene in the presence of methyl aluminoxane (MAO) as co-catalyst. The reaction conditions were first optimized in C2 catalyst case, where the highest 1-hexene product was achieved at the catalyst concentration, temperature and ethylene pressure of 1.5 × 10−3 M, 40 °C, and 8 bar, respectively. During this optimization and under specific reaction conditions, a switching behavior from ethylene trimerization to polymerization was also detected, as an undesired reaction. At the optimized conditions, synthesized catalysts showed the following trend toward both 1-hexene yield and selectivity: C1>C2>C3>C4. Then, to shed light on the possible reaction mechanisms and to confirm the activity trend obtained in experimental section, density functional theory (DFT) calculations were employed. In this line, obtained results for activity trend in the simulation studies fit well with the experiments. According to both experimental and DFT results, the highest catalytic activity was observed in the presence of the catalyst with a cyclohexane middle bridge (C1).

Ceria-based compounds are very attractive materials for electrochemical applications, such as solid oxide fuel cells (either electrolyte or anode [1,2]), oxygen storage materials [3], oxygen sensors or catalysts of partial oxidation of hydrocarbons [4,5]. They present high mobility of oxygen ions [3], high oxygen storage capacity (OSC) [3], attractive redox catalytic properties [1–5], chemical compatibility with water and carbon dioxide at high temperatures [4] and sufficient resistance to reduction under relatively low oxygen partial pressures [4]. The high oxygen mobility in ceria promotes the mechanism of carbon removal, which in turn, should contribute to the stability of the catalysts on hydrocarbon conversion reactions [3].Undoped ceria (CeO2) has a fluorite-type structure with 8-fold coordination of cations. In reductive atmospheres the oxygen vacancies are compensated by the reduction of Ce4+ to Ce3+. As a result, n-type electronic conduction through small polaron thermally activated hopping occurs [4,6]. To introduce the oxygen vacancies and to increase the ionic conductivity of these compounds, cerium atoms in the structure can be substituted with some aliovalent cations. Among them one can find e.g., Gd3+ or Sm3+ [7,8]. Another interesting type of dopant is praseodymium [9,10], which shows a significant redox activity under reductive conditions. At reduced pO 2 Pr acts in a similar manner to Gd or Sm. It is a fixed valent dopant non prone to changes in the oxygen vacancy concentration with temperature [9,10]. At high pO 2 the concentration of oxygen vacancies becomes strongly dependent on temperature and oxygen partial pressure [9,10].In order to obtain a ceria-based compound with a considerable concentration of oxygen vacancies, the strategy of aliovalent co-doping of ceria with Gd and Pr has been suggested in the literature [9]. In this case oxygen nonstoichiometry is formed as a result of charge compensation. Another pair of dopants, Nd and Sm, led to improved electrical conductivity thanks to the lowering of an association enthalpy of the oxygen vacancy and the dopant ions [11–13]. Other factors, which may play a role in co-doping effect are: raise of a configurational entropy, modification of an elastic strain in the crystal lattice and changes in the grain boundary composition [11–13].In this paper, aliovalent co-doping procedure by Sm and Pr has been suggested. Then oxygen vacancies should be formed to compensate effectively negatively charged Sm3+, Pr3+ and Ce3+ that partially substitute Ce4+. Assuming ideal behaviour for the reduction of Pr and Ce in the lattice (invariant values of reaction entropies and enthalpies) the expected defect reactions leading to a formation of oxygen vacancies can be described by the following reactions [9]: (1) S m 2 O 3 → C e O 2 2 S m C e ′ + V O ⋅ ⋅ + 3 O O x (2) 2 P r C e x + O O x ↔ 2 P r C e ′ + V O ⋅ ⋅ + 1 2 O 2 ( g ) (3) 2 C e C e x + O O x → 2 C e C e ′ + V O ⋅ ⋅ + 1 2 O 2 ( g ) where O O x is the oxide ion in its lattice site, V O ⋅ ⋅ is an oxygen vacancy, S m C e ′ is Sm3+ ion in Ce4+ ion site, P r C e x and P r C e ' are Pr4+ and Pr3+ ions in Ce4+ site, respectively.It is clear from the above equations that both dopants lead to a formation of oxygen vacancies, which may have a positive effect on the partial oxidation of hydrocarbons catalysed with the use of these compounds. Therefore, in this work various nanocrystalline compounds of Pr and Sm co-doped ceria (with up to 20 mol.% of dopants) were fabricated by the reverse microemulsion synthesis method. Next, they were deposited in the form of layers on the surface of SOFC anode in aim to act as electrochemically active materials for the biogas reforming process. The aim of such a SOFC anode modification was to investigate the influence of these functional layers on a lifetime and efficiency of the commercially available solid oxide fuel cell operating under biogas without the need of an external reformer.The following compositions: CeO2-δ, Ce0.9Sm0.1O2-δ, Ce0.9Pr0.1O2-δ, Ce0.8Pr0.15Sm0.05O2-δ, Ce0.8Pr0.1Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ have been fabricated via a reverse microemulsion method. A detailed description of the applied procedure is reported elsewhere [14].The phase composition of the investigated materials was analysed using the X-ray diffraction method (XRD) by an X'Pert Pro MPD Philips diffractometer with Cu Kα (1.542 Å) radiation at room temperature. The size of crystallites was estimated based on the Scherrer formula: C = kλ/[(Be−Bt)cosθ], where C is an average diameter of the crystalline grain, k is a constant (assumed to be 0.9), λ is the X-ray wavelength, θ is the diffraction angle, Be is the measured width of a peak profile and Bt is the instrumental width of a peak. The XRD patterns were also analysed by the Rietveld refinement method using a HighScore Plus software with the pseudo-Voigt profile function applied. As a starting point of the analysis, crystal structure parameters of CeO2 (Fm-3m space group) were used [15]. The morphology of fabricated materials was examined using the FEI Quanta FEG 250 Scanning Electron Microscope (SEM). The thermal expansion coefficient of doped-ceria pellets (previously sintered at 1000 °C for 2 h) was determined using the Netzsch DIL 402 PC dilatometer operating in 100–1000 °C temperature range under nitrogen atmosphere with 3 °C/min heating/cooling rate.To form pastes, the obtained powders were ground in a mortar for about 1 h with ESL 403 organic binder (ElectroScience Laboratory, USA). The prepared pastes were deposited on the anode surface of a traditional 1-inch Solid Oxide Fuel Cell (Ni-YSZ anode, YSZ electrolyte and LSM-YSZ cathode). The catalytic layer was a circle of 16 mm diameter and 30 μm thickness. Finally, the modified fuel cells with deposited catalytic layers were fired at 1000 °C for 2 h.Such prepared fuel cells were mounted in a measurement rig [16]. They were heated up to 800 °C with argon delivered to the anode side and then, to reduce nickel oxide, humidified hydrogen (3% H2O) was supplied at 800 °C for 30 min and further at 750 °C for 20 h. After this time the hydrogen was replaced by wet synthetic biogas (3% H2O) consisting of methane and carbon dioxide mixed at a volume ratio of 60:40. The total flow rate of the inlet gas mixture was 21 cm3 min−1, ensuring constant gas supply to the FTIR system at the outlet of the fuel cell. This study was focused more on the comparison of the activity of additional catalytic layers and its influence on degradation rate, disregarding the optimal fuel utilisation factor discussed in other papers [17,18]. Two types of electrical measurements were collected during fuel cell operation in biogas: a current density versus voltage and a current density versus time at 0.65 V for the 90 h at 750 °C. A scheme representing a general procedure of the experiment applied in this work is shown in Fig. 1 .Simultaneously with electrical tests, an analysis of the composition of the outlet gases from SOFC was performed using a Fourier Transformed Infrared Spectroscopy (PerkinElmer Spectrum 100 with ZnSe optical windows). FTIR spectra were collected every 10 min within the wavenumber range of 4000–500 cm−1 with a resolution of 4 cm−1. Concentrations of methane, carbon dioxide and carbon monoxide were then calculated. Although H2 gas is not visible in FTIR spectra, after a calibration process, we were able to determine its concentration as a difference from 100% of summed CH4, CO2 and CO concentrations. Such an approach is correct and in line with expectations. Moreover, the conversion rates of CH4 and CO2 as well as the yields of H2 and CO were calculated. A detailed description of the measuring system and analysis methods was reported elsewhere [16].The XRD patterns of Ce(Pr,Sm)O2-δ powders fabricated by a reverse microemulsion method are presented in Fig. 2 . All diffraction peaks can be attributed to CeO2-δ, which indicates that all materials are single-phase. Therefore it can be stated that the amount of dopants applied in this work (20 mol%) is below the solubility limit of Pr and Sm in ceria, what is in agreement with literature reports [19]. The obtained XRD data also allowed to estimate the size of the crystallites in the fabricated compounds. The results of the analysis are shown in Table 1 . It can be found that all of the materials are nanocrystalline and that addition of dopant reduces the average size of crystallites (from ca. 10 to 7 nm). The unit cell parameters of analysed compounds both with the goodness of fit (GOF) values for Rietveld refinement are also presented in Table 1. The lowest lattice parameter was found for pure CeO2-δ and Ce0.9Pr0.1O2-δ, because Ce4+ and Pr4+ ions have the smallest and comparable ionic radii, as shown in Table 1. However, when both Pr and Sm dopants were introduced into ceria, then there is no linear dependence between lattice parameter and the amount of dopant, what suggests that praseodymium exists in mixed-valence state (3+/4+) in these compounds [4,9].The morphology of fabricated powders examined using Scanning Electron Microscopy is shown in Fig. 3 . All doped-ceria compounds have similar, uniform microstructure with round-shape grains of an average size of 20 nm. The only exception is pure CeO2-δ in which two types of grain shape are visible: round and flakes-like. The former has an average size of 20 nm and the latter up to 100 nm.In the next step the chemical compatibility of the fabricated compounds with the NiO-YSZ anode material was investigated. For this purpose the powders of the doped ceria materials were mixed with NiO-YSZ powder at 50 vol% ratio, ground in an agate mortar, uniaxially pressed into pellets and subsequently sintered at 1000, 1100 and 1200 °C for 2 h. Then the XRD analysis was performed. In Fig. 4 , a part of XRD pattern of Ce0.8Pr0.1Sm0.1O2-δ –NiO/YSZ composite after sintering at different temperatures for 2 h is shown. It is representative also of other compositions. It clearly indicates that at temperature above 1000 °C a chemical reaction between ceria and YSZ phase takes place. As a result, the secondary phase of Ce0.33Zr0.67O2-δ is formed. This observation allows us to conclude that catalytic layers of ceria deposited on the surface of NiO-YSZ anode should be sintered at temperature not higher than 1000 °C to prevent the zirconium diffusion. Nevertheless, a potentially negative effect of the existence of this additional Ce–Zr–O phase depends on the part of the fuel cell in which it is formed [21–27]. For example, it was reported by Patel et al. [24] that when this phase occurs at the Ni/YSZ/CeO2 anode operating with direct hydrocarbon feeds, it plays a key role in suppressing carbon formation and associated cell cracking. When Ciementi et al. [25] used Zr0.35Ce0.65O2-δ to decorate the Ni-YSZ anode operating in methanol fuel, they also found that the addition of this compound not only enhanced the coking resistance due to its oxygen storage capability, but also increased the activity of the anode for fuel electro-oxidation due to the increased conductivity, as well as it affected the type of carbon that was formed [26]. On the other hand, if this phase forms also at the anode-electrolyte interface, then it will deteriorate the ionic conductivity of the thin YSZ electrolyte, lowering the performance of a whole fuel cell [27].In the next stage of the experiment the doped-ceria powders were mixed with an organic binder to the form of a paste and deposited on the surface of NiO-YSZ fresh anode. Then they were sintered at 1000 °C for 2 h in the air. The cross sections of these samples were analysed by SEM to examine the quality of the interface between the ceria catalytic layer and the anodic support. Exemplary SEM images of Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ interfaces are shown in Fig. 5 . They are representative of two groups of compounds: without and with praseodymium dopant. Catalytic layers without Pr have quite uniform microstructure, both in the case of grain size and porosity. In the interface with the anode support there are no visible cracks and no traces of delamination. When Pr is used as a dopant then the quality of the layer is quite poor. Large agglomerates, non-uniform porosity as well as a poor adhesion to the support are visible.To better understand the source of this phenomenon the dilatometry studies of fabricated doped-ceria materials were performed. For reference also the NiO-YSZ anode was examined. The results are presented in Fig. 6 . The calculated values of the expansion coefficient in defined temperature ranges are shown in Table 2 . The thermal expansion coefficient of NiO-YSZ is equal to 12.4 × 10−6 K−1 and is in agreement with literature reports [28]. Among presented compounds only Ce0.9Sm0.1O2-δ fits well to NiO-YSZ anode (TEC = 12.6 × 10−6 K−1). This observation explains a very good quality of the interface between the Ce0.9Sm0.1O2-δ catalytic layer and the anode support. It can also be noticed that all Pr-doped ceria samples show linear expansion below 400 °C. The expansion coefficients measured in this temperature range are close to that of NiO-YSZ. However, the dilatometry curves display a non-linear behaviour with an inflection point at ~550 °C. This deviation from linearity increases with increasing Pr content. Such behaviour has been previously reported in the literature for Pr-doped ceria compounds [9,10,29,30]. It is explained by a chemical strain originating from the combination of slight contraction of the unit cell upon formation of oxygen vacancies and expansion of the unit cell upon partial reduction of praseodymium from Pr4+ to Pr3+ (according to Eqs. (1) and (2)) [9,10]. Therefore, the quality of the interface between the NiO-YSZ anode and the Pr-containing ceria catalytic layer may be poor. In consequence, it may be less adherent to the substrate and thus be more prone to mechanical damage. However, this weaker interface should not affect the catalytic properties of the layer itself. As long as effective gas diffusion can occur in the layer, biogas reforming chemical reactions should be performed, regardless of contact with the support. In turn, electrocatalytic reactions will require good electrical contact between the NiO-YSZ substrate and the layer (to allow transport of O2− ions), but even point contact enabling the formation of a percolation path should be sufficient and the layer will fulfil its role.After the analysis of structural properties of the fabricated compounds and their compatibility with the NiO-YSZ anode support they were investigated as an additional anode catalytic layers in SOFCs fueled by biogas. The results of electrical measurements are presented as current density plots versus time in Fig. 7 . The data were normalised to the standard value of 100% in the moment of fuel switching from hydrogen to biogas. Such a presentation allows us to exclude the effect of different gas diffusion through a layer due to a different microstructure, which may influence the absolute value of a power density. To make it possible to know the actual current density values at which the fuel cells operated, an I–V plot for the reference SOFC without a layer was added to Supplementary Materials (Fig. S4). In Fig. 7 one can see that after fuel switching from hydrogen to biogas a rapid drop of current density takes place. This is due to fuel dilution with CH4 and CO2 and low fuel utilisation factor in this experiment. Among all investigated compounds the Ce0.8Pr0.05Sm0.15O2-δ layer seems to be the most resistant to fuel change whereas the Ce0.8Pr0.15Sm0.05O2-δ provides the biggest drop of current density after fuel switching. However, after the initial deterioration Ce0.8Pr0.05Sm0.15O2-δ ensures the best long-term stability, which was not observed for other investigated catalytic layers. For the rest of the presented layers further biogas feeding causes progressive degradation of the cell, which is responsible for a constant decrease in the current density (even up to 10% within 90 h of biogas feeding). Therefore one may state that none of these layers is a perfect catalyst for a direct internal reforming of biogas. However, comparing these results with the performance of a reference fuel cell (without a catalytic layer) in a similar experiment, then a beneficial effect of using the catalytic layer is clearly visible. All of the investigated layers give lower current density drop after hydrogen/biogas switching which may be explained by the ability of more effective direct internal reforming of biogas.To better understand the direct internal reforming of biogas and its influence on SOFC performance it is necessary to analyze also the composition of the outlet gases from the fuel cell. All possible reactions occurring at the anode side are presented in Table 3 . Among them, one can find three undesired processes: CH4 pyrolysis (5), Boudouard reaction (6) and CO reduction (7) leading to a formation of a solid carbon, which can block the active area of the catalyst, impede the diffusion of fuel to the triple phase boundary (TPB) as well as it can even destroy the anode structure. The first reaction is an endothermic process, while the others are exothermic.The results of the in situ FTIR analysis for the selected representative compositions: Ce0.9Pr0.1O2-δ and Ce0.9Sm0.1O2-δ (monodoped) and Ce0.8Pr0.1Sm0.1O2-δ (co-doped) are presented in Fig. 8 as time dependencies of concentration of the particular outlet gases and corresponding catalytic parameters: the conversion rates of CH4 and CO2, the CO and H2 selectivities and the yields of CO and H2. The results obtained for all investigated compounds are given in the Supplementary Materials (Figs. S1 and S2). A detailed description of how these parameters were calculated can be found in our previous paper [16].The most stable in time composition of outlet gases were noticed for SOFC with the Ce0.9Sm0.1O2-δ layer. The initial increase of CH4 and CO2 concentration is related to a dilution of initial synthetic biogas mixture (60% of CH4 and 40% of CO2) by the hydrogen remaining in the measuring rig as well as by the initial very intensive internal reforming performed in the whole volume of the catalyst. After a few hours of operation under biogas a kind of equilibrium state is reached. This stabilisation is visible also in corresponding catalytic parameters (Fig. 8 right): conversion rates of CH4 and CO2, the CO and H2 selectivities and the yields of CO and H2. Among two other presented catalysts the worst is Ce0.9Pr0.1O2-δ, regarding the lowest stability in time as well as the least effective internal reforming (the highest amount of unreacted fuel with simultaneous the lowest amount of products). The co-doped Ce0.8Pr0.1Sm0.1O2-δ composition is clearly between the other two compounds. It should be noted that additional beneficial effect of co-doping is visible in more detailed non-equilibrium chemical analysis, which was performed based on FTIR measurements. It allowed to determine a contribution and direction of the particular chemical reactions to the direct internal reforming process and to recognise which of them are mostly responsible for carbon deposition. A detailed description of the procedure was shown in our previous paper [16]. The results for Ce0.8Pr0.1Sm0.1O2-δ are shown in Fig. 9 . Both time dependencies of reaction quotients (Qr) for reactions (1)–(4) from Table 3 (Fig. 9 left) and carbon activity coefficients (αC,r) for reactions (5)–(7) from Table 3 (Fig. 9 right) are presented.It is generally agreed that Qr~10−3 stands for the situation when mostly reactants are present in reaction area, while Qr~103 when nearly all substances are products and finally if 10−3<Qr < 103 then significant amounts of both reactants and products are visible [16,34]. Therefore, related to our results, steam reforming of methane (1) is a dominant reaction, most probably due to the existence of an observable amount of water in reaction atmosphere. Both dry reforming (2) and RWGS (3) oscillate around an equilibrium, what is in agreement with other literature reports [35]. In turn, methanation reaction (4) is fully shifted towards reactants and takes part in decomposition of CH4 rather than its formation.The next parameter, carbon activity coefficient (αC,r) should be equal to 1 at the equilibrium point. When αC,r > 1, then a solid carbon formation is promoted [36,37]. Based on our results it can be concluded that both Boudouard reaction (6) and CO reduction (7) are rather shifted towards reactants (are lower than unity) and do not lead to carbon deposition on the anode side. This is probably due to an addition of water into fuel stream and a formation of additional water molecules by electrochemical hydrogen oxidation (8), which promote rather carbon gasification than deposition. Only αC,r for CH4 pyrolysis is significantly higher than 1 over whole measurement time, which clearly indicates that this reaction is responsible for carbon accumulation in investigated SOFC.This dominance of a steam reforming of methane (1) in a complex internal biogas reforming process as well as CH4 pyrolysis (5) as a main reaction responsible for carbon deposition was also observed for all other investigated SOFCs with catalytic layers. The only difference between them is the dynamics of time changes of the Qr and αC,r parameters. Therefore time dependencies of Qr for steam reforming and αC,r for CH4 pyrolysis were set in one graph for all samples and shown in Fig. 10 . To better understand these graphs it should be explained that the rate of time changes of quotients' values can illustrate how far from equilibrium point is each of reactions at a given time. Indirectly, a course of plotted function can deliver an overall view on how efficiently different reactions are trying to reach their equilibrium points. The higher is the change of calculated Qr in time, the bigger is the difference between the actual concentration of products and equilibrium composition [38]. On the other hand, the time changes of αC,r parameter, that is a reciprocal of Qr, should be understood in an opposite way. The more rapidly particular reactions responsible for coking move away from the equilibrium point, the lower might be the rates of these reactions, leading to slower carbon accumulation [38]. To prove it, additional plots of carbon balance in time can be drawn. This parameter is calculated as a difference between the numbers of moles of carbon in the inlet and the outlet stream of gases and gives us the information about an average rate of carbon deposition during dwell time. A comparison of time changes of carbon balance for all investigated SOFCs can be found in Fig. 11 . The lower is the decrease in carbon balance over time, the more stable is the operation of an analysed fuel cell. A rapid drop of this parameter is an undesired phenomenon, as it may indicate that the reforming process has slowed down (most probably due to the carbon accumulation and/or a change in the microstructure of the catalyst).Therefore, based on the data collected in Figs. 10 and 11 and explanation given upwards, it may be concluded that Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ are the most attractive compounds towards steam reforming of methane, whereas Ce0.8Pr0.15Sm0.05O2-δ is the most stable in time for this reaction, but not so effective. The activity of Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ leads to more efficient conversion of methane and production of CO/H2, what is in agreement with a composition of outlet gases and catalytic parameters shown in Fig. 8. Moreover, in general, it also corresponds well with higher values of current density (Fig. 7) obtained for these fuel cells. However, current density depicts us an efficiency of fuel utilisation and the ability to perform electrochemical reactions in particular compounds, what may be (but it doesn't have to be) in agreement with concentration of outlet gases.Regarding the carbon accumulation in particular catalytic layers, the Ce0.9Pr0.1O2-δ is the least stable (the biggest decrease of carbon balance over time, see Fig. 11) while the carbon deposition due to CH4 pyrolysis is the slowest (the biggest increase of αC,r over time, see Fig. 10 right). However, the latter one goes also in pair with the slowest carbon removal indicated by a reverse direction of Boudouard reaction (6) and CO reduction (7). Finally, the SOFC with Ce0.9Pr0.1O2-δ is very unstable in time, what corresponds well with the results of electrical measurements (Fig. 7). The deterioration of its catalytic parameters is comparable with that observed for a reference fuel cell without any catalytic layer [16].All these observations also allow to conclude that high praseodymium content in ceria-based compounds is not desirable. It does result in a slight improvement in catalytic properties, but undoubtedly it leads to the problems with TEC mismatch between the catalytic layer and NiO-YSZ support. In consequence, many cracks in the catalytic layer/anode interface occur, limiting the electrical contact between these constituents. Further, it deteriorates electrical properties of the operating SOFC, as well as limits electrochemical reactions leading to the oxidation of deposited solid carbon (reaction 11).Finally, it is also worth to notice that in time dependence plots for all analysed parameters of Ce0.9Pr0.1O2-δ compound there is a characteristic extremum point appearing at ~65 h of biogas feeding. At this point there is a local maximum of current density (Fig. 7), corresponding with a maximum of carbon balance (Fig. 11) and a minimum of carbon activity coefficient for CH4 pyrolysis (Fig. 10 right). It correlates with a significant change in a fuel composition (Fig. 8), where unreacted methane starts to be a dominant component in an outlet stream of gases. Then, for a short time, a more intensive carbon accumulation takes place, giving a better electrical contact on the anode side and leading to higher current density. After 3–5 h a new equilibrium state is achieved and the internal biogas reforming further proceeds. The post-mortem SEM image of Ce0.9Pr0.1O2-δ layer is shown in Fig. 12 . It is uniform, with slight cracks randomly distributed. No visible traces of deposited carbon can be found, nor in the surface as well as in the cross section. This result is consistent and representative with other layers. It confirms, that although the investigated Pr and/or Sm doped ceria does not improve the DIR-SOFC electrical parameters, it protects from a significant deposition of a solid carbon, allowing SOFC to operate with biogas for a much longer time than without the additional layer. Even if a small amount of carbon is deposited at the end of the experiment, it seems to be very hard to determine its amount. Although the carbon balance (Fig. 11) could be integrated over time, it should be remembered that this is a dynamic measurement and depositing carbon is oxidised and removed from the cell on an ongoing basis. Only a structural form of possible carbon deposits could be determined using Raman spectroscopy [39,40] or Temperature Programmed Oxidation (TPO) [40–43]. However, for our layers, these methods did not give any reasonable results, as the amount of deposited carbon was too low.The CeO2-δ, Ce0.9Sm0.1O2-δ, Ce0.9Pr0.1O2-δ, Ce0.8Pr0.15Sm0.05O2-δ, Ce0.8Pr0.1Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ oxides, deposited in the form of layers on the surface of SOFC anode directly fed by biogas, were studied in aim to determine the influence of Pr and Sm on the fuel cell performance.Regardless of the composition, the applied reverse microemulsion synthesis method allowed to obtain single-phase, nanocrystalline powders. It was found that in order to avoid the reaction between these oxides and the anode material (NiO-YSZ) sintering temperature should not exceed 1000 °C, since above that temperature a formation of the secondary Ce1-xZrxO2-δ phase was noticed.It was also shown that Pr-doped ceria catalytic layers suffered from poor adhesion to the NiO-YSZ anode support due to a large mismatch in total thermal expansion coefficients of these materials. These observations are in agreement with previous literature reports. However, co-doping with Sm decreased the TEC significantly, leading to a better adhesion of the layer to the anode.The studies of the influence of the catalytic layers on the direct internal reforming of biogas showed that Pr dopant deteriorates the properties of pure ceria. Single samarium doping or, if appropriate, an addition of a small amount of praseodymium is preferred. The Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ materials are the most attractive towards steam reforming of methane, which is a dominant reaction among all processes occurring simultaneously in direct internal reforming of biogas. Their high activity led to more efficient conversion of methane and production of CO/H2, what was in agreement with a composition of outlet gases and catalytic parameters. In turn, CH4 pyrolysis was found to be a dominant reaction responsible for carbon accumulation on the anode side, but both carbon activity coefficient and carbon balance parameters confirmed that the Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ show the highest stability over time and thus are the most attractive candidates for catalytic materials enhancing direct internal reforming of biogas in SOFC.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 Science Center under grant No. NCN 2017/26/D/ST8/00822.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.2020.07.146.

The Pr and Sm co-doped ceria (with up to 20 mol.% of dopants) compounds were examined as catalytic layers on the surface of SOFC anode directly fed by biogas to increase a lifetime and the efficiency of commercially available DIR-SOFC without the usage of an external reformer. The XRD, SEM and EDX methods were used to investigate the structural properties and the composition of fabricated materials. Furthermore, the electrical properties of SOFCs with catalytic layers deposited on the Ni-YSZ anode were examined by a current density-time and current density-voltage dependence measurements in hydrogen (24 h) and biogas (90 h). Composition of the outlet gasses was in situ analysed by the FTIR-based unit. It has been found out that Ce0.9Sm0.1O2-δ and Ce0.8Pr0.05Sm0.15O2-δ catalytic layers show the highest stability over time and thus are the most attractive candidates as catalytic materials, in comparison with other investigated lanthanide-doped ceria, enhancing direct internal reforming of biogas in SOFCs.

Methane (CH4) is a commercially important fuel and alternative chemical resource replacing petroleum oil [1,2], and it is abundant in shale gas, natural gas, and organic-waste-digested biogas [3,4]. While clean liquid fuels and platform chemicals can be obtained through the gas-to-liquid process using methane-derived synthesis gas [5–8], oxidative or non-oxidative coupling of methane can also produce olefins and paraffins for further manufacturing of polymers and aromatic compounds.Oxidative coupling of methane (OCM) is a promising approach for the production of higher hydrocarbons and olefins compared to conventional synthesis gas-based processes as the exothermic reaction does not include equilibrium limitations and has lower energy requirements. During the OCM process, reactions occur on the catalyst surface and in the gas phase [9–12]. CH4 is adsorbed on the catalyst surface and its CH bond decomposes into methyl radicals and hydroxyl groups on the oxide surface. The methyl radicals diffuse into the gas phase, while the surface hydroxyls are removed to form water and leave oxygen vacancies on the catalyst surface. This surface reaction is called CH4 activation (surface reduction). The oxygen vacancies are filled by oxygen species in the gas phase through oxygen activation (surface oxidation). In the gas phase, coupling of the methyl radicals produces C2H6 [13]. C2H6 and paraffins can undergo oxidative dehydrogenation to produce olefins as the CH activation energy of C2H6 is similar to that of CH4 [14]. Furthermore, the deep oxidation of oxygen-containing intermediates or radicals can form COx (CO2 and CO) and hydrocarbons [10,11,15,16].Among OCM catalysts, Mn-Na2WO4/SiO2 is a highly selective and active catalyst [9,17–21]. Partial replacements of Mn and WO4 2− with other metals (Ti, Co, Fe, Cu, Ni, Zn, Y, Zr, Mo, Pd, La, Ce, Nd, Eu, Tb, and Hf) [22–25] and anions (MoO4 2−, SO4 2−, PO4 3-, CO3 2−, SiO3 2−, and P2O7 4-) [26–28], respectively, have been attempted; however, the tri-component Na-W-Mn catalyst still exhibits the highest OCM activity [23]. CH4 is activated at the W sites, thereby leading to the reduction of W6+ to W5+, whereas the Mn compound supplies oxygen to re-oxidize W5+ to W6+; this was recently confirmed by means of a combined in-situ X-ray diffraction (XRD) and operando analysis [3,17,29,30]. Mn-Na2WO4/SiO2 is activated in the reaction temperature range 700–900 °C [31,32]. However, at such high temperatures, deep oxidation to CO2, and the decomposition of hydrocarbons into coke can occur, which can cause the loss of valuable products and possibly, plugging, because of the coke formation. Perovskite catalysts, which exhibit a high stability and require low activation temperatures, have also been reported for the OCM [11,33–38]. Perovskite, with an ABO3 formula, contains A- and B-site cations forming cuboctahedral and octahedral oxygen anions, respectively. The A- and B-site cations can be doped with lanthanides, alkaline metals, alkaline-earth metals, and transition metals. The doping increases the mobility and activity of oxygen in the perovskite structure and improves the activation of the CH4 and O atoms at low temperatures during the OCM. For example, in a previous study performed in our lab, a SrO-deposited BaTiO3 perovskite catalyst achieved a 17.6 % C2+ (olefins and paraffins) yield at 725 °C; however, the low C2+ selectivity (45–50 %) and a lower degree of oxidative dehydrogenation led to a low olefin selectivity [33].In this study, a hybrid BaTiO3 perovskite and Mn-Na2WO4 catalyst was prepared for use in the OCM, in which high olefin and paraffin selectivities were achieved at a lower temperature (700 °C) than required for the conventional Mn‐Na2WO4/SiO2 catalyst (approximately 800 °C). The effects of the metal (Na, Mn, W)-matrix (BaTiO3) interactions on the catalytic OCM results were also investigated for the prepared catalysts. A detailed analysis of the W and Mn redox reactions that occur on BaTiO3 was performed to elucidate the mechanism of the low-temperature activated OCM catalysts.All chemicals were used as received without further purification. Titanium isopropoxide (Ti(OCH2CH2CH2CH3)4, 97 %) and barium nitrate (Ba(NO3)2, 99 %) were purchased from Sigma-Aldrich (Milwaukee, Wisconsin, USA). Sodium tungstate dihydrate (Na2WO4∙2H2O, 99 %) was purchased from Yakuri Pure Chemicals (Kyoto, Japan). Manganese nitrate hexahydrate (Mn(NO3)2∙6H2O, 98 %) was purchased from Kanto Chemicals (Tokyo, Japan). Silica gel 60 (0.060–0.2 mm, 70–230 mesh) and citric acid (C6H8O7, 99.5 %) were purchased from Alfa Aesar (Ward Hill, Massachusetts, USA). Deionized (DI) water (18.2 MΩ∙cm) was prepared using an aquaMAX-Ultra 370 series water purification system (YL Instruments, Anyang, Korea).Citric acid, barium nitrate, and sodium tungstate dihydrate were added to DI water in an alumina cup at room temperature under continuous stirring (360 rpm) for 10 min. Next, manganese nitrate hexahydrate and titanium isopropoxide were consecutively added dropwise to the mixture. The mixture was heated to 80 °C and aged until a transparent gel (metal-citrate complex) formed. The gel was dried in air at 140 °C for 3 h, leading to the formation of ivory powder. The prepared powder was further crushed and calcined in air at 900 °C for 6 h. The detailed compositions of the catalysts are listed in Table S1. The prepared catalysts were denoted as m-n-BTMW, where m = Mn/W (atom/atom), n = (Mn + W)/(Ba + Ti) (atom/atom), and Ba/Ti = 1.08 atom/atom [33]. The other catalysts prepared in this study consisted of Ba/Ti = 1 atom/atom for BaTiO3 (BTO), Ba/Ti = 1 atom/atom and Mn/(Ba + Ti) = 0.4 atom/atom for the Mn-doped BTO (Mn-BTO), Ba/Ti = 1 atom/atom and W/(Ba + Ti) = 0.4 atom/atom for Na2WO4-doped BTO (W-BTO), Mn/W = 3.5 atom/atom for the complex consisting of manganese oxide and Na2WO4, and Ba/Ti = 1 atom/atom, Mn/W = 3.5 atom/atom, and (Mn + W)/(Ba + Ti) = 0.4 atom/atom for the Mn and Na2WO4-doped BTO (Mn-W-BTO). Mn‐Na2WO4/SiO2, containing 5 wt% Na2WO4 and 2 wt% Mn, was prepared by the slurry method as described in a previous study [39]. Silica gel (3 g) was dispersed in DI water (70 mL), and sodium tungstate dihydrate (0.1811 g) was added to the slurry. This was followed by the addition of manganese nitrate hexahydrate (0.1178 mL) while stirring. The prepared powder was calcined in air at 800 °C for 5 h. The BET surface areas of catalysts were measured by N2 physisorption analysis using a Micromeritics ASAP 2020 apparatus (Norcross, Georgia, USA) (Table S2).O2 and CO2 temperature-programmed desorption (O2-TPD and CO2-TPD, respectively), and H2 temperature-programmed reduction (H2-TPR) were performed using a BELCAT-B (MicrotracBel, Osaka, Japan) with a thermal conductivity detector (TCD). A mass spectrometer (BELMASS, MicrotracBEL, Osaka, Japan) was also connected to the BELCAT-B to identify the products during the TPD and TPR measurements. For O2-TPD, the catalysts were pretreated at 900 °C for 1 h in a 5% (vol/vol) O2/He flow, followed by treatment in a 5% (vol/vol) O2/He flow at 50 °C for 1 h. For the measurement, the temperature was increased to 900 °C at a heating rate of 5 °C/min in a He flow. For the CO2-TPD analysis, the catalysts were pretreated at 550 and 850 °C for 1 h under a He flow, followed by treatment in a 5% (vol/vol) CO2/He flow at 50 °C for 1 h. For the analysis, the temperature was increased to 900 °C at a heating rate of 5 °C/min in a He flow. For H2-TPR, the catalysts were oxidized in a 5% (vol/vol) O2/He flow at 900 °C for 1 h, and then cooled to 50 °C, followed by flushing with Ar at 50 °C for 20 min. The measurement was performed in the temperature range 50–900 °C at a heating rate of 5 °C/min.The X-ray diffraction (XRD) results were obtained at 25, 600, 700, and 800 °C using an X’Pert3 PRO powder diffractometer (Malvern PANanalytical, Malvern, UK) with Cu Kα1 radiation (λ =1.54059 Å), operated at 40 kV and 30 mA. At the given temperature, the catalysts were treated in a CH4 flow to obtain the XRD results of the CH4-treated catalysts. The catalysts were flushed with N2 and oxidized in an air flow to obtain the XRD results of the air-treated catalysts. The fractions of hexagonal BaTiO3 phase in the catalysts were roughly measured using the diffraction peak intensity ratio ((103)hexagonal/[(103)hexagonal + (111)tetragonal]) [40], where (103)hexagonal and (111)tetragonal are the diffraction peak intensities of hexagonal (103) and tetragonal (111), respectively.The X-ray photoelectron spectroscopy (XPS) results were obtained using an angle-resolved X-ray photoelectron spectrometer (Theta Probe AR-XPS system, Thermo Fisher Scientific, Waltham, Massachusetts, USA) with a monochromated Al Kα X-ray source (hν = 1486.6 eV) operated at 15 kV and 100 W at the Korea Basic Science Institute (Busan, Korea). The binding energies of all the XPS data were calibrated using the C 1s peak at 284.59 eV (Fig. S1). The Oads/(Oads+Olatt), Mn2+/(Mn2++Mn3++Mn4+), and Ti3+/(Ti3++Ti4+) ratios were calculated using the corresponding deconvoluted peak areas, where Oads and Olatt are the adsorbed and lattice oxygen atoms on the surface, respectively.The morphologies and elemental compositions of the catalysts were determined using a TEM (Talos F200X, FEI, Hillsboro, Oregon, USA) equipped with high-performance energy dispersive X-ray spectroscopy (EDS, Super-X EDS system, Bruker, Billerica, Massachusetts, USA) at the Korea Institute of Science and Technology Advanced Analysis Center (Seoul, Korea). A STEM (Titan TM 80-300, FEI, Hillsboro, Oregon, USA) connected to a high-angle annular dark-field (HAADF) detector was also used to observe the catalysts.The Raman spectra were obtained using a Raman microscope (In Via Raman Microscope, Renishaw, Wotton-under-Edge, UK) with a 532 nm laser excitation source. The Raman spectra were collected at 25, 600, 700, and 800 °C while the catalysts were heated from room temperature to 800 °C under a N2 flow.The OCM was performed in a fixed-bed reactor system under atmospheric pressure (Fig. S2). A straight quartz tube reactor with a height of 370 mm and an inner diameter of 6 mm was used. The catalyst (0.2 g), located close to the furnace thermocouple position, was placed on supporting quartz wool. The remaining reactor volume was filled with inactive zirconia silica beads. The catalyst was heated from room temperature to 650–800 °C in a mixed O2 (20 mL/min) and N2 flow (30 mL/min) at a heating rate of 10 °C/min. A reactant mixture of CH4, O2, and N2 (19.4, 6.5, and 6.5 mL/min, respectively, achieving CH4/O2/N2 = 3/1/1 mol/mol/mol) was fed into the reactor. Weight hourly space velocity (WHSV) was fixed at 9720 mL h−1 g−1. Water vapor produced during the reaction was removed using a cold trap (−2 °C). The gas products (CO, CO2, CH4, C2, and C3 compounds) were analyzed after 30 min of reaction at each reaction temperature using a gas chromatography system (7890A, Agilent Technologies, Santa Clara, California, USA) equipped with a ShinCarbon ST micropacked column. The products were quantified using a TCD and a methanizer-connected flame-ionization detector (FID). The selectivity of molecule i (Si, %), methane conversion (XCH4, %), O2 conversion (XO2, %), yield of molecule i (Yi, %), and carbon-based olefin-to-paraffin ratio (olefin/paraffin, mol/mol) were calculated using the following equations: (1) S i % = N i ×   F i ∑ N i ×   F i   × 100 (2) X C H 4   ( % ) = F C H 4 ,   f e e d   - F C H 4   F C H 4 ,   f e e d × 100 (3) X O 2   ( % ) = F O 2 ,   f e e d   - F O 2   F O 2 ,   f e e d × 100 (4) Y i ( % ) =   S i   ×   X C H 4 100 (5) S C 2 + % = S C 2   H 2 + S C 2   H 4 + S C 2   H 6 + S C 3   H 6 + S C 3   H 8 (6) S C O x % = S C O + S C O 2 (7) O l e f i n P a r a f f i n ( m o l / m o l ) =   2 ×   F C 2 H 4 + 3 ×   F C 3 H 6 2 ×   F C 2 H 6 + 3 ×   F C 3 H 8 where Ni is the number of carbon atoms in molecule i, Fi is the molar flow rate of molecule i in the product mixture, FCH4, feed is the molar flow rate of methane in the feed, FO2, feed is the molar flow rate of O2 in the feed, C2+ is the mixture of olefins and paraffins, and COx is the mixture of CO and CO2. The carbon balance is defined as the ratio of carbon molar flow rate out of the reactor to the carbon molar flow rate of the feed. The carbon balance is a measure of the carbon loss that is caused by the formation of solid products (coke) or heavy gases (C4+) during the reaction. The oxygen balance roughly indicates the difference between the ideal and real oxygen consumptions, as reported in our previous works (see the supplementary data) [11,41]. (8) C a r b o n   b a l a n c e   ( % ) = 2 ×   F C 2 + 3 ×   F C 3 + F C O 2 +   F C O +   F C H 4 F C H 4 , f e e d × 100 (9) Oxygen balance (%) = (O 2 consumed by the formation of hydrocarbons and CO x ) + F O 2 F O 2 , feed ×100 The prepared mixed oxides, 3.5-0.4-BTMW, containing Ba, Ti, Mn, Na, W, and O contain two main types of well-defined nanostructures (Fig. 1 A): BaTiO3-based hexagonal particles, containing highly dispersed Mn, Na, and W (Fig. 1B), and MnO2-rich rods (Fig. 1C).The hexagonal particles contain highly dispersed Ba, Ti, Mn, Na, and W (Figs. 1B and S3), although the overall structure is hexagonal BaTiO3 (Fig. 2 A and B) [42]. On the surface of the particles, Na2WO4 is present on top of Na-Mn-Ti-O (Figs. S4 and S5A), which is deposited on the BaTiO3 matrix (Figs. S4 and S5B). These observations indicate that the excess Mn present, after incorporating into the Mn-doped BaTiO3 structure, exists outside the BaTiO3 particles. The W atoms, which are larger than the Mn atoms, could not be incorporated into the BaTiO3 matrix and Na-Mn-Ti-O; thus, separate Na-W-O particles were formed (Fig. S4). Notably, the ratio of Na/W = 0.39 atom/atom (determined by EDS analysis, Fig. S5B), which is lower than that of Na2WO4 (Na/W = 2 atom/atom), indicates that the small Na+ cations easily migrate to Na-Mn-Ti-O through the BaTiO3 particles. The presence of Ti in the Na-Mn-Ti-O structure confirms that Ti was substituted into the MnO6 octahedra in Na-Mn-Ti-O, with the Na layers interposed between the MnO6 layers [43]. The observed distance between the layers is 4.9 Å, which is lower than that of Na0.8MnO2 (5.1 Å) reported in the literature [44]. This indicates that the Na content in Na-Mn-Ti-O is lower than in Na0.8MnO2 (i.e., x < 0.8).The nanorods contain 1 × 1 (T1) and 2 × 1 (T2) MnO2 tunnel structures in the bulk of the rods, as depicted in the STEM images (Fig. 2C and D) [45–47]. These can accommodate the Na+ ions of Na2WO4. The EDS results also indicate a high concentration of Ti, not Ba, (Fig. S6A), suggesting that Ti-dispersed MnO2 rods are formed. In addition to the MnO2-based rods, further observations of the Na2WO4 particles on the surfaces of the rods (Fig. S7A) suggest that a Na-Mn-Ti-O layer covered by a Na2WO4 layer formed through a reaction between Na2WO4 (as the Na source) and MnO2 [48]. This will be discussed in Section 3.2.The structures of the BTMW catalysts were elucidated by performing XRD analysis. The prepared BTO, Mn-BTO, and Mn-W-BTO were also observed. The XRD results indicate incorporation of Mn and a less amount of W into BaTiO3 (Fig. 3 ). Tetragonal BaTiO3 perovskite (t-BaTiO3, PDF#81-2201) is present in all the catalysts, indicating that BaTiO3 forms the base structure of all the mixed oxide catalysts. Furthermore, small diffraction peaks corresponding to BaCO3 (PDF#41-0373) are present in the XRD results of BTO and Mn-BTO. For Mn-BTO, peaks representing TiOx (Ti9O17, PDF#85-1061) and MnO2 (PDF#44-0142) are present, indicating the substitution of Ti with Mn and the addition of excess Mn, respectively. For W-BTO, strong sharp diffraction peaks representing Na2WO4 are present and this confirms a weak interaction between Na2WO4 (or W) and BaTiO3. For the Mn-W-BTO and 3.5-0.4-BTMW catalysts, the incorporation of Mn into the TiO6 octahedra distorted the tetragonal BaTiO3 structure as a consequence of the Jahn-Teller effect [49,50]. The high Mn content led to the formation of hexagonal BaTiO3, which coexists with tetragonal BaTiO3. The formation of Na-Mn-Ti-O (NaxMn(Ti)O2, PDF#21-1140 for Na0.44Mn(Ti)O2) [51,52] is attributed to a reaction between sodium manganese oxide and the Ti atoms, which were replaced by Mn from the Mn-doped BaTiO3. These observations correspond to the TEM, STEM, and TEM-EDS results (Figs. 1 and 2). Because 3.5-0.4-BTMW contains a higher Ba content (Ba/Ti = 1.08 atom/atom) than Mn-W-BTO (Ba/Ti = 1.0 atom/atom), Ba2TiO4 (PDF#72-0135) was used as a reference for 3.5-0.4-BTMW.The formation of highly dispersed Na-W-O layers on the BaTiO3-rich particles was further confirmed by CO2-TPD (Fig. 4 ) and TEM/STEM/TEM-EDS (Fig. S8). As basic catalysts are considered good OCM catalysts [12,53], the basicity of the m-n-BTMW catalysts (m = 3.5, n = 0–0.4) was determined by CO2-TPD (Fig. 4A and B). CO2 adsorption and desorption did not occur on the BTMW catalysts annealed at 850 °C under an inert He flow (Fig. 4A). Significant CO2 adsorption and desorption is observed at approximately 700 °C for the catalysts pre-annealed at 550 °C under an inert He flow (Fig. 4B), which is, however, attributed to the thermal decomposition of BaCO3, present on the BTMW catalysts, to BaO and CO2 because the curves of the catalysts pre-annealed under an inert He flow at 850 °C does not exhibit strong peaks at 700 °C. The presence of BaCO3 in the Ba-containing mixed oxides was also confirmed by the XRD results of BTO and Mn-BTO, which exhibit diffraction peaks for BaCO3 (Fig. 3a and b). Interestingly, the peak attributed to the decomposition of BaCO3 to BaO and CO2 at approximately 700 °C decreases with an increase in the (W + Mn) content and is not present in the result of 3.5-0.4-BTMW. This indicates that BaCO3 is not present in the BTMW catalysts that contain W and Mn (Fig. 4B). The strong interaction between Na-W-O and BaTiO3 may suppress the formation of BaCO3 in the structure. In the absence of BaCO3 decomposition peaks, the two broad peaks in the result of 3.5-n-BTMW (without Mn and W, n = 0) at 100–400 °C and 400–800 °C are assigned to weak and strong basic sites, respectively. Compared to 3.5-0-BTMW, the results of Mn-BTO and W-BTO did not exhibit clear CO2 desorption peaks, which can be attributed to the poor basicity of Mn-BTO and formation of the Na-W-O layer which fully occupies the surface of W-BTO. Notably, the TCD peaks observed in the results of 3.5-0.3-BTWM, 3.5-0.4-BTWM, and Mn-BTO at 400–800 °C can be attributed to O2 formation, not to CO2 desorption, during the thermal decomposition of the manganese oxides, as indicated in the O2-TPD results (Fig. S9).The TEM and STEM images of W-BTO confirm the formation of a Na-W-O shell on the BaTiO3 surface (Figs. S8A–C), which is also observed in the SAED image (Fig. S8D). EDS also revealed the formation of a Na-W-O layer on the BaTiO3-rich particle surface (Fig. S8E). Among the 3.5-n-BTWM catalysts containing varying amounts of n, BaTiO3 was fully covered by Na2WO4 in 3.5-0.2-BTWM or in the catalysts with n ≥ 0.2 (Fig. 4B).The formation of nanostructures, as depicted by TEM (Fig. 1), were confirmed by H2-TPR (Figs. 5 and S10). Among the components, the reduction of Mn was focused with the assumption that Mn is an oxygen supplying component. The H2-TPR results of Mn2O3, Na2WO4, BaTiO3, and the other Na-Mn-W-Ba-Ti-O composites were obtained to elucidate the complex structure of the BTMW catalysts (Fig. S10 and Table S3).By adjusting m = Mn/W between W-BTO (m = 0) and Mn-BTO (m = ∞) (Fig. 5A), the peaks at 500–700 °C increases with an increase in the amount of Mn added; thus, these peaks are attributed to the reduction of Mn species, Tα (Mn4+ to Mn3+) and Tβ (Mn3+ to Mn2+). With the addition of Mn, the Tα and Tβ peaks appeared at 544–578 °C and 600–629 °C, respectively. W-BTO, not containing Mn, exhibits a weak reduction at 400–550 °C, which can be attributed to the reduction of BaTiO3 (Fig. S10), and very small reduction peaks for the W species at 600 °C or higher (Figs. 5A and S10). The Tα and Tβ reduction peaks do not significantly change for the compounds containing W (W = 2.3–28.6 %), however, the Tα peak shifts to a lower temperature by approximately 100 °C for Mn-BTO (no W species). This indicates that the W species suppress the Tα reduction at lower temperatures. In addition, the Tβ reduction, or the oxygen supply from Mn3+, should be easier for m-0.3-BTMW with m = 1–3.5 exhibiting the lower reduction temperatures, compared to those with m = 0.11 and 4–9.A change in the reduction of Mn was also observed by adjusting n = (Mn + W)/(Ba + Ti). For Mn-BTO (Fig. 5A) in which n = 0.1 (Fig. 5B), Mn-BTO (containing MnO2 and BaTi(Mn)O3) is reduced at higher temperatures (Tα =483 °C, Tβ =662 °C) compared to Mn-BTO with n = 0.1 (Tα =461 °C, Tβ =609 °C). BaTi(Mn)O3 is the structure in which Ti of BaTiO3 was partially substituted with Mn. This indicates that MnO2 increases the reduction temperature of Mn3+ to Mn2+ (Tβ). As depicted in Fig. 5A, the surface-occupying W species increases the reduction temperatures of Tα and Tβ. For n = 0.1 depicted in Fig. 5B, lower reduction temperatures are observed for Tα and Tβ compared to Mn-BTO with n ≥ 0.2–0.3. These observations can be attributed to the presence of surface W species which were not incorporated into the BTO structure partially occupying the BTO surface. Less BaTi(Mn)O3 and more Mn oxide can form on the BTO surface if n = (Mn + W)/(Ba + Ti) increases; this leads to an increase in the reduction temperatures of Tα and Tβ. Mn-W (n = ∞) exhibited the highest reduction temperature of 748 °C. For n = 0.2–0.7, in addition to the changes in the reduction temperature, the Tβ peak intensity does not significantly change, while the Tα peak intensity increases, confirming the formation of a larger amount of NaxMn(Ti)O2 (PDF#21-1140 in Fig. 3).The structures of the catalysts during the OCM process were determined through in-situ XRD. The analysis was performed through catalytic CH4 activation and subsequent air oxidation at 25, 600, 700, and 800 °C (Fig. 6 and Table 1 ). Na2WO4 forms on the surface of the W-containing catalysts (W-BTO and Mn-W-BTO), which transform from a cubic structure (c-Na2WO4, PDF#74-2369) at room temperature to an orthorhombic structure (o-Na2WO4, PDF#20-1163) at 600 °C [54], and then, melt at 600–700 °C, with the complete disappearance of the Na2WO4 peaks at 700 °C (Fig. 6B). The Na2WO4 melt can highly disperse WO4 2− on the catalyst surface and decrease the reduction temperature of W during the OCM reaction at a temperature of 700 °C or greater, thereby improving the CH4 activation [55].For BTO that does not contain W, Mn, and Na (Fig. 6A), the room temperature tetragonal BaTiO3 (t-BaTiO3, PDF#81-2201) transforms to an elongated cubic structure at 600–800 °C, followed by a small fraction of the cubic domains transforming into hexagonal structure (h-BaTiO3, PDF#82-1175) at 700 °C. The presence of CH4 or air does not significantly affect the phase behavior of BaTiO3. BaTiO3 forms a cubic perovskite phase (c-BaTiO3, PDF#79-2263) with a slightly elongated c-axis (a ≈ c, c/a = 1.002–1.003) at 600–800 °C (Table 1). The elongation along the c-axis of the cubic structure decreases the dipole movement of Ti4+ along this axis in octahedral TiO6, forming a tetragonal-like deformed structure, which is stabilized as tetragonal BaTiO3 (c/a = 1.007) at room temperature [56,57]. The hexagonal BaTiO3 phase (11.1 %) forms at 700 °C. Notably, the BaCO3 diffraction peaks observed at lower temperatures, disappear at 800 °C. The disappearance of the BaCO3 peaks corresponds with the decomposition of BaCO3 at temperatures lower than 850 °C, as indicated in the CO2-TPD results (Fig. 4A–B).With the addition of W to BTO to prepare W-BTO, cubic BaTiO3, without its hexagonal polymorph, is observed at 600–800 °C (Fig. 6B), unlike what is observed for pure BTO, which undergoes a transformation from elongated cubic to hexagonal phase (Fig. 6A). Although only cubic BaTiO3 is observed for W-BTO in a flow of air, the addition of CH4 at 800 °C leads to elongation of the cubic BaTiO3 structure by expanding the c-axis (c/a = 1.002 at 800 °C) and forming more Ti3+ and oxygen vacancies [58].Tetragonal and hexagonal BaTiO3 phases co-exist at room temperature in Mn-BTO. The tetragonal phasetransforms partially to a hexagonal phase when exposed to a flow of CH4, which reversibly transforms back to the tetragonal phase in a flow of air (Fig. 6C and Table 1). Under CH4 flow, the Mn in the bulk BaTi(Mn)O3 is reduced by the activation of CH4 molecules and the creation of oxygen vacancies, which dimerizes TiO6-TiO6 (corner-shared octahedra in the tetragonal phase) into Ti2O9 (face-shared octahedra in the hexagonal phase) (Scheme 1 ) [59]. The peaks corresponding to TiO2-x (PDF#85-1061 for Ti9O17, [60]) disappear under a flow of CH4 in the temperature range 700–800 °C. This disappearance of peaks is not observed at a temperature of 600 °C or less, indicating the activation of CH4. These peaks are observed under a flow of air. However, the Mn species in MnO2 on the surface of BaTi(Mn)O3 are not reduced at 600 °C in a flow of CH4, unlike the reduction observed for bulk BaTi(Mn)O3. The reduction of MnO2 is suppressed by the oxygen atoms that are supplied from bulk BaTi(Mn)O3 to MnO2, as suggested by the H2-TPR results (Fig. 5). At higher temperatures (700–800 °C), MnO2, under a flow of air, partially transforms to Mn3O4 (PDF#18-0803) [61,62], which is further reduced to MnO (PDF#72-1533) with a switch in the gas flow to CH4. This process is reversible, as a switch back to air re-oxidizes MnO to Mn3O4. Under the flow of CH4, Mn3+ (Mn3O4) on the surface is reduced to Mn2+ (MnO).For Mn-W-BTO (Fig. 6D), the BaTiO3 structure transformation depends on the temperature and the CH4/air environment, similar to that of Mn-BTO (Table 2 ). NaxMn(Ti)O2 was analyzed in a temperature range from room temperature to 800 °C in a flow of air. A shift to lower 2θ values is observed with an increase in the temperature (2θ = 15.98, 36.95, and 43.02° at room temperature vs. 2θ = 15.56, 36.74, and 42.61° at 700 °C and 2θ = 15.51, 36.74, and 42.52° at 800 °C) and several new peaks (2θ = 19.09, 19.39, 33.93°) appear at 700 and 800 °C. Thus, NaxMn(Ti)O2 is reconstructed by the thermal conversion of Mn4+ to Mn3+, as observed in the CO2-TPD and O2-TPD results (Figs. 4 and S9) [43]. The mobile Na+ cations migrate into the bulk Mn-Ti-O structure to neutralize the electronically unsaturated state (Mn3+ compared to Mn4+), which forms NaxMn(Ti)O2 (x<1, Na0.7Mn(Ti)O2 in this case) in a flow of air at 700 and 800 °C [51]. Thus, a large fraction of Mn3+ species in NaxMn(Ti)O2, which is composed of MnO6, MnO5 anions, and Na+ cations [52], improves the formation of MnO5 polyhedron anions. When Mn-W-BTO is exposed to a flow of CH4 for 30 min, CH4 decomposes into methyl radicals and surface-adsorbed hydrogen species, and Mn in Na0.7Mn(Ti)O2 is reduced to Mn2+, which deconstructs the mixed oxide into Mn3O4 and MnO. Although MnWO4 is expected to be formed from Mn-Na2WO4/SiO2 through a reaction between the Mn2+ cations and excited WO4 2− anions [17], in this study, MnO is formed instead of MnWO4. This indicates that the WO4 2− anions can easily combine with other cations including Ba2+.The crystal structures of the m-n-BTMW catalysts, determined at temperatures up to 800 °C, are summarized in Scheme 2 . At room temperature, the cubic Na2WO4 crystal occupies the surface of both NaxMn(Ti)O2 and BaTi(Mn)O3, topped by small BaTi(Mn)O3 and NaxMn(Ti)O2 particles, respectively (Figs. S4 and S6B). NaxMn(Ti)O2, which contains a lower quantity of Na, has a structure similar to MnO2, where MnO6 in MnO2 structure is partially substituted with TiO6 building tunnels constructed from TiO6 and MnO6. In BaTi(Mn)O3, the incorporation of Mn leads to the formation of hexagonal and tetragonal BaTiO3 structures. The hexagonal BaTiO3 contains a large fraction of Mn4+, in which MnO6 and TiO6 coexist in a face-shared octahedral arrangement. The tetragonal BaTiO3 contains a large fraction of Mn3+, which forms corner-shared MnO6 and TiO6 octahedra. At 700–800 °C, however, the Na2WO4 crystals melt to a liquid phase, in which the Na+ cations and WO4 2− anions are highly mobile. Mn4+ in NaxMn(Ti)O2 is thermally reduced to Mn3+, and the Na+ cations migrate to NaxMn(Ti)O2 to neutralize the electronically less stable state. The WO4 2− anions more easily combine with Ba+ and Na+ cations in NaxMn(Ti)O2 and BaTi(Mn)O2, which highly disperses W on the catalyst surface.The activity of the catalysts in the OCM was determined at 700–800 °C using a CH4/O2/N2 ratio of 3/1/1 mol/mol/mol (Table 2). The BTO, Mn-BTO, Mn-W-BTO, and 3.5-0.4-BTMW catalysts exhibit high CH4 (20.3–39.1 %) and almost 100 % O2 conversions over the entire temperature range, while the W-BTO and W-Mn catalysts exhibit lower CH4 (≤ 8.6 %) and O2 (≤ 21.9 %) conversions. Another factor that influences the CH4 activation ability is the dehydrogenation activity of the catalysts, indicated by the olefin/paraffin ratios. Mn-W-BTO and 3.5-0.4-BTMW exhibit higher dehydrogenation activities (olefin/paraffin = 1.3–2.2 mol/mol), while Mn-BTO, W-BTO, and W-Mn exhibit lower dehydrogenation activities (olefin/paraffin = 0.0–0.8 mol/mol).For BTO, a moderate C2+ selectivity (48.6 %) is observed with a lower dehydrogenation activity (olefin/paraffin =0.9 mol/mol) at 700 °C (Table 2). CH4 reacts with the lattice oxygen (Olatt) in perovskite BaTiO3 to create oxygen vacancies on the catalyst surface and form Ti3+ in BaTiO3. The larger fraction of Ti3+ present in the spent BTO catalyst indicates that the adsorption of oxygen (Oads) in the oxygen vacancies is slower than the loss of Olatt during the OCM reaction (Table 3 and Fig. S12). The amount of Oads on the spent BTO and Mn-BTO catalysts is higher (Oads/(Oads + Olatt) = 0.14 and 0.16, respectively) than on the other catalysts (Oads/(Oads + Olatt) = 0.08–0.09), which may improve the deep oxidation of the methyl radicals to CO and CO2 (Table 3 and Fig. S13) [11].Mn-BTO and W-BTO fully oxidize the CH4 molecules, and high COx selectivities (90.0 % and 81.1 %, respectively) are achieved at 700 °C. A high oxidation activity and COx selectivity (1.0–3.6 % CO and 80.6–89.0 % CO2) is achieved by Mn-BTO, which consists of MnO2 but not W on the catalyst surface (Fig. 6C). An increase in Ti3+ and Mn2+ in the spent Mn-BTO catalyst is observed in the XPS results of the catalysts (Table 3, Figs. S12, and S14), which indicates that the oxygen atoms of MnO2 and BaTi(Mn)O3 activate the CH4 molecules to produce methyl radicals. For W-BTO, the high W coverage on the BTO surface, because of the incorporation of less W into BaTiO3 compared to Mn into BaTiO3, suppresses the reaction on the BaTiO3 surface, which significantly decreases the O2 conversion from 99.9 % for BTO to 1.3–5% for W-BTO. The higher amount of Ti3+ in the spent W-BTO catalyst indicates that the oxygen vacancies in BaTiO3 is not rapidly oxidized, leading to the formation of more oxidized tungsten species, which leads to further oxidation of the reactants to COx products (selectivity of 13.6–35 % for CO and 33–67.5 % for CO2).A significantly high C2+ selectivity (64.6–66.3 %) and high olefin/paraffin ratio (1.3–1.4 mol/mol) are observed at 700 °C for the Mn-W-BTO and 3.5-0.4-BTMW catalysts, which contain BaTiO3, Mn, and Na2WO4 as active components. The faster oxidation of the oxygen vacancies in BaTi(Mn)O3 and NaxMn(Ti)O2, which exhibits a change of W(+5)—□ (□ as an oxygen vacancy) to W(+6)—O, is confirmed by the decrease in Ti3+ and Mn2+ in the spent catalysts (Table 3, Figs. S12, and S14). The oxygen supply in the system suppresses further reduction of W(+5), which is formed through the in-situ reduction of W(+6), increasing the C+2 selectivity and O2 and CH4 conversions. 3.5-0.4-BTMW exhibits a slightly higher C2+ selectivity, because of the formation of Ba2TiO4 (Fig. 3) which improves its performance in the OCM reaction [33].3.5-0.4-BTMW was determined to be the optimum OCM catalyst based on its OCM activity and optimization of the BTMW catalyst composition performed at 700 °C. Through the optimization of m and n, it was confirmed that the Mn species activate gas-phase O2 to supply O atoms to the W species. First, m = Mn/W in the m-0.3-BTMW catalyst was manipulated (Fig. 7 A and B). In the absence of Mn and at small m = Mn/W values, low O2 and CH4 conversions are observed, which reaches 96.8 % at m ≥ 2. The conversion of CH4, which is activated by the oxygen supply, slightly decreases with an increase in m = Mn/W. This can be attributed to difficulties associated with the supply of oxygen from the Mn3+ species, or the higher temperature required to reduce Mn3+ to Mn2+ at a higher m = Mn/W, as depicted in Fig. 5A.The value of n = (W + Mn)/(Ba + Ti) in the 3.5-n-BTMW catalysts was also adjusted (Fig. 7C and D). At n = 0–0.1, the 3.5-n-BTMW catalysts contain a limited amount of Mn and W species on the surface, and their catalytic activities are similar to that of BTO. High COx selectivities and O2 conversions are obtained on the exposed BaTiO3 surfaces incompletely covered by W. At n = 0.2, although the BTO surface is fully occupied by Na2WO4 (see H2-TPR results, Fig. 5B), a lower fraction of Mn species, including NaxMn(Ti)O2 and BaTi(Mn)O3, forms on the surface, suppressing O2 activation and decreasing the O2 conversion to 74 %. At n = 0.3–0.4, more Mn species form, leading to an O2 conversion of approximately 100 % and a CH4 conversion of 38.5–39.1 %; a C2+ selectivity of 66.3–66.5 % was achieved. At n ≥ 0.5, the catalyst contains less (Ba + Ti) and more (Mn + W). The W species that fully cover the BTO surface can limit the oxygen supply from BaTiO3, leading to a decrease in the O2 and CH4 conversions. The temperature required to reduce Mn3+ to Mn2+ also increases with an increase in n = (Mn + W)/(Ba + Ti), suppressing the oxygen supply and decreasing the O2 and CH4 conversions.To achieve the optimum yield of C2+, m should be between 2 and 4, and n between 0.3 and 0.4 at 700 °C, as indicated in Fig. 7B and D. The catalytic OCM activity was stable for 100 h, as observed on the representative optimal catalyst (3.5-0.3-BTMW) (Fig. S15). With an increase in the reaction temperature, the C2+ yield decreases, which indicates that deep oxidation to COx occurs in expense of the formation of the desired C2+ compounds.The catalytic activity of 3.5-0.4-BTMW in the OCM was further compared with that of the low-temperature active BaTiO3 perovskite and Mn-Na2WO4/SiO2 catalysts (Fig. 8 ). The CH4 and O2 conversions achieved by 3.5-0.4-BTMW at 700 °C are higher than those by Mn-Na2WO4/SiO2 in the temperature range 775–800 °C (Table S4) [39,41,63–65]. BaTiO3 exhibits the highest CH4 and O2 conversions at 675 °C; however, the conversions decreased below those of the other catalysts at higher temperatures. These observations indicate that higher temperatures are required for the activation of CH4 and O2 on the Mn and W species. 3.5-0.4-BTMW achieved the highest C2+ selectivity at 700 °C, but a lower olefin/paraffin ratio compared to Mn-Na2WO4/SiO2 at 775 °C. Because the oxidative dehydrogenation of paraffins to olefins are accompanied by the production of COx, a lower olefin/paraffin ratio corresponds to a higher C2+ selectivity. BaTiO3, as the more active oxidation catalyst, achieves a lower C2+ selectivity and olefin/paraffin ratio, or higher COx selectivity, compared to the other catalysts, although it is activated at the lowest temperature (650 °C).The spent catalysts were characterized prior to proposing a reaction mechanism. The high-temperature XRD and Raman spectroscopy results of the spent 3.5-0.4-BTMW catalyst in a flow of inert N2 are depicted in Fig. 9 . No MnO is observed at 25–800 °C in the XRD result of 3.5-0.4-BTMW (Fig. 9A), while Mn2+ (MnWO4) is detected in the XRD result of Mn-Na2WO4/SiO2 [3,66]. NaxMn(Ti)O2 is thermally decomposed to Mn2O3(PDF#06-0540) and Mn3O4 at 700–800 °C under a flow of N2; however, this is not observed under a flow of air (Fig. 6D). The high-temperature Raman results (Fig. 9B) indicate the presence of Mn2O3 and Mn3O4 at 700–800 °C under a flow of N2, which is confirmed by the bands at 200, 321, and 348 cm−1 corresponding to the out-of-plane bending modes of Mn2O3 and Mn3O4 [67,68]. The Raman peaks at 535, 659, and 750 cm-1 indicate the formation of hexagonal BaTiO3 [69], while the peak at 628 cm-1 corresponds to the stretching mode of O-Mn-O, and the shifted peak at 685 cm-1 is attributed to Mn2O3 which formed through the decomposition of NaxMn(Ti)O2 [70]. The peaks at 827 and 922 cm-1 correspond to the asymmetric and symmetric stretching vibrations of Na2WO4, respectively [54]. Furthermore, the peak at 915 cm-1 corresponds to BaWO4 [71], indicating that the free WO4 2- anions strongly interact with BaTi(Mn)O3 and NaxMn(Ti)O2 to form BaxWO4 and NaxWO4 on the catalyst surface, respectively. Through these interactions, highly dispersed single WO4 2- anions, which are not oligomerized, can form on the surface of the catalyst.Based on the above discussion, a reaction mechanism for the BTMW catalysts is proposed, as indicated in Scheme 3 . First, CH4 interacts with the oxygen atoms in W—O (in NaxWO4 or BaxWO4) to form methyl radicals in the gas phase and oxygen vacancies in W—□ on the catalyst surface. In the gas phase, coupling of the methyl radicals leads to the formation of C2H6. Second, the oxygen vacancies on the catalyst surface are occupied by oxygen present on the surface of NaxMn(Ti)O2, which reduces Mn3+ to a mixture of Mn2+ and Mn3+ (in Mn3O4). Mn2+ and Mn3+ (on the surface) are re-oxidized to Mn3+ by the oxygen supplied by bulk BaTi(Mn)O3. Reduction of Mn3+ to Mn2+ in NaxMn(Ti)O2 may not occur because of the oxygen supplied by BaTi(Mn)O3. The partial reduction of Mn3+ to a mixture of Mn2+ and Mn3+ can destroy the electronic balance in the crystal structure, improving the migration of the free surface Na cations into the bulk structure to form NaxMn(Ti)O2. When re-oxidized, the fraction of Mn3+ in NaxMn(Ti)O2 increases, and the Na cations are removed from the bulk to the surface. In bulk BaTi(Mn)O3, the reduction of Mn produces more oxygen vacancies to dimerize TiO6 into Ti2O9 in perovskite BaTiO3 and transform the tetragonal structure to a hexagonal BaTiO3 structure. In the final step, Mn2+ in BaTi(Mn)O3 is re-oxidized to Mn3+ by O2 gas to transform the hexagonal phase back to tetragonal phase.m-n-BTMW catalysts, hybrid catalysts consisting of perovskite BaTiO3 and Mn-Na2WO4, have been prepared and applied in the OCM. The strong interactions between BaTiO3, Na2WO4, and Mn created active sites composed of NaxMn(Ti)O2, BaTi(Mn)O3, and NaxWO4 under the OCM reaction conditions. The strong interaction between NaxMn(Ti)O2 and BaTi(Mn)O3 promoted the activation of oxygen, supplying oxygen to W on the surface of the catalysts at low temperatures, and increasing the C2+ selectivity through the activation of CH4 on NaxWO4 at temperatures close to the melting temperature of Na2WO4 (approximately 700 °C). In addition, fast recovery of the Mn oxidation state in these Mn compounds via the activation of CH4 increased the CH4 conversion achieved by the m-n-BTMW catalysts. The lower reducibility of Mn in NaxMn(Ti)O2/BaTi(Mn)O3 in the m-n-BTMW catalysts led to desirable activity for the OCM reaction compared to the conventional Mn‐Na2WO4/SiO2 catalyst; an optimal reaction temperature and increased C2+ yield was achieved. The optimized m-n-BTMW catalyst, exhibiting the best OCM activity at 700 °C, contained m and n values of 2–4 and 0.3–0.4, respectively. Lien Thi Do: Conceptualization, Methodology, Investigation, Writing - original draft. Jae-Wook Choi: Validation, Methodology. Dong Jin Suh: Validation. Chun-Jae Yoo: Methodology, Supervision. Hyunjoo Lee: Resources. Jeong-Myeong Ha: Conceptualization, Methodology, Writing - review & editing.The authors report no declarations of interest.This research was supported by the C1 Gas Refinery Program (2015M3D3A1A01064900) and the Technology Development Program to Solve Climate Changes (2020M1A2A2079798) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT.Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.apcatb.2021.120553.The following are Supplementary data to this article:

The oxidative coupling of methane (OCM) using hybrid catalysts containing BaTiO3 perovskite and Mn-Na2WO4 exhibited high activity and high selectivity at low temperature: 66.3 % C2+ (olefins and paraffins) selectivity and 25.9 % C2+ yield at 700 °C which is 100 °C lower than that used for Mn-Na2WO4 catalysts (800 °C). Upon the preparation of complex catalysts, the insertion of Mn into BaTiO3 (BaTi(Mn)O3) and the formation of NaxMn(Ti)O2 improved the oxygen-supplying ability of the catalysts. Additionally, strong interactions between the WO4 2− anions and BaTi(Mn)O3 or NaxMn(Ti)O2 stabilized the WO4 2− anions on the surface and improved the methane activation ability of the catalysts for the favorable production of hydrocarbons. The nanoscopic modification of catalysts was confirmed using H2–temperature-programmed reduction, transmission electron microscopy, X-ray diffraction, X-ray photoelectron spectroscopy, and Raman spectroscopy results.

Transformation of biomass, especially non-edible and waste materials, to high-value chemicals, fuels, and materials is a promising approach for reducing dependence on fossil resources. Recently, interest among researchers in the production of lactic acid has been increasing due to its growing market demand. Lactic acid is a naturally occurring organic acid that is one of the main ingredients in the food, cosmetic, and textile industry. It is currently considered as one of the most potential feedstock monomers [1,2], that can be converted to chemicals such as pyruvic acid, acrylic acid, 1,2-propanediol, 2,3-pentanedione, lactate esters and polylactic acid (PLA) [1,3–7]. Lactic acid can be produced by chemical synthesis or by the fermentation of carbohydrates that are present in the biomass. Today, lactic acid is commercially produced mainly through the fermentation of sugars, including glucose and sucrose derived from starchy feedstocks [1,2,8–11]. However, the current fermentative production of lactic acid has several drawbacks. These include environmental and scaling-up issues arising from the raw material choices, long reaction times, waste generation as well as from the separation problems in recovering the pure lactic acid [2,9,11]. Therefore research groups have focused on novel chemocatalytic methods, which are more desirable and cost-effective, for directly converting cellulosic biomass or sugars into lactic acid or its alkyl esters [1,5,12,13].The proposed reaction route (Scheme 1 ) in the conversion of cellulosic biomass-based hexoses such as glucose includes three key steps: (1) the isomerization of C6 monomers (glucose) into fructose, (2) the retro-aldol reaction of fructose to C3 triose intermediates, such as dihydroxyacetone, pyruvaldehyde, and glyceraldehyde, followed by (3) the conversion of trioses to lactic acid via several tandem reactions. In the reaction, two molecules of lactic acid are formed from one molecule of hexose. The main byproducts, 5-(hydroxymethyl)furfural (HMF), levulinic acid, and formic acid are also formed [14–16].The traditional chemocatalytic methods for the conversion of biomass to lactic acid or its esters from lignocellulose through C6 monosaccharides and C3 trioses have been investigated over the last decade. Catalytic research with homogeneous catalysts has been focusing on the use of metal salt-based catalysts using Sn, Al, and Pb, as well as transitional metal salts, such as Zn, Ni, Fe, Co, and Cr [16–19]. Alkaline conditions have been used in the synthesis to improve the conversion and yield of lactic acid [5,17,20]. Although some of these homogeneous water-soluble salts can selectively convert glucose into lactic acid, these salts are generally expensive, highly toxic, corrosive, and/or difficult to recover afterwards [21]. Besides traditional homogeneous catalysts, solid heterogeneous catalysts are more attractive from an industrial perspective. Heterogeneous catalysts, such as Zr, Cr, Sn, Mo, W, and Pb metal oxides, as well as supported metal catalysts on zeolites and aluminum and silica oxides have been tested for the conversion of trioses, sugars, and cellulose to lactic acid [10,14,22–27]. Compared to some homogenous catalysts, the lactic acid yields obtained with heterogeneous metal oxide catalysts were lower. Marianou et al. tested SnCl2, SnO and Sn/γ-Al2O3 catalysts in glucose conversion to lactic acid with yields of ca. 33 %, 18 % and 20 %, respectively [26]. Takagi et al. used γ-AlO(OH) in the conversion of glucose to lactic acid in aqueous phase resulting yield of ca. 30 % [27]. Sn-beta-zeolites provided slightly higher yields of lactic acid. Dong et al. studied conversion of glucose with SnO2, ZnO, Sn-Beta and Zn-Sn-Beta obtaining lactic acid yields of 6%, 12 %, 23 % and 48 %, respectively [15]. However, only a few studies on the use of activated carbon in the production of lactic acid can be found in the literature. Some studies on the oxidation of glycerol to lactic acid with carbon-supported Pt and Pd noble metal catalysts have been carried out [28,29]. Onda et al. tested the conversion of glucose to lactic acid and gluconic acid by noble metals supported on activated carbon in an alkaline, aqueous solution with the lactic acid yield of 43 % [30]. Zhang et al. used activated carbon together with metallic Zn and Ni to improve the hydrothermal conversion of glucose into lactic acid (Y = 55 %) [20]. However, high temperatures and/or alkaline conditions were used in these studies.In catalysis, the support can play an important role, increasing the surface area and the stability of the catalyst [31,32]. The support may also improve the catalytic activity by acting as a co-catalyst and its chemical properties can be changed by specific surface functional groups and physical properties tuned by controlling the pore structures. Carbon-based support materials have been used for catalytic applications because of their properties, such as high surface areas, high thermal and chemical stability, low corrosion capability, and easy recovery from the reaction mixture [31]. Moreover, when compared to alumina and silica supports, activated carbon supports are less expensive, and the active phase can be recovered after use by burning away the carbon support [31]. As an attractive point, waste and residue biomasses can be used as a raw material in the preparation of activated carbon.In this study, various activated carbon-based heterogeneous metal oxide catalysts were tested in the conversion of glucose to lactic acid in aqueous solution. Sn, Al, and Cr oxides were used as catalysts on activated carbon supports. The prepared activated carbon supports and catalysts were characterized by multiple techniques. As a raw material for activated carbon, hydrolysis lignin, a waste fraction from lignocellulosic biomass hydrolysis was used. The effect of the metal and the activated carbon support prepared from hydrolysis lignin by chemical or steam activation was tested in a pressurized batch system. Finally, the effect of changes in the reaction conditions (temperature, time, and pressure) on the conversion of glucose and yield of lactic acid were studied, and catalyst reusability experiments were conducted. To our knowledge, metal oxides supported on biomass-based activated carbon have now been used to convert glucose to lactic acid in aqueous solution for the first time.Hydrolysis lignin from the biomass hydrolysis process was obtained from Sekab Ab, Sweden. The following catalyst preparation materials were used: anhydrous AlCl3 (99 %) from Alfa Aesar, CrCl3∙6H2O (98 %) from VWR, SnCl2∙2H2O (99 %) from Merck, ZnCl2 (97 %) from VWR, and HNO3 (65 %) from Merck. Anhydrous glucose (99 %) from Alfa Aesar, fructose (99 %) from Acros Organics, formic acid (>98 %) from Merck, hydroxyacetone (95 %) from Alfa Aesar, 5-(hydroxymethyl)furfural HMF (98 %) from Acros Organics, anhydrous lactic acid (98 %) from Alfa Aesar, levulinic acid (98 %) from Acros Organics, NaOH from VWR, NaHCO3 (99.7 %–100.3 %) from Alfa Aesar, Na2CO3 (99.5 %) from Alfa Aesar, and HCl (32 %) from Merck were used as reagents and/or standard materials.The activated carbon support was prepared from hydrolysis lignin by chemical activation with ZnCl2 or by physical activation using steam. Composition of used hydrolysis lignin is provided in Table S1 (Supplementary material). Hydrolysis lignin was dried in the oven at 105 °C and crushed to a particle size of < 0.42 mm for further use. Chemical activation was done by impregnation of the dried lignin with zinc chloride using a 2:1 mass ration of ZnCl2:biomass. ZnCl2 dissolved in H2O was mixed with the biomass for 3 h at 85 °C and then dried in the oven for about 3 days at 105 °C until a constant weight was reached. The carbonization and activation of the dried ZnCl2-impregnated lignin was done in a stainless steel tube in a tube furnace (Nabertherm RT200/13) at 600 °C for 2 h using a heating ramp of 10 °C min−1. During the thermal heating process, the reactor was flushed continuously with inert N2 gas (at a flow rate of 10 ml min−1). Carbonization, followed by physical activation of the biomass with steam, was done in a one-step process in a stainless-steel tube in a tube furnace using a heating ramp of 10 °C min−1 to a temperature of 800 °C; at the target temperature steam was added by feeding water at a flow rate of 0.5 ml min−1 into the reactor for 2 h. During the thermal heating process, the reactor was flushed continuously with inert N2 gas (at a flow rate of 10 ml min−1). The resulting activated carbon supports were washed with hot water, dried overnight at 105 °C, and crushed and sieved to a fraction size of 0.1–0.42 mm. The supports were named ACZ (AC zinc chloride-activated and water-washed) and ACS (AC steam-activated and water-washed). To modify the surface, the support materials were treated with 3 mol L−1 HNO3. The supports were named ACZN (AC zinc chloride-activated and HNO3-treated) and ACSN (AC steam-activated and HNO3-treated). The treatment was performed in a round-bottom flask, with a ratio of 10:1 mass ratio of acid:support and heated for 4 h at 85 °C. After the acid treatment, the supports were filtrated and washed with hot distilled water until a constant pH was obtained, and dried in the oven at 105 °C.Prior to the incipient wetness impregnation method, the pore volumes of the support materials were measured by the N2-physisorption method to calculate the volume of the impregnation solution. The amount of metal salts (SnCl2∙2H2O, AlCl3, and CrCl3∙6H2O) added by impregnation on the support were calculated by assuming the targeted concentration of metal (Sn, Al, Cr) in the catalyst was 2.5–10 wt.% of the total catalyst mass. Metal salts were added in distilled water, with a drop of concentrated HCl to dissolve the precursor salts. The impregnation solution was mixed with the support, matured for 4–5 h at room temperature, and finally dried in an oven at 105 °C for 16 h. The catalysts were thermally treated in a quartz tube in a tube furnace under a nitrogen atmosphere, using a constant flush of N2 (at a flow rate of 10 ml min−1). The thermal treatment was carried out at 350 °C, using a 5 °C min-1 ramp and a 3 -h holding time at the target temperature. All catalysts were tested in the reaction without further reduction treatment in a hydrogen atmosphere.Specific surface areas (SAs) and pore size distributions were determined from the physisorption adsorption isotherms using nitrogen as the adsorbate. Determinations were performed with a Micromeritics ASAP 2020 instrument (Micromeritics Instrument, Norcross, GA, USA). Portions of each sample (0.2 g) were degassed at a low pressure (0.27 kPa) and a temperature of 140 °C for 3 h to remove the adsorbed gas. Adsorption isotherms were obtained by immersing the sample tubes in liquid N2 (−196 °C) to achieve constant temperature conditions. Gaseous nitrogen was added to the samples in small doses, and the resulting isotherms were obtained. SAs were calculated from the adsorption isotherms according to the Brunauer–Emmett–Teller (BET) method. The precentral distribution of pore volumes (vol.%) was calculated from the individual volumes of micropores (pore diameter < 2 nm), mesopores (pore diameter 2–50 nm), and macropores (pore diameter > 50 nm) using the density functional theory (DFT) model.The morphology of the catalyst particles was studied using a JEOL JEM-2200FS energy-filtered transmission electron microscope (EFTEM) equipped for scanning transmission electron microscopy (STEM) at the Centre for Material Analysis, University of Oulu. The STEM model is used for images, energy-dispersive X-ray spectroscopy (EDS) analysis, and quantitative mapping of the catalyst. The catalyst samples were dispersed in pure ethanol and pretreated in an ultrasonic bath for several minutes to create a microemulsion. A small drop of the microemulsion was deposited on a copper grid pre-coated with carbon (Lacey/Carbon 200 mesh copper) and evaporated in air at room temperature. The accelerating voltage in the measurements was 200 kV, while the resolution of the STEM image was 0.2 nm. The metal particle sizes were estimated visually from high-resolution STEM images of each sample.The metal contents of the supports and catalysts were measured by ICP-OES using a Perkin Elmer Optima 5300 DV instrument. Samples weighing 0.1–0.2 g were first digested with 9 ml of HNO3 at 200 °C for 10 min in a microwave oven (MARS, CEM Corporation). Then, 3 ml of HCl was added, and the mixture was digested at 200 °C for 10 min. Finally, 1 ml of HF was added, and the mixture was again digested at 200 °C for 10 min. Excess HF was neutralized with H3BO3 by heating at 170 °C for 10 min. Afterwards, the solution was diluted to 50 ml with water, and the elements were analyzed by the ICP-OES method.The total ash content was determined by using SFS-EN 14,775 standard method.X-ray diffractograms were recorded with the PANalytical X′Pert Pro X-ray diffraction (XRD) equipment using monochromatic CuK α1 radiation (λ =1.5406 Å) at 45 kV and 40 mA. Diffractograms were collected in the 2θ range of 5°–80° at 0.017° intervals, with a scan step time of 110 s. The crystalline phases and structures were analyzed with the HighScore Plus program.X-ray photoelectron spectroscopy analyses were performed using the Thermo Fisher Scientific ESCALAB 250Xi XPS System. The catalyst samples were placed on an indium film, with a pass energy of 20 eV and a spot size of 900 μm; the accuracy of the reported binding energies (BEs) was ±0.3 eV. Sn, Al, Zn, Cr, O, C, and N elemental data were collected for all samples. The measured data were analyzed with the Avantage V5 software. The monochromatic AlKα radiation (1486.7 eV) was operated at 20 mA and 15 kV. Charge compensation was used to determine the presented spectra, and the calibration of the BEs was performed by applying the C1s line at 284.8 eV as a reference. The approximate detection depth of the analysis was < 10 nm.Elemental analysis of the prepared AC supports was performed by using a Flash 2000 CHN-O Organic elemental analyzer by Thermo Scientific. The ground and dried sample (about 1 mg) was placed in the analyzer and mixed with vanadium pentoxide (V2O5, 10 mg) to enhance the burning. The prepared sample was combusted at 960 °C for 600 s using methionine as a standard for the elements: C, H, and N, whereas the standard used for oxygen was 2,5-(bis(5-tert-butyl-2-benzoaxazol-2-yl)thiophene (BBOT).The surface acidity and basicity of the ACs and catalysts were characterized according to the Boehm titration method [33,34]. The samples (0.1–0.2 g) were weighed and separately mixed with 50 ml of 0.01 mol L−1 solutions of HCl, NaOH, NaHCO3, or 0.005 mol L−1 Na2CO3, and shaken for 72 h in sealed vials at room temperature. The solutions were filtered with ashless filter paper. Acidic groups were determined by the back-titration method—taking 10 ml of each filtrate, mixing with 20 ml of 0.01 mol L−1 HCl and finally back-titrating with 0.01 mol L−1 NaOH using potentiometric titration. The acidic groups on the AC were calculated [35] based on the theory that NaOH neutralizes carboxylic, lactonic, and phenolic groups and Na2CO3 neutralizes carboxylic and lactonic groups, while NaHCO3 neutralizes only carboxylic groups. Basic sites were calculated by back-titration with 0.01 mol L−1HCl; the total number of basic groups were calculated with the assumption that HCl neutralizes the basic groups on the AC surface.Fourier Transform Infrared (FTIR) spectra were obtained using an ATR-FTIR spectrometer (Perkin Elmer Spectrum One) with diamond/ZnSe crystal. The scans were obtained in the spectral range of 4000–650 cm–1 with a resolution of 4 cm–1, and 20 scans for each sample. Reference (blank) FTIR spectra were obtained from clean crystal.For catalyst testing, HEL’s manual DigiCAT pressure reactor with the hotplate and stirrer system and three parallel Mini-Range stainless steel reactors (50 mL), each with individual manometers for pressure control, was used. In a typical reaction, the catalyst was added to the reactor with 0.100 g of glucose in 20 ml of ultrapure H2O. The reactor was purged with nitrogen and continuously mixed at 500 rpm. Heating was started after purging with nitrogen. When a reaction temperature of 180 °C was reached (in about 25 min), nitrogen was fed into the system until the desired final reaction pressure of 30 bar was attained (in about 5 min). The temperature was maintained constant through the heating of the external aluminum block in which the parallel reactors and external temperature probe were placed. After the reaction time (0–300 min), samples were collected from the reactor through the outlet vent and filtrated with a 0.45 μm (polyethersulfone) membrane filter for further product analysis. Two tests were run in parallel and the error was presented as a percentage of the average standard deviation of the two parallel tests. Recycling of the catalyst was performed after filtering the catalyst out from the reaction solution and washing it with 10 ml ethanol three times. The catalyst was then used under the same reaction conditions as in the previous cycle. Due to collection loss, the weight of the catalyst was slightly lower than the initial weight after recycling; the corresponding glucose loading was reduced to keep the weight ratio of catalyst to glucose constant.The product analysis was performed with a Shimadzu High-Performance Liquid Chromatograph (HPLC) with a Shodex RI detector. The quantification was based on external calibration using standard solutions of glucose, fructose, lactic acid, levulinic acid, HMF, formic acid and hydroxyacetone. The liquid samples were analyzed with a Shodex SUGAR SH1821 column (8.0 mm ID × 300 mm) with pre-column SUGAR SH-G using 5 mmol L−1 H2SO4 as a mobile phase with flow rate 1.0 ml min−1 and a column temperature of 40 °C. Conversion and yield were calculated from the results of the quantification by HPLC by using the following equations: (1) C o n v e r s i o n   ( % )   = C g l u c o s e   i n i t i a l - C g l u c o s e   a t   e n d   o f   r e a c t i o n C g l u c o s e   i n i t i a l (2) Y i e l d   ( % )   = C m e a s u r e d   L a c t i c   A c i d   i n   t h e   s a m p l e   C t h e o r e t i c a l   m a x .   o f   L a c t i c   A c i d   i n   t h e   s a m p l e The theoretical maximum of lactic acid in the sample (Eq. 2) is calculated by assuming that 2 mol of lactic acid is obtained from 1 mol of glucose.Compounds from product mixture were identified using Agilent 8890 GC System equipped with a mass detector (MS) and an Agilent HP5-MS Ultra Inert GC column (30 m x 0.25 mm x 0.25 mm). The oven temperature was programmed at 70 °C for 1 min, then increased to 280 °C at 10 °C min−1 and kept for 5 min. The injection volume was 1 ml and He flow 1 ml min−1. Mass spectra were collected with an electron impact ionization of 70 eV. The Full-scan acquisition was performed with the mass detection range set at m/z 35–500. Data acquisition and analysis were executed by 5977B GC/MDS (Agilent Technologies).The BET SAs, average pore diameters, and DFT pore volumes and pore distributions of the AC supports and catalysts were measured by N2-physisorption analysis, and the results are listed in Table 1 . According to the analysis, the SA and the pore volume values of chemically activated ACZ were higher (1595 m2 g−1 and 0.78 cm3 g−1) than those of steam-activated ACS (760 m2 g−1 and 0.47 cm3 g−1). After treatment with HNO3, the SAs and pore volumes of both ACS and ACZ decreased about 30 %, probably due to addition of functionalities (see XPS and Boehm titration analysis) or due to the collapse of the pore walls [36–38]. However, the mesoporous volumes of both catalyst supports (ACZN and ACSN) treated with HNO3 were similar, and the main difference was in the micropore volumes; the chemically activated support ACZN had more micropores than the ACSN support. The ACZN support was mainly microporous carbon, with 66 vol.% micropores and 34 vol.% mesopores. In contrast, ACS and ACSN were about 50:50 vol.% of micro:meso pores. For all supports, macropore volumes were zero. With metal impregnation, SAs and pore volumes decreased, indicating the addition of metal in the pores (Table 1). The addition seemed to occur mainly in the mesopores, since a greater decrease in their volumes was detected, though microporous volumes were also partly filled. For some catalysts, a greater decrease in porosity seemed to occur, in the following order: Sn/Al5/2.5@ACS > Sn/Al5/5@ACZN > Sn/Al5/2.5@ACZN > Sn/Cr5/2.5@ACZN > Sn10@ACZN > Sn/Al5/[email protected] contents of the supports and catalysts were measured by total ash content analysis and by ICP-OES. Results of the ash analysis and ICP-OES analysis from activated carbon supports ACZ, ACZN, ACS, and ACSN are presented in Table S2 (Supplementary material). Without acid treatment, the chemically activated ACZ support contained about 5 wt.% of zinc (Table S2) and total metal content was 7.6 wt.% by ash analysis. After ACZ was treated with HNO3, zinc metal was removed from the support and ash content decreased to zero. The steam-activated support ACS contained 0.5 wt.% of Ca and Na; other metals, such as Fe, Mn, Mg, K, and Zn were present at less than 0.1 wt.% and total ash content was 2.3 wt.%. After acid treatment, total residual metal content was less than 0.3 wt.% for both ACZN and ACSN.The active metal contents of the catalysts Sn, Al, and Cr were measured by ICP-OES analysis. The results are presented in Table 1. From the ICP-OES analysis, determined Sn and Al contents were close to the targeted ones. Slightly lower amounts of Sn and Al were detected from the non-acid-treated catalyst Sn/Al5/2.5@ACS. In some cases, acid treatment and the presence of oxygen functionalities (see XPS and Boehm-titration) on the surface has been claimed to be important, by making the surface more accessible for metal precursors at the catalyst preparation and impregnation step—i.e., acting as anchoring sites for the metal precursors on the surface—or by making the surface more hydrophilic, so that the metal precursors can adsorb to the internal surface of the pores [31,39,40].The morphology of the catalyst particles was studied with an EFTEM/STEM microscope. The STEM mode was used for images and combined with EDS analysis and quantitative mapping to detect the elemental composition of the materials. The STEM-high angle annular dark field (STEM-HAADF) image of the Sn/Al5/2.5@ACSN catalyst and the quantitative mapping presented in Fig. 1 show an equal distribution of aluminum and tin on the surface of the AC. The STEM-HAADF images from all the supports and catalysts are presented in Fig. S1 and Fig. S2 (Supplementary material). From the images of the supports, large particles of zinc oxide are seen on ACZ (Fig. S1 c), which disappear from ACZN after acid treatment (Fig. S1 d). From the catalyst images (Fig. S2), the particle sizes on the surfaces were estimated to be less than 10 nm, with a particle size of approximately 3–5 nm; some aggregations were also detected from the surfaces (∼ 20 nm). The N2-physisorption analysis indicated that mainly mesopores and some of the micropores were filled by impregnation of the catalyst particles on the support. The STEM images verified that small nanoparticles were able to enter into the support pores. However, some of the smallest pores can be blocked by the particles at the pore entrance. Especially for the catalysts Sn/Al5/5@ACZN and Sn/Al5/2.5@ACZN (Figs. S2 b and c), there appeared to be some aggregations, which could be due to an accumulation of nanoparticles in the pores. These could be the reason for the larger decrease in the BET SA and pore volumes. Overall, the particles seemed to be distributed uniformly on surfaces with small particle sizes.XRD analysis was performed to verify the metal phases of the catalysts. The diffraction patterns are presented in Fig. S3 (Supplementary material). For all supports and catalysts, broad peaks at 23.8° and 44.2° were detected, representing amorphous carbon. For ACZ support, peaks at 31.7°, 34.3°, 36.1°, 47.4°, 56.5°, 62.7°, 67.8°, and 68.9°, representing ZnO (JCPDS file No. 04-003-2106) were detected, which disappeared after acid treatment (see Fig. S3 a, XRD pattern of ACZ and ACZN). For Sn-catalysts (Fig. S3 b), the oxidized metal phase of tin(IV)oxide (SnO2) (JCPDS file No. 00-001-0657) was detected, though the peaks representing SnO2, at around 26.3°, 33.5°, 51.7°, and 64.2°, were rather small for bi-metal catalysts. For Al and Cr, no clear peaks could be detected, probably because of the small particle size or a low concentration of the catalyst material. Overall, diffractograms presented no high crystallinity for Sn, Al, or Cr, indicating the presence of amorphous phases and/or very small particles on the surface of the support, which were also seen on the STEM images.The metal phases of the catalysts were confirmed with XPS (Fig. S4 and Table S3, Supplementary material). According to the analysis, metal oxides with peaks at 487.2 eV and 495.6 eV (Fig. S4 a) corresponded to Sn3d5/2 and Sn3d3/2 [25], respectively, representing tin(IV) oxide (SnO2), and with peaks at 74.5 eV (Fig. S4 b) corresponded to Al2p oxide, most likely representing AlO(OH) [41], were detected from the spectrums of catalysts containing Sn and Al on support ACS, ACSN, and ACZN. The XPS spectrum peaks of the Sn/Cr catalyst on ACZN (Figs. S4 a and c) at 487.2 eV and 495.6 eV corresponded to Sn3d5/2 and Sn3d3/2, respectively, representing SnO2, and peaks at 577.4 eV and 586.9 eV corresponded to Cr2p3/2 and Cr2p1/2 [42], respectively, representing Cr2O3. The support ACZ spectrum (Fig. S4 d) peaks at 1045.7 eV and 1022.6 eV corresponded to Zn2p1/2 and Zn2p3/2, respectively, representing ZnO [43].The surfaces of the AC supports were analyzed with XPS, FTIR and Boehm titration for information about their content and functionality. The elemental analysis (C,H,N,O) of the support material was compared to the total carbon, oxygen and nitrogen content obtained by XPS; however, elemental analysis analyses the total bulk material, while XPS analyses only the uppermost layer. Examples of the C1s and O1s spectra from the XPS analysis are presented in Fig. S5 (Supplementary material). The peaks from the C1s spectra indicated that most of the carbon was present as graphitic conjugated carbon (at 284.8 eV) and non-conjugated carbon (at ∼285 eV). Also, carbon-oxygen type functionalities were present at 286.6 eV (from phenolic, alcoholic, or etheric functional groups) and at 288.8 eV (from carboxylic, anhydride, ester, or lactone groups) [44–47] (Table S3, Supplementary material). The highest total carbon content, detected by both XPS and elemental analysis, was in steam-activated ACS and it decreased after the chemical treatments due to the addition of heteroatoms (Table 2 ). The more drastic decrease in the total carbon content, detected by XPS rather than by elemental analysis, indicated that the functional groups were attached to the surface. The XPS O1s analysis of the chemically activated ACZ showed a higher total oxygen content than that of ACS, most likely due to oxygen atoms bound to Zn as ZnO [48], which were present from the preparation step. When activated carbon was treated with nitric acid, the total oxygen content of the ACSN was three times higher than on the surface of the plain steam-activated catalyst support (ACS) (Table 2). The nitric acid treated support ACZN showed a higher content of total oxygen than ACZ and four times higher than ACS. Similarly, confirmed by elemental analysis, the oxygen content of the HNO3 treated supports was about four times higher than of the untreated ones and was highest in the ACZN support. The increase of oxygen functionalities after HNO3 treatment was detected mainly as oxygen functionalities from the O1s scan at 532.3 eV (Table S3, Supplementary material), which can be identified as carbonyl oxygen from functionalities such as lactone, ester, carboxylic or anhydride, and oxygen atoms from phenol or ether groups. However, the identification is not clear since the same functionality can give a signal at different BEs, and it also depends on the fittings of the peaks [44,45,47,49–51]. It has been reported that oxidation by nitric acid treatment increases the oxygen content on the activated carbon, and especially the number of acidic functionalities such as carboxyl groups [36,38,51,52]. A small increase in the nitrogen content was also detected on ACs after nitric acid treatment. The nitric acid oxidation is also known to result in a number of nitro groups via the nitration mechanism of aromatic ring [53]. This was indicated by the FTIR spectrum of ACZN (Fig. S6 a, Supplementary material).The acidic and basic group concentration of the supports was studied with the Boehm titration method (Table 2). The titration indicated that only minor amounts of acidic functionalities and some basic functionalities (e.g., chromene or pyrone [31]) were present on ACS. The explanation in the literature is that the high temperatures (800 °C) used for steam activation can destroy most of the functional groups on the surface of AC [51], seen also from the FTIR spectrum (Fig. S6 a). A slightly higher total acidic group concentration was observed on chemically activated ACZ as compared to the steam-activated ACS. This could be due to the lower temperature (600 °C) used in the preparation of ACZ, leaving more functional groups on the surface, or from the presence of ZnO. According to the Boehm method (Table 2), after nitric acid treatment, a higher content of total acidic groups, at 1.2 mmol g−1 and 2.7 mmol g-1 were detected on ACSN and ACZN, respectively, than on untreated ACS and ACZ, due to an increase in carboxylic, lactonic, and phenolic groups (Fig. S7, Supplementary material). This was in good agreement with the results from the XPS analysis. No basic groups were detected on the ACZ, ACZN, or ACSN surfaces with Boehm titration.After the addition of metals to the supports, the total carbon content decreased, while the oxygen content increased, most likely due to the addition of metal oxides or hydroxides, which can be seen from the metal-oxygen bonds at 531.2 eV or 532.3 eV [54–56] (Table S3, Supplementary material) according to the XPS analysis. A decrease in nitrogen content was also noted after the addition of metal (Table S3), indicating that nitrate/nitro groups introduced to the surface during oxidation with nitric acid decreased after catalyst preparation, since the nitro groups begin to decompose at about 270 °C [53]. Boehm titration analysis was conducted for the catalysts by back-titrating with NaOH to determinate their total acidic group concentration. Besides the acidic groups (carboxylic, lactonic, and phenolic) on the aromatic carbon framework, inorganic components such as metal hydroxides may take up protons and/or precipitate during the acidification step in the Boehm titration procedure, thereby affecting the total acidity of the sample [57]. The total acidity increased after the addition of metal oxides to ACS and ACSN, and was 1.1 mmol g−1 and 1.7 mmol g-1 for Sn/Al5/2.5@ACS and Sn/Al5/2.5@ACSN, respectively. This is most likely due to oxidized metals or acidic species created during the metal impregnation on AC, seen also as an increase in O1s XPS analysis at 532.3 eV, indicating an increase in metal-oxygen bonds or carbon-oxygen bonds (Table S3, Supplementary material). On the other hand, for the catalysts impregnated on ACZN, the total acidity did not change after the addition of metals and was about 2.6 mmol g−1 for Sn10@ACZN, Sn/Al5/2.5@ACZN and Sn/Cr5/2.5@ACZN. Further investigations using Boehm titrations were performed for the heat-treated ACs. Heat treatment was performed in a similar manner as for the catalysts—at 350 °C for 3 h under N2. As a result, the total acidity decreased by about 20 %, and mainly the carboxylic groups seemed to break down (Fig. S7, Supplementary material); these are groups that break easily at lower temperatures [44,51]. This could be why the total acidity did not increase with the addition of metal oxides on the ACZN support—it contained more carboxylic groups, which could have decomposed in the heat treatment process during the catalyst preparation. Presence of carbonyl groups after catalyst preparation and heat treatment on carbon was verified from the FTIR spectrum (Fig. S6 b), however, peaks indicating the presence of nitro groups were absent. This was in agreement with the XPS and titration analyses.Preliminary studies with chlorides of Sn, Al, and Cr, and chlorides of Sn + Al, Al + Cr, and Sn + Cr metal combinations were done to study the conversion of glucose to lactic acid under the following conditions: 2 h at 180 °C, 30 bar, and 500 rpm, using 0.100 g of glucose in 20 ml of H2O. SnCl2*2H2O, AlCl3, or CrCl3*6H2O were used as single catalysts at a concentration of 0.1 mmol, and at concentrations of 0.05 mmol each metal in combination. The reaction parameters and catalyst amounts were adapted from Deng et al. [16]. Results from the test are shown in Fig. 2 . For all the homogeneous metal salts tested, except SnCl2, 100 % conversion of glucose was obtained. The lactic acid yield of the single Sn catalyst was low (7%). Single Cr and Al catalysts had higher lactic acid yields (∼20 %); however, levulinic acid was produced in an almost 1:1 ratio with lactic acid. The lactic acid yield from the Al + Cr combination was about the same as that from single Al and Cr catalysts, though a lower amount of levulinic acid was produced. The highest lactic acid yields, 37 % and 36 %, were obtained from the Sn + Cr and Sn + Al combinations, respectively. Based on the preliminary study results, the best working metal combinations (tin combined with aluminum or chromium) were selected as heterogeneous catalysts on activated carbon for further analyses of lactic acid production.The conversion of glucose (0.100 g) by heterogeneous metal catalysts supported on activated carbon was studied at 180 °C, 500 rpm, 30 bar with 0.100 g of the catalyst in 20 ml of H2O (Fig. 3 ). The effect of the metal was studied using a carbon-supported tin oxide catalyst as well as tin oxide combined with oxides of aluminum or chromium on the carbon support. Furthermore, the ratio of the metals in the catalysts was modified. Chemically activated and nitric acid-treated activated carbon (ACZN) was used as support. ACZN treated in a manner similar to the catalysts, in a N2 atmosphere at 350 °C for 3 h, was used as a reference sample.After two hours, without a catalyst in the reaction system, the conversion of glucose was 48 %, and HMF was the main synthesis product with a yield of 23 % with a minor lactic acid yield of 5%. When ACZN was added in the reaction, a slightly higher fructose yield was noticed; however, the lactic acid and HMF yield was about the same as that without the catalyst. With the addition of 10 wt.% of tin to the support (corresponding to 0.084 mmol of Sn in 0.1 g of catalyst), the conversion was almost complete (94 %) and the formation of lactic acid increased to 19 % with Sn10@ACZN. Results indicated that the presence of SnO2 in the support promotes the conversion of glucose and the formation of lactic acid instead of HMF. With the addition of aluminum at 5 wt.% with 5 wt.% of tin (0.042 mmol of Sn and 0.19 mmol of Al), the lactic acid yield increased to 27 %, and complete conversion of glucose was achieved with the Sn/Al5/5@ACZN catalyst. However, when the aluminum/tin ratio was changed to 5/2.5 wt.% (0.042 mmol of Sn and 0.093 mmol of Al), the highest lactic acid yield (31 %) was reached with the Sn/Al5/2.5@ACZN catalyst. The trend indicated that the addition of a small amount of Al improved the formation of lactic acid, while the higher loading did not improve it further. Deng et al. noticed that for homogeneous catalysts, selectivity for lactic acid was highest when the molar ratio of Sn/(Sn + Al) was 0.5, and it decreased for lower or higher ratios [16]. In our studies, we see a similar trend, with lactic acid yield decreasing in the order Sn/Al5/2.5@ACZN > Sn/Al5/5@ACZN > Sn10@ACZN, where the Sn/(Sn + Al) molar ratio was 0.30, 0.19, and 1.00, respectively. Lewis acids, such as tin, are found to catalyze the isomerization of glucose into fructose, as well as the retro-aldol reaction of hexoses into trioses [10,26] (see Scheme 1). Other research groups have studied glucose conversion with supported Sn10/γ-Al2O3 catalysts that yielded 20 % lactic acid and 25 % HMF [26]. Holm et al. tested Sn-zeolites and obtained a lactic acid yield of 26 % in a water solution [10]. Aluminum oxide was also used as a solid Lewis acid catalyst for converting trioses, with a lactic acid yield of 28 % [27]. Rasrendra et al. found that the aluminum and chromium metal salts were the most active catalysts converting trioses such as dihydroxyacetone and glyceraldehyde to lactic acid in water [19]. Takagaki et al. found that a chromium oxide catalyst could easily transform pyruvaldehyde to lactic acid, with a yield of 46 % [22]. Xia et al. found that the Cr/(Cr:Sn) molar ratio of 0.5 on a Cr-Sn-Beta zeolite gave the highest yield of lactic acid in the conversion of glucose [58]. However, in this study, the addition of chromium oxide with tin oxide to the Sn/Cr5/2.5@ACZN catalyst (0.042 mmol of Sn and 0.044 mmol of Cr, at a molar ratio of 0.5), did not improve the production of lactic acid, as compared to the Sn/Al catalyst. Instead, the addition of chromium oxide seemed to direct the reaction towards the production of HMF and levulinic acid, in contrast to the Sn/Al catalyst.To conclude, the highest yield of lactic acid produced was with the combination of Sn with Al, at a ratio of 5:2.5 (Sn/Al wt.%) on ACZN. This combination was selected for use in the following examinations for conversion of glucose to lactic acid.The effect of the catalyst support on the conversion of glucose to lactic acid was studied at 180 °C, 30 bar, 500 rpm using 0.100 g of the catalyst (0.042 mmol of Sn and 0.093 mmol of Al) and 0.100 g of the glucose in 20 ml of H2O. The activated carbon supports ACS, ACSN, ACZ, and ACZN were tested as reference samples (treated in an N2-atmosphere at 350 °C for 3 h similar to the catalysts). The supports were impregnated with the Sn/Al catalyst at a ratio of 5/2.5 wt.%. ACZ was used as a support only after treating it with nitric acid, since it contained 5 wt.% zinc (see Table S2). The results of the conversion and main product yields are presented in Fig. 4 . Other byproduct yields were not quantified.After two hours, the conversion of glucose in the batch system for ACSN and ACZN reached about the same conversion rate as that without the catalyst (Fig. 3), and no notable differences in the yield of lactic acid were observed. With ACS, a higher conversion rate was noticed, but the lactic acid yield was about the same as before. Furthermore, slightly higher fructose and HMF yields were noticed with an increasing order of ACS < ACSN < ACZN, which was in the same order as the increase in total acidity of the ACs (see Table 2). This could be due to the Brønsted acids (acidic groups) on the ACs, which can catalyze fructose to HMF and its derivatives [59]. In contrast, ACZ yielded 27 % lactic acid and a 97 % conversion rate, and the HMF yield shifted to levulinic acid. It was found that ZnO in the support (see ICP-OES, XRD, and XPS results) can act as a catalyst and convert glucose to lactic acid [15,34]. However, ACZ was not used as a support for Sn/Al without treating it first with nitric acid to remove residual ZnO as it already contained the active metal oxide.With the addition of Sn and Al oxides to the ACS support, the Sn/Al5/2.5@ACS catalyst provided glucose conversion of 98 % and a lactic acid yield of 24 %. With the acid-treated Sn/Al-catalysts, the glucose conversion rate reached 100 % and the highest lactic acid yields of 31 % and 34 % were reached with the Sn/Al5/2.5@ACZN and Sn/Al5/2.5@ACSN catalysts, respectively. Also, there was a shift from HMF to levulinic acid as the main byproduct. Results indicated that nitric acid treatment of the catalyst support had an impact on the conversion of glucose to lactic acid, seen in comparison to the untreated catalyst support (Sn/Al5/2.5@ACS). Further analysis of the reaction mixture was performed with GCMS. GCMS chromatograms and MS spectra (Fig. S8 and S9, Supplementary material) revealed that lactic acid was the main product with all catalysts. Also, other monocarboxylic acids such as formic acid and acetic acid were identified from the product mixture. With more acidic catalysts (Sn/Al5/2.5@ACSN and Sn/Al5/2.5@ACZN) byproducts formed besides the HMF and levulinic acid were other furfural derivatives such as 5-methyl furfural and 2,5-furandicarboxaldehyde and amount of byproducts varied during the reaction time (Fig. S10, Supplementary material). Compounds such as dihydroxyacetone, glyceraldehyde were also identified from the product mixture indicating that reaction towards lactic acid was going through the formation of dihydroxyacetone as intermediate (Fig. S10). The main byproduct with ACZ catalyst was identified as acetol (hydroxyacetone) (Fig. S8). It has been found that acetol and lactic acid are formed competitively from the same intermediate, i.e. pyruvaldehyde, via hydrogenation on metal surfaces [60–62]. Without the presence of external hydrogen in the reaction system, the hydrogenation can occur via transfer hydrogenation by hydrogen donor such as formic acid for example [63]. With ACZ, the acetol production was prominent compared to more acidic Sn/Al catalysts, which on the other hand, directed the reaction towards HMF and its derivative’s production. Other byproducts formed such gaseous products or humines in the reaction mixture were not analyzed. It has been claimed that Lewis acid sites on heterogeneous catalysts are crucial for lactic acid production, while the Brønsted acid sites play no role at all [64]. Rasrendra et al. found that Brønsted acids, namely H+ ions, had a positive effect on the conversion of dihydroxyacetone to pyruvaldehyde while Lewis acid sites played a key role with conversion of trioses to lactic acid [65]. Clippel et al. studied dihydroxyacetone conversion to lactic acid and it’s esters with bifunctional carbon-silica catalysts and demonstrated that the presence of weak Brønsted acid sites originating from oxygen-containing functional groups in the carbon part were crucial in accelerating the dehydration reaction [66]. This indicated that acidic sites on the carbon surface could participate on the conversion of dihydroxyacetone to pyruvaldehyde and further to higher yields of lactic acid, explaining why more acidic Sn/Al5/2.5@ACSN was performing better than less acidic Sn/Al5/2.5@ACS (see chapter 3.1.1). Furthermore, when comparing Sn/Al5/2.5@ACSN and Sn/Al5/2.5@ACZN, the higher acidic group concentration on ACZN compared to ACSN did not seem to have notable effect on the conversion rate or the lactic acid yield or on changes in the byproduct yields after addition of Sn/Al. The higher SA of the Sn/Al5/2.5@ACZN catalyst also did not seem to contribute to the higher activity or lactic acid yield for the catalyst, as compared to the steam-activated Sn/Al5/2.5@ACSN with a lower SA. Both catalysts had almost similar mesoporous structures, indicating that having a more microporous structure is not beneficial for lactic acid production. It is possible that the molecules are too big to fit in the smallest micropores, or that the desorption of the products is difficult. This could be also the reason why higher acidic group concentration on the Sn/Al5/2.5@ACZN did not have an effect on the conversion rate or the lactic acid yield if the active acidic sites in the smallest micropores were not accessible for the reactants or were blocked. It has been shown that oxidized carbon catalysts, specially carboxylic groups on carbon, take place in dehydration reactions, however, not only the number of the acidic groups but also their location and accessibility are relevant for the reaction [67–69]. Results indicated that the environmentally hazardous ZnCl2 treatment step could be removed from the pretreatment process of the AC support, since the more microporous surface has no beneficial role to play in the conversion process or in the yield of lactic acid after the addition of Sn/Al to the support. Even though, it was noted that presence of ZnO and a lower amount of acidic groups in the catalyst seemed to produce lactic acid and inhibit the formation of HMF and its derivatives, the addition of Sn/Al oxides on acidified catalyst support resulted in the best lactic acid production levels. Moreover, the addition of zinc on the support cannot be controlled since it is left over from the activation process. Detailed studies regarding the preparation of carbon-supported zinc oxide catalysts and their utilization in lactic acid production are ongoing and will be reported at a later date.Based on the discussion above, we propose a reaction route presented in Scheme 2 . as the reaction pathway for production of lactic acid and main byproducts with the AC based catalyst. The detailed reaction mechanism remains in question at the moment.Based on studies of the effect of metal and support interactions on glucose conversion to lactic acid, the Sn/Al5/2.5@ACSN catalyst was selected for optimization studies in the aqueous phase. The effects of catalyst loading, time, temperature, and pressure on the conversion of glucose to lactic acid were studied. Glucose at a concentration of 0.100 g was used and mixed at 500 rpm in 20 ml of H2O in all the experiments. The effect of catalyst loading at a temperature of 180 °C, with increasing catalyst amounts (0.050, 0.100, 0.150, and 0.200 g) was studied (Fig. 5 ). A catalyst load of 0.050 g was too low, as 100 % conversion of glucose was not achieved even after 3 h. With higher catalyst loading (≥ 0.100 g), the glucose conversion rate reached 100 % within 60 min, and a catalyst concentration of 0.100 g yielded 34 % of lactic acid. Using 0.150 g of the Sn/Al5/2.5@ACSN catalyst, a yield of 37 % was achieved after 30 min. The highest yield of lactic acid (42 %) was achieved in 30 min using 0.200 g of the catalyst, and the yield was constant even at a reaction time of 180 min. An overview of some of the used heterogeneous catalysts in the conversion of biomass to lactic acid is presented in Table 3 . Comparing our results to those, our catalyst provided reasonable yields of lactic acid in glucose conversion.The effect of the reaction temperature on the conversion of glucose to lactic acid was studied at 160, 180 and 200 °C using 0.200 g of the Sn/Al5/2.5@ACSN catalyst and 30 bar reaction pressure (Fig. 6 ). The complete conversion of glucose at 160 °C took 120 min to achieve and was slower than at higher temperatures (180 and 200 °C) where the reaction rate increased and conversion was almost complete (≥ 98 %) at 20 min. Besides, the increase in the lactic acid yield from 35 % to 42 % was detected when the temperature was increased from 160 °C to 180 °C, but a further increase in the temperature did not increase the lactic acid yield. During the reaction temperature increase, byproducts (HMF and levulinic acid) were formed faster (Fig. S11, Supplementary material). At 200 °C, the color of the reaction solution was much darker (Fig. S12, Supplementary material), even though the lactic acid yield was not higher, indicating that other, unwanted soluble byproducts (including polymeric materials, such as humines) could have formed from the decomposition of reaction products at higher temperatures [22]. This suggested that a temperature of 180 °C was optimum for the reaction and a further increase in temperature was unfavorable for the selectivity of the reaction.Different types of atmospheres notably influence the yields of lactic acid, and an atmosphere of pure nitrogen was found to yield the highest amount of lactic acid compared to oxygen and air [70,71]. Sun et al. found that an increase in pressure gave higher lactic acid yields up to 40 bar (He), after which the yield started to decrease [72]. An atmosphere of nitrogen was used in our catalytic studies. The effect of increasing pressure on the conversion rate and the yield of lactic acid was tested at ∼5, 20, 30, and 40 bar. The pressure increase was carried out by the addition of inert gas (N2) into the system. The pressure of the reaction itself was about five bar, when no excess pressure was added. As seen in Fig. 7 , the pressure increase from 5 to 40 bar did not seem to have a major effect on either the conversion rate or the yield of lactic acid. A slightly lower conversion rate was noted after 20 min when no pressure was added, but the difference was only 4%, and after 60 min, the conversion rate was 100 % for all pressures. The lactic acid yield was about the same at every tested pressure. This was noted as a positive result since excess pressure does not need to be added to the system.Reusability tests of the Sn/Al5/2.5@ACSN catalyst were performed to obtain information about the stability of the catalyst. Tests were performed at 180 °C, at 5 bar (pressure of the reaction itself) and 30 bar with 0.2 g of the catalyst and 0.1 g of glucose in 20 ml of H2O. The catalyst was filtered after 120 min, washed with ethanol, dried in the oven at 105 °C, and tested in the same conditions as in the first run at four cycles. Ethanol wash was selected instead of water after reuse, as the lower-polarity solvent could access the catalyst pores easier. As a result of the reuse (Fig. 8 ), glucose conversion decreased from 100 % to 86 % after the first cycle, and finally to 80 % after four cycles. Yield of lactic acid decreased from about 40 % after the first cycle to 15 % after four cycles. More fructose was detected in the reaction solution, indicating the reaction was not completed, and yield of HMF increased from 5% to 15 %. No differences were detected in the conversion rates and yields at the different reaction pressures used (∼5 to 30 bar).After catalyst reuse, the reaction liquid was analyzed using the ICP-OES method, and the metal content was determined after each recycle of the catalyst. The hot, acidic medium can promote the solubility of some metal oxides and cause deactivation of the heterogeneous catalysts by leaching the metal or metal oxides [73]. In these experiments, the pH of the final product solution was about 2. After the first run, 0.2 % of tin was leached from the initial amount of Sn added to the catalyst; 15 % of aluminum was leached from the initial metal content added to the catalyst. After the second run, leaching was 0% for Sn and 1.0 % for aluminum. This indicated that some of the Al was leaching; however, it was relatively minor, and Sn was quite stable on the catalyst. The SA and the pore volumes of the catalyst were also determined after four recycles. The BET SA of the catalyst was 97 m2 g−1, with a pore volume of 0.06 cm3 g−1, resulting in an 82 % decrease of the BET SA and a 75 % decrease of the pore volume, leading to the conclusion that most of the SA and porosity was lost during the use of the catalyst. Blocking of the catalytic sites on the porous structure of catalyst by adsorbed reaction products or carbon deposits seems to be the cause of the catalyst deactivation. Moreover, the product distribution after fourth reuse seems to be similar as with the plain AC support, indicating blocking of the active metal sites on catalyst support. In the aqueous solution, carbonaceous byproducts, such as humines, can be formed during the reaction, which can block the pores, deposit on the active sites, and finally deactivate the catalyst [66,74,75]. In other research, calcination at mild conditions was used successfully to reactivate the catalyst after residue deposition [76]. After fourth reuse, the catalyst was calcined in air at 300 °C for two hours to remove the deposits on the catalyst surface. After calcination, The BET SA of the catalyst was 640 m2 g−1, indicating that the pore structure was opened. However, lactic acid yield was the same as after second reuse (Fig. 8), suggesting that leaching of aluminum was also the reason for decrease in lactic acid yield after reuse. This verified the fact that co-operation of the two metals; tin and aluminum was important for the catalyst selectivity to lactic acid.The aim of this work was to illustrate the possibility to utilize lignocellulosic side stream, hydrolysis lignin as a raw material for activated carbon, which in turn could be used as catalyst support for metal oxide catalysts. The modification of the AC support by chemical treatments with ZnCl2 and HNO3 was also studied. The prepared lignin-based activated carbon-supported tin-, aluminum- or chromium-containing catalysts were studied in the conversion reaction of glucose to lactic acid. All of the tested carbon-supported metal oxide catalysts showed a high rate of glucose conversion (> 94 %) and were able to convert glucose to lactic acid in 20–120 min, depending on the reaction conditions. The addition of tin oxide along with aluminum oxide resulted in higher lactic acid production yields, in contrast to chromium oxide, which directed the reaction towards byproduct formation. It was noted that activated carbon support prepared by chemical activation with zinc chloride caused the deposition of zinc oxides on the support; this by itself was able to convert glucose to lactic acid and should be studied more in the future. Moreover, treatment of the AC surface with nitric acid had a positive effect on the lactic acid yield when tin and aluminum oxides were added on to the support, and was related to the higher acidity of the catalyst. The highest yield of lactic acid was produced with steam-activated and nitric acid-treated carbon support containing Sn/Al oxides at 5/2.5 wt.% (Sn/Al5/2.5@ACSN)—within 20 min, the catalyst Sn/Al5/2.5@ACSN yielded 42 % lactic acid at 180 °C without addition of excess pressure. However, the reusability experiments indicated that the catalyst was not stable in an aqueous solution, resulting in a decrease in lactic acid yield within four cycles. This was probably caused by the deposition of carbonaceous byproducts, such as humines, on the catalyst active sites and leaching of aluminum oxides as active metal. In conclusion, lignin-based activated carbon supports provided interesting results and opened a window for later studies with lactic acid production. Furthermore, improved catalyst and reaction condition design is needed for better selectivity and to obtain the recyclable catalyst without deactivation. Riikka Kupila: Conceptualization, Investigation, Formal analysis, Writing - original draft. Katja Lappalainen: Conceptualization, Supervision, Writing - review & editing. Tao Hu: Formal analysis, Writing - review & editing. Henrik Romar: Writing - review & editing. Ulla Lassi: Supervision, Writing - review & editing.The authors declare no conflict of interest.Authors R.K. and K.L. would like to thank the Green Bioraff Solutions Project (EU/Interreg/Botnia-Atlantica, 20201508) for funding this research. Hanna Prokkola is thanked for performing the elemental analysis. Mia Pirttimaa and Sari Tuikkanen are thanked for methods developed with HPLC analytics.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcata.2021.118011.The following are Supplementary data to this article:

In this study, heterogeneous biomass-based activated carbon-supported metal oxide catalysts were prepared and tested for lactic acid production from glucose in aqueous solution. Activated carbons were produced from hydrolysis lignin by chemical (ZnCl2) or steam activation and modified with a nitric acid treatment and Sn, Al, and Cr chlorides to obtain carbon-based metal oxide catalysts. The modification of the carbon support by nitric acid treatment together with Sn and Al oxides led to an increase in lactic acid yield. The highest lactic acid yield (42 %) was obtained after 20 min at 180 °C with the Sn/Al (5/2.5 wt.%) catalyst on steam-activated carbon treated by nitric acid. Reusability of the catalyst was also studied with the conclusion that the deposition of carbonaceous byproducts and leaching of Al oxides led to a decrease in catalyst selectivity to lactic acid.

While worldwide energy consumption continues to grow, unfortunately, nearly 88% of the current energy economy relies on fossil fuels [1]. It is only a matter of time before fossil fuels become either completely depleted or unprofitable to retrieve. Despite their outsized share of the energy portfolio, the era of fossil fuels is coming to an end [2]. Apart from the problem of diminishing availability, the negative externalities of fossil fuels have posed a prominent risk to the global ecosystem. At present, the world’s main source of energy is the combustion of fossil fuels. The byproducts of this combustion (e.g., CO2, NO x , SO x , and fine particles) seriously pollute the air, soil, and water [3–5]. It is urgent that we adopt a fresh mindset in order to find solutions to these problems and to devise a future with a more secure and sustainable energy supply. Renewable energy will play a vital role in the world’s energy future. However, there is a market barrier that stems from a major difference between renewable and conventional energy sources [6–9]. The amount of energy (mainly in the form of electricity) yielded by renewable energy sources can change unpredictably over a short period of time. For example, solar systems only produce energy when the sun is shining. Unfortunately, other renewable sources such as wind and tidal movements also possess adverse inconstancy [10]. This inconstancy makes the current generation of renewable energy less reliable than fossil-fuel-derived energy, because its output is highly dependent on weather conditions (i.e., clouds or wind) and time (i.e., day or night). For renewable energy to be practical on a very large scale, highly efficient electricity conversion and the high-density storage of electricity are required to enable energy-distribution technologies.Electrochemical hydrogen–water conversion (H2 + O2 ↔ H2O) is a clean and efficient execute solution for a sustainable energy system [11,12]. To be specific, renewable energy can be converted into chemical energy stored in hydrogen (H2) through water electrolysis [13]. Inversely, hydrogen molecules can be electrochemically recombined into water (H2O) in order to output electricity through fuel cells. In this energy system, hydrogen acts as the energy carrier, and the energy conversion is independent of thermal cycles [1,14,15]. Because it is based on electrochemical reactions, hydrogen–water conversion can drastically reduce the release of climate-changing gases and compounds that are harmful to the natural environment and to human health. However, for the purposes of practical application, hydrogen must first be obtained, then be stored, and finally be converted back to water to release the stored energy [16]. In order to achieve this target, efficient, low-cost water electrolysis and fuel cell technology must be efficiently integrated together. Electrochemical processes are the core of these energy conversion technologies, including the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) in the water electrolyzer, and the oxygen reduction reaction (ORR) and the hydrogen oxidation reaction (HOR) in the hydrogen–oxygen fuel cell [17,18]. The output performance of the aforementioned energy conversion technologies is significantly influenced by the efficiency of these four electrochemical reactions. Thus, the most critical problem in this sustainable energy system is how to effectively catalyze these reactions on the catalytic electrode surface in order to attain the lowest overpotential and highest current density [19–22]. Apart from the potential drop induced by electrochemical reactions, the electrical resistances and transport-related resistances affect the overall cell voltage of the water electrolysis and fuel cell. Accelerating the electron and proton transfer and the product emission by optimizing the electrode structure is another problem that requires extensive attention.This article provides a comprehensive review of recent advances toward the structural engineering of electrocatalytic catalysts for electrochemical hydrogen–water conversion. Two major issues are addressed in this review: ① the origin of energy dissipation in the electrochemical hydrogen–water conversion system; and ② the structural design of electrocatalysts for high energy-conversion efficiency, driven by the combination of fundamental science and practical technology. In the second section, after a brief introduction of the electrochemical process that occurs in hydrogen–water conversion, we present a review of the energy dissipation in the two functioning technologies—that is, water electrolysis and the fuel cell—from a practical standpoint, and use classical kinetics to analyze the key barrier of the electrochemical reactions occurring on the catalyst surface. With the aid of scaling relations among reactive intermediates, we develop a framework to understand catalytic trends, which ultimately provides rational guidance toward the development of improved catalysts for a wide range of reactions. In the third section, we summarize general strategies on designing higher-performance electrocatalysts and discuss their advantages and drawbacks. Featured examples of state-of-the-art electrocatalysts that have been achieved for each reaction by structural design are presented in this part, demonstrating the successful combination of synthetic chemistry, electrocatalytic chemistry, and computational chemistry. The last section outlines the key scientific problems in the electrochemical hydrogen–water conversion system and provides further development direction for catalyst design for a renewable and clean energy system with a high energy efficiency.As illustrated in Fig. 1 , two different functioning technologies are involved in this renewable and clean energy system: water electrolysis and the fuel cell. The two electrolyzers are mainly composed of four parts: the electrolyte (e.g., H2O), the ion-exchange membrane (e.g., a Nafion membrane), the anode electrode, and the cathode electrode [23,24]. In order to accelerate water splitting, the two electrodes are always coated with a highly active and stable catalyst layer. In the water electrolyzer, electrical energy is consumed to split water into gaseous hydrogen (H2) and oxygen (O2). Taking acid water electrolysis as an example, water is oxidized to form oxygen molecules and protons at the anode, as the electrons pass through the external circuit and the protons pass through the membrane down to the cathode. Meanwhile, the protons and electrons combine at the cathode to form hydrogen molecules.The electrochemical reaction that occurs in the fuel cell is the exact opposite of the water electrolysis process [25]. A spontaneous “cold” combustion of hydrogen and oxygen occurs in the fuel cell device, in which hydrogen acts as the fuel and oxygen acts as the oxidizer. In general, hydrogen diffuses to the anode surface by penetrating through the electrode pores. Through the catalytic action of the catalyst layer, the adsorbed hydrogen is ionized and releases an electron at the electrode. Next, the hydrogen ions that pass through the electrolyte and the electrons that pass through the external circuit all reach the cathode, recombine with oxygen molecules to form water molecules, and release electricity. The heat generated by the internal reaction and electrical resistance can be removed by applying suitable water or air cooling systems. Table 1 summarizes the half-cell reactions of water electrolysis and the fuel cell in different media. The four reactions can be grouped into two reversible reaction couples: hydrogen-involving HER and HOR with an equilibrium potential (U 0) of 0 V versus reversible hydrogen electrode (vs RHE); and oxygen-involving ORR and OER with a U 0 of 1.23 V vs RHE. From a chemical point of view, hydrogen–water conversion is composed of two redox couples: The water/oxygen couple at high potential and the water/hydrogen couple at relatively low potential [26]. In acidic media, hydrated protons transfer charges from the anode to the cathode, while hydroxide ions act as charge carriers from the cathode to the anode in alkaline electrolytes.According to the thermodynamics of hydrogen–water conversion, reversible water electrolysis and the H2–O2 fuel cell have the same electrical onset potential of 1.23 V under standard conditions. However, the actual onset potential for the two reactions is far from the standard electrical potential. Under practical operating conditions, the cell voltage of the H2–O2 fuel cell and of water electrolysis is always below 0.9 V and higher than 1.8 V, respectively, even utilizing state-of-the-art noble metals as the electrocatalysts [2,27–29]. In practice, in order to drive the electrochemical reaction process, there are a number of barriers that must be overcome, including the electrical resistance of the circuit, activation energies of the electrochemical reactions, blockage of the electrode surfaces by the product gas bubbles and water, and ionic transfer resistances across the electrolyte solution [24]. These barriers, which require a sufficient electrical energy supply, greatly reduce the energy conversion efficiency and cause the work potential to fall short of the thermodynamic potential, which is the so-called phenomenon of polarization (or overpotential, or overvoltage) [30]. Fig. 2 shows the resistances (i.e., the barriers) in a typical cell with a liquid electrolyte [31,32]. The first resistance from both ends (R a Ω and R c Ω) is the external electrical circuit resistance, which includes the internal resistance of the wiring and connections at the anode and cathode, and the resistance of the electrons across the catalyst layers, which are not always good electronic conductors [33]. R a ct and R c ct originate from the overpotential of the half-cell reactions on the surface of the anode [34]. R a D and R c D are caused by the diffusion layers close to the electrode surface when mass transport phenomena are involved or gaseous species are formed [35]. As for water electrolysis, the partial coverage of the electrode surface by generated bubbles hinders the contact between the electrode and electrolyte. Similarly, the product water acts as a blockage between the electrode and the H2 and O2 input, inducing a resistance of mass transport in an alkaline fuel cell (AFC). R ions originates from the transport of ions in the electrolyte and R sep stems from the resistance of the cell separator [36]. A similar situation is found in other types of cells, such as zero-gap and proton-exchange membrane (PEM) cells [37,38].The resistances in cell systems can be classified into three categories: activation resistances (losses due to electrochemical reactions), ohmic resistances (losses due to ionic and electronic conduction), and concentration resistances (losses due to mass transport). Each of the three major losses contributes to the characteristic shape of the current–voltage (i–V) curve of the electrochemical cell [39,40]. Fig. 3 shows a typical i–V curve for water electrolysis and for a fuel cell. In the case of water electrolysis (Fig. 3(a)), the current starts to flow across the cell above the thermodynamic electrolysis voltage of 1.23 V. Additional voltage is required to overcome the resistances discussed above. At low current densities, the voltage drops caused by ohmic resistances are small, and the reaction activation overvoltage accounts for the dominant part of the voltage drop. The logarithmic shape of the polarization curve (the Tafel area) is attributed to the charge transfer phenomena at the anode and cathode. As the overvoltage increases further, the reaction activation barrier decreases, and the shape of the polarization curve becomes linear. This linear shape indicates that the ohmic resistance is now the key kinetic parameter of the cell. In the case of the fuel cell (Fig. 3(b)), the activation resistances mostly affect the initial part of the curve, the ohmic resistances are mainly apparent in the middle section of the curve, and the concentration resistances are significant in the tail. Although the reactions that occur in the two functioning technologies are reversible, the shapes of the i–V curves are not the same: the i–V curve for water electrolysis generally obeys the Butler–Volmer model even at very high overpotentials, while the i–V curve for a fuel cell tends to show a constant value at high overpotentials due to the limitation in the mass transfer rate.To improve the energy efficiency of the two electrochemical cells and thus improve the performance of the energy system, an understanding of these resistances must be grasped in order to minimize them. Ohmic losses are caused by the electrode material’s resistance to the electron flow and the electrolyte’s resistance to the ion flow; these can be reduced by utilizing highly conductive materials as the wiring and electrode substrate, and by diminishing the distance between the two electrodes [41–43]. The concentration losses, which are attributed to mass transport, can be relieved by increasing the pressure of the gaseous reactants or the concentration of the liquid electrolyte [44,45]. The two kinds of voltage drops mainly depend on the cell design and operation conditions. In addition to the above two resistances, the majority (> 60%) of the voltage drop in an electrochemical cell is induced by the Gibbs free-energy change for the endergonic transformation of the half-cell reactions [39]. Depending on the direction of the reactions, the activation polarizations greatly increase or decrease the anode voltage where the oxidation reaction takes place, and decrease or increase the cathode voltage where the reduction reaction occurs.In electrochemistry, the Butler–Volmer relationship is used as the primary departure point to relate the overvoltage η across a metal–electrolyte interface to the current density j (in A·cm−2) across this interface [46]: (1) j = j 0 e α n F η / ( R T ) - e - ( 1 - α ) n F η / ( R T ) where η is the overvoltage—that is, the difference between the actual voltage across the interface and the equilibrium voltage; j 0 is the exchange current density in A·cm−2; α is the coefficient of the charge transfer; n is the number of electrons transferred in the electrochemical reaction; F ≈ 96 485 C·mol−1 is the Faraday constant; R is the constant of a perfect gas (0.082 J·(K·mol)–1); and T is the absolute temperature in K. The Butler–Volmer equation basically reveals that the current produced by an electrochemical reaction increases exponentially with the activation overvoltage and exchange current density. In fact, improving the reaction energy efficiency focuses on increasing j 0 , which represents the “rate of exchange” between the reactant and product at equilibrium. Taking the forward reaction for simplicity and including the concentration effects, j 0 is defined as follows: (2) j 0 = n F c f e - Δ G a c t / ( R T ) where c is the reactant surface concentration, f is the decay rate to products, and ΔG act is the activation energy barrier for the forward reaction. Eq. (2) clearly shows that decreasing the size of the activation energy barrier (ΔG act) will increase j 0 under a given environmental condition. In the actual reaction, only species in the activated state can undergo the transition from reactant to product. In fact, the activation energy of the reactions is strongly influenced by the electrode material [47]. A catalytic electrode is a site for species activation and transition. Using a highly catalytic electrode can significantly lower the activation barrier for the reaction, and therefore provides a way to dramatically increase j 0 . To reduce the activation energy of the electrode reactions, continuing research efforts are focusing on the design of efficient catalytic electrode materials based on an understanding of the relationship between the activation energies, electrode materials, and surface configurations.In terms of the reaction mechanism, the activation energy barrier (ΔG act) can be quantified by the Gibbs free-energy change (ΔG max) for the rate-determining step (RDS) at the equilibrium potential, and its theoretical value on different catalytic materials can be calculated by means of density functional theory (DFT) calculation. In this way, the relationship between the activation energy and the electrode material is built, as a volcano-shaped plot is obtained by plotting j 0 versus ΔG max. The most common shape of the volcano plot is the HER rate description based on the Langmuir type of adsorption with the maximum located near the position where the hydrogen adsorption free energy (ΔG H*) is zero [48]. In the HER, the reaction species is first adsorbed on the catalyst surface to form the reaction intermediate (M–Hads). After the aforementioned Volmer step, hydrogen molecules can be formed by the coupling of an electron and a proton in the electrolyte through a Heyrovsky step, or their direct combination via a Tafel step [30,49,50]. As a result, the ΔG H* is the overall decisive rate for HER [51,52]. In recent years, the DFT-calculated ΔG H* has been widely used as the activity descriptor [53–55], for many traditional metals, metal composites/metal alloys, and nonmetallic materials. As shown in Fig. 4 (a), different metals show significant differences in HER exchange current density, and the highly active metals (e.g., Pt) located near the top of the volcano plot possess optimal ΔG H* [56]. If the catalytic material has a weak adsorption force on hydrogen, the hydrogen atom can barely be absorbed on the surface of the material, and the overall reaction rate is determined by the adsorption step of hydrogen (Volmer step). On the other hand, a too-strong adsorption of hydrogen atoms onto catalytic materials results in difficulty breaking the M–Hads bond to form H2, and the RDS is the desorption step (Heyrovsky/Tafel). As the reverse process of HER, the RDS of HOR is the dissociative adsorption of H2 on the catalyst surface, which involves electron transfer from the surface to the σ* antibonding orbital of the H2 molecule [57]. Consequently, the interaction of M–Hads also plays a dominant role in the kinetics of HOR, and the activity of HOR follows the same trend as HER on noble metal surfaces due to the high reversibility of these two reactions (Fig. 4(b)) [58–61].In addition to the hydrogen-involved reactions, the relationship between j 0 and ΔG max can be applied to the oxygen-related reactions that occur in hydrogen–water conversion. As shown in Figs. 4(c) and (d), the shapes of volcano plots for these reactions are quite similar, except for the reaction intermediates determining the reaction rate. In general, the ORR includes either a four-electron pathway to reduce oxygen to water, which is attractive for fuel cells, or a two-electron pathway, which is desirable for producing hydrogen peroxide [62]. In fact, a direct four-electron mechanism can either be a dissociative or associative process, depending on the oxygen dissociation barrier on the catalyst surface [63]. As a result, the oxygen adsorption strength (ΔG O*) is associated with the ORR activity to construct a volcano plot (Fig. 4(c)) [63,64]. For metals that bind oxygen too strongly, the reaction rate is limited by the removal of O* or OH* species. For metals that bind oxygen too weakly, the reaction rate is limited by splitting of the O–O bond in O2 (dissociative mechanism) or, more likely, by the transfer of electrons and protons to the adsorbed O2 (associative mechanism), depending on the applied potential [63]. As indicated by the volcano plot in Fig. 4(c), there seems to be some room for improvement, as even platinum (Pt) is not at the absolute peak. Metals with a somewhat lower oxygen-binding energy than Pt should have a higher ORR activity. Based on the above thermodynamic volcano plot, Viswanathan et al. [65] and Hansen et al. [66] developed a microkinetic model for ORR given that the OH binding energy is varying. They found a kinetic activity volcano that is in close agreement with the thermodynamic activity volcano, and identified an activity optimum at a 0.1 eV weaker O* binding than Pt(111) for the reduction of O2 to H2O through a four-electron pathway.The OER volcano plot has a long history starting in 1984, when Trasatti used the transition enthalpy from the lower to higher oxidation state of metal in metal oxides as a descriptor for the OER electrocatalytic activity of oxide electrodes [67]. That pioneering work viewed the OER process as a transition between two different configurations of the surface coordination complex. Accordingly, all metal oxides that are difficult or easy to oxidize are not very active for the OER. Difficult oxidization means that the intermediates are weakly adsorbed; therefore, water dissociation is the RDS. On the other hand, easy oxidization indicates that the intermediates are strongly adsorbed, and the removal of the O* or OH* species is the RDS. In this case, the OER reactivity has been related to the oxygen adsorption free energy ΔG O* [68,69], as in the case of ORR. However, the single descriptor of ΔG O* for OER activity is incomplete, as the four-electron OER involves multiple intermediates (OOH*, OH*, and O*), the binding energies of which are strongly correlated and can hardly be decoupled [63,70]. A linear scaling relation exists between the binding energies of the different surface intermediates [70]; that is, if the energy associated with one reaction step is changed, the energies of the others also change. Thus, Man et al. [70] took the difference between the energy states of two subsequent intermediates (ΔG O*−ΔG OH*) as a descriptor for the catalytic activity of several compounds, including rutile, perovskite, spinel, rock salt, and bixbyite oxides (Fig. 4(d)), whose activity obeys the volcano plot quite well. In fact, the binding energies of OH* and OOH* (whether in OER or ORR) are related to each other by a constant energy value of approximately 3.2 eV in broad classes of metal oxide materials, regardless of the binding site [70,71]. As a result of this non-ideal scaling between OOH* and OH*, a real catalyst generally shows a minimum theoretical overpotential of 0.3–0.4 V [63,72,73], even for materials at the top of the OER and ORR volcano plots, including the extensively studied RuO2 for OER [70] and Pt-based catalysts for ORR [74].It is noticeable that the volcano plot appropriately demonstrates the Sabatier principle [75]; that is, an ideal catalyst should bind the reaction intermediates neither too weakly nor too strongly. In other words, optimal catalytic activity can be achieved using a catalyst surface with appropriate binding energies for reactive intermediates. To be specific, the best approximation to an ideal HER/HOR catalyst would be a material that is capable of minimizing the absolute value of ΔG H*, and the ideal ORR and OER catalysts would be able to optimize the ΔG O* and ΔG O*−ΔG OH*, respectively. In fact, aside from decreasing the activation barrier, there is another significant way to increase j 0 , which is not apparent from Eq. (2): that is, to increase the number of possible reaction sites per unit area [76–79]. j 0 represents the current density, or the reaction current per unit area, and the area for current density is generally based on the projected geometric area of an electrode. The true electrode surface area of an electrode with an extremely rough surface can be orders of magnitude greater than the geometric electrode area, and can thus provide many more reaction sites. Therefore, the effective j 0 of a rough electrode surface will be greater than that of a smooth electrode surface. Another simple way to increase the density of active sites is to enlarge the amount of catalyst on a given electrode. However, an excessive amount of catalyst will hinder the charge and proton transfer on the electrode surface. As a result, the activity of the electrode does not increase linearly with the amount of catalyst.In conclusion, there are two general methods for increasing the activity (or rate of reaction) of an electrocatalyst system: ① improving the intrinsic activity of each active site; and ② increasing the density of active sites on a given electrode. Both methods have pros and cons. The difference in intrinsic activity between different catalysts may be more than ten orders of magnitude, while the difference in activity caused by catalyst loading will be only 1–3 orders of magnitude. Improving the intrinsic activity of each active site is the most fundamental and effective way to achieve high activity, and its realization must be based on a deep understanding of the reaction mechanism and material properties. Increasing the number of active sites is an easier strategy, but the activity growth is limited. At the same time, activity promotion by increasing catalyst loading is obtained through the sacrifice of increasing the electrode cost and the charge and proton transfer blockage. In practice, the two methods are not mutually exclusive and can ideally be implemented simultaneously, thus greatly enhancing the activity of catalyst.It is well known that the current density of a catalyst increases as the density of the active sites increases. Exposing more active sites is important in achieving high catalytic performance. Nanoarchitecture has been considered to be the most effective strategy, as it allows the density of active sites to be directly enriched and utilized, and thus efficiently optimizes the electrocatalytic activity [80–84]. The relation between the actual active surface area and the overall performance of an electrocatalyst was first recognized in transition-metal alloy systems. As early as 1982, Brown et al. [85] found that an alloy surface is generally rougher than that of a single metal, and can provide more active sites for a catalytic reaction. With the aid of nanostructuring and selective etching of molybdenum (Mo) in Ni–Mo alloys [85–87], the surface area of Ni–Mo alloys greatly increased, resulting in an obvious improvement in catalytic reactivity. With the rapid development of synthesis techniques, a series of electrocatalytic nanomaterials with different morphologies have been achieved in the past decade, including nanocages, nanofibers, nanoflowers, nano-foam, nano-nets, nano-needles, nano-rings, nano-shells, and nanowires [77,88–91]. Faber et al. [92] reported metallic cobalt disulfide (CoS2) as a highly active catalyst for HER, and demonstrated the crucial role of geometric structure in determining its overall catalytic performance. Compared with the common morphologies of nanoparticles and nanofilm, an increase in active surface area drastically improves the HER catalytic performance of microstructured and nanostructured electrodes (Fig. 5 ), endowing the CoS2 nanowire electrodes with overpotentials as low as 145 mV for driving a current density of −10 mA·cm−2. In addition, nanostructuring possesses a dual function in terms of both operational stability and reaction rate through the facilitation of mass transport and the removal of generated gas bubbles or water from the catalyst surface. Our group synthesized Mo2C/C with a two-dimensional (2D) lamellar structure via a controllable synthesis using a self-assembly and pre-shaping strategy [93]. The highly dispersed Mo2C nanoparticles and the 2D lamellar structure effectively boosted the mass and charge transfer across the Mo2C active sites, facilitating the electrochemical HER process. Moreover, our group further synthesized a series of three-dimensional (3D) nanostructured catalytic materials, including the NiCo2(SOH) x nanoflower [94], the coral-like FeNi(OH) x [95], the Ni–VC nanoboscage [96], Ni–Mo2C nanowire [97], and the Ni(OH)2@Ni2P nanopillar [98]. All these materials possessed highly active surfaces, fast electron transfer, and gas escape channels, which are beneficial for catalyzing water electrolysis.To be specific, a rapid loss resulting from water flooding may take place in ORR catalysts, apart from deactivation during a long operation time [99,100]. Water flooding will interrupt the O2 supply to active sites as the porous channels are obstructed by the accumulation of water, resulting in the termination of ORR in the flooded region [101,102]. In order to quantify the mass transfer and anti-flooding performance with the pore characteristics of electrocatalysts in a fuel cell, Wang et al. [103] designed a special “rattle-drum”-like work electrode for ORR catalysts. Benefiting from a bigger pore volume and regular pore arrangement, the dual-porosity Pt/C catalyst exhibited four times the quantified mass transfer and anti-flooding efficiency of a commercial catalyst. In fact, different types of pores have special functions in the ORR process. Mesopores and macropores may be significant in mass transport during the ORR process [104–106], while micropores are beneficial for hosting most catalytic sites [107,108]. With the purpose of constructing hierarchically porous structures, the sacrificial template method has been widely adopted, using silica colloid [105,109–111], ordered mesoporous silica [106,112–114], polystyrene microspheres [115], and some other oxides [116,117] as templates. For example, a colloidal silica template was used by Liang et al. [118] for the synthesis of N-doped carbon catalyst with a high specific surface area of 1280 m2·g−1, a hierarchically porous structure with meso/micropore distribution. However, the subsequent removal of the template can be time consuming and usually requires the usage of a strong acid or alkaline solution, which is dangerous to researchers and harmful to the environment. To avoid these disadvantages, our group developed a morphology-controlled approach using NaCl as the template, in which the template can be removed using hot water and then recycled [119–122]. A nanostructured polyaniline (PANI) with a special structure was encapsulated in the NaCl crystal via salt recrystallization, and then accurately converted into a carbon nanomaterial under high temperature (Fig. 6 ). Moreover, a mass of pores was created in the carbon nanomaterial by gasification in a closed nanoreactor. Due to the multiple types of pores and high utilization of active sites, the 3D Fe/N–C catalysts exhibited excellent catalytic performance toward the ORR.Facet engineering is another widely studied method to modulate the catalytic performance of materials for a given reaction. The reactivity of catalytic materials is highly related to their exposed facets because the adsorption strength of the intermediate species of the catalytic reaction varies greatly on different surface facets of the catalyst. Facets are always denoted by Miller indices. The exposed facet(s) of a nanomaterial strongly correlate to the shape of the nanoparticle [123,124]. In general, faceted nanomaterials can be categorized into low-index and high-index faceted types [125]. Low-index facets are those for which the sum of the three components of the Miller indices (hkl) is small, whereas a high-index facet contains at least one Miller index greater than unity.The structural sensitivity of the reactivity of low-index faceted nanomaterials has been demonstrated with regard to single-crystal Pt for HER. The surface morphology of Pt(hkl) with well-defined surfaces was confirmed by scanning tunneling microscopy (STM, Fig. 7 ), and the degree of activity was observed to follow the order (110) > (100) > (111) in alkaline solution [126]. Importantly, the activity in alkaline and acid electrolyte was found to differ substantially [127–129]. This pH effect involved structure–function relationships in the HER, and was further studied by Strmcnik et al. [130] with a focus on both Pt(111) and Pt(111) modified by Pt islands. Compared with the HER activity of the pristine Pt(111) surface, the HER on the Pt islands/Pt(111) electrode was found to be 5–6 times more active in alkaline solution, but only 1.5 times more active in acid electrolyte. The effect of pH on HER activity was shown to be due to the special ability of edge-step sites to dissociate water [131–135]. In addition, the activity order of Pt(hkl) is in line with the density of low-coordinated Pt atoms due to the accelerated water dissociation step on their metal surfaces [136–139].A similar phenomenon was observed in the case of ORR activity on single-crystal Pt surfaces following a descending order in nonadsorbing HClO4 electrolytes [140]. However, when the electrolyte was replaced by H2SO4, Pt(100) was found to be more active than Pt(111) [141–143]. This activity difference was attributed to the special adsorption behaviors of the bisulfate anion on Pt(111). The bisulfate anions can be adsorbed on the Pt(111) surface more strongly than on Pt(100), resulting in an impeditive ORR process. Systematic investigations confirmed that the different properties of the respective facets have a significant effect on their catalytic performance. Following this work, many studies focused on developing a facet-control method to construct more active facets on a catalytic surface, from ideal single-crystal metals to more practical nanomaterials [144–150]. Narayanan and El-Sayed [144] were the first to demonstrate facet-controlled synthesis in the case of Pt nanocrystals, including (100)-terminated nanocubes, (111)-bounded nanotetrahedra, and nanospheres with both (111) and (100) facets. A high-temperature organic phase method was reported by Wang et al. [145] for the synthesis of monodispersed (100)-terminated Pt nanocubes for ORR. The facet-controlled Pt nanocubes showed a specific ORR activity that was more than double the activity of a commercial Pt catalyst in acid electrolyte.Because of the higher density of low-coordinated atoms, steps, edges, and kinks, metals and compounds with high-index facets generally possess greatly improved reactivity in comparison with typical low-index materials [151]. However, such high-index facets are thermodynamically instable due to their higher surface energy [152,153]. Accordingly, the synthesis of high-index faceted nanomaterials has become a herculean task. In the last few years, a wide variety of synthetic protocols have been exploited for the synthesis of metallic nanomaterials with high-index facets for the purpose of improved catalytic reactivity. Yu et al. [154] developed a simple reduction route in aqueous solution to prepare Pt concave nanocubes (c-NCs) enclosed by high-index facets of (510), (720), and (830). Pt c-NCs have also been fabricated by using glycine to manipulate the reduction kinetics of H2PtCL6 [155]. Employing electrochemical means, Tian et al. [156] synthesized tetrahexahedral (THH) Pt nanocrystals with facets including (730), (210), and (520). In addition to these single-metal materials, multi-metallic high-index faceted nanocrystals have been developed [157–160]. As shown in Fig. 8 , Luo et al. [159] reported a new class of Pt3Fe zigzag-like nanowires (z-NWs) with stable high-index facets and a nanosegregated Pt-skin structure. These unique structural features endowed the Pt-skin Pt3Fe z-NWs with a mass and specific ORR activity of 2.11 A·mg−1 and 4.34 mA·cm−2, respectively, at 0.9 V vs RHE.With the purpose of reducing catalyst cost, the design and preparation of low-cost metal nanomaterials with different exposed reactive facets have been a recent trend in facet engineering [161–167]. Su et al. [168] studied the growth mechanism of the NiO crystal and found that the surface energy of the NiO facets followed the order of (100) < (113) < (101) ≈ (110). Han et al. [161] provided a template-free hydrothermal method for the controllable fabrication of a surface-tailored Co3O4 nanocube (NC), nanotruncated octahedron (NTO), and nanopolyhedron (NP), with facets of (001), (112), (001) and (111). The different crystal planes endowed the Co3O4 nanocrystal with different exposed surface atomic configurations of the Co2+ and Co3+ active sites. The unusual (112) plane-enclosed Co3O4 nanoparticle on reduced graphene oxide (rGO) with abundant Co3+ sites exhibited superior activity for both OER and ORR. In addition to these metal oxides, many other metal compounds with special facets have been reported. Feng et al. [163] further synthesized Ni3S2 nanosheet arrays with stable (2 1 ¯ 0) facets, and demonstrated them to be efficient and ultra-stable electrocatalysts for HER and OER. Wang et al. [162] obtained flower-like nickel phosphide with different crystalline structures (Ni5P4 and Ni2P) and ascribed the excellent HER activity to the hierarchical structure with high-energy (001) facets.Apart from tuning the exposed facets of the nanocrystal, modulating the atomic scale arrangement (i.e., the transformation of the crystalline phase) can affect the intrinsic activity of a catalyst, due to the fundamental changes in its physical and chemical properties. Transition-metal disulfides with several unique polymorphs are typical cases that have been widely studied. Among these polymorphs, the metastable 1T phase has recently aroused great research interest owing to its metallic behavior, which is beneficial to electrocatalytic processes [169–174]. Lukowski et al. [169] synthesized metallic 1T-MoS2 nanosheets from semiconducting 2H-MoS2 via a lithium intercalation method. The resulting 1T-MoS2 exhibited a dramatic improvement in electrocatalytic HER performance compared with the corresponding 2H polymorphs (Fig. 9 ). Similarly, metallic 1T tungsten disulfide (1T-WS2) was further synthesized by Lukowski et al. [173] via a simpler microwave-assisted intercalation. The polymorph engineering endowed the 1T-WS2 with faster electrical conductivity and more intensive active sites, boosting its HER activity.Not only can unique properties be induced by atomic arrangement, but the active sites of 1T-catalysts may also differ from those of the traditional 2H-structured phase. In this context, Voiry et al. [175] obtained highly conductive 1T-MoS2 nanosheets with excellent HER activity after removing excess negative charges from the surface of chemically exfoliated MoS2 nanosheets. Interestingly, after partial oxidation of the 1T- and 2H-MoS2, a sharp contrast in HER activity changes was observed. Although there was almost no shift in the HER activity of 1T-MoS2 after edge oxidation, the activity of 2H-MoS2 seriously decreased. It is well known that the edges of the usual 2H-MoS2 crystal are the main active sites for HER. The significant difference in HER activity between the partially oxidized 1T- and 2H-MoS2 nanosheets revealed that the main active sites of 1T-MoS2 for driving HER catalysis are not the edges of the nanosheets, but the basal planes of the nanosheets.The catalytic performance of metal oxides may also vary with their crystal phases. Our group found that reversing spinel crystalline structure has a great influence on the ORR catalytic activity of spinel (Fig. 10 ) [176,177]. By adjusting the iron (Fe) content, the spinel structure of a Co–Fe-based crystal can be changed from its normal structure to the inverse structure and then back again [178]. The electrochemical results revealed that the inverse spinel {Co}[FeCo]O4/NG (nitrogen doped graphene) had the best ORR activity, outperforming commercial Pt/C. DFT results further disclosed that the higher ORR activity of the inverse-structured {Co}[FeCo]O4 could be ascribed to the modulated oxygen adsorption energy and elongated adsorbed oxygen bond induced by the dissimilarity effect of Fe and Co atoms at the octahedral site. The effect of crystal phase on the reaction pathway of ORR has also been studied. Karunagaran et al. [179] synthesized four kinds of iron oxide nanoparticles with different phases incorporated inside 3D rGO aerogels and determined their electrochemical, catalytic, and electron transfer properties for ORR. The results showed that ORR was catalyzed by all four catalysts via a two-electron pathway under higher potentials (0.70 V). On the other hand, when the potentials decrease to 0.20 V, rGO composites containing magnetite, maghemite, and goethite proceeded via four-electron transfer kinetics, whereas the hematite-containing composite went through two-electron transfer kinetics.Configuration distortion induced by the Jahn–Teller effect for transition-metal compounds has also been studied in relation to the electrocatalytic performances of such compounds [180]. Recently, Liu et al. [181] observed an obvious structural distortion in Co3S4 atomically thin nanosheets (CSATNs) via the ultrasound exfoliation treatment of an intermediate Co3S4/TETA hybrid precursor. The structural distortion of CSATNs generates an electronic configuration change. Compared with bulk samples, the shift from spectral to lower magnetic fields (Figs. 11 (a) and (b)) implies that the spin state of Co3+ in the octahedral sites (t2g 4eg 2) of CSATNs adjusts from low spin to high spin. High-angle annular dark field (HAADF) images showed that the octahedral coordinated cations were solely exposed in the planes, which further revealed the existence of Jahn–Teller elongation (Figs. 11(c)–(f)). Due to the synergistic adjustment in the atom and electron configuration, CSATNs possess significantly enhanced OER performance in comparison with bulk samples. In fact, the Jahn–Teller effect is attributed to the uneven electron distribution of the central ions in degenerate d orbitals (t2g or eg). Thus, the filling state of electrons in the eg orbital may have a significant role in the catalytic properties of the transition-metal compounds. A volcano relationship between the intrinsic ORR activity and the filling states of the eg orbital in the B ions of perovskite-based oxides (ABO3) was discovered by Suntivich et al. (Fig. 11(g)) [182]. The perovskite-based oxides with only one electron filling in the eg orbital (defined as eg ≈ 1) were demonstrated to possess the highest ORR activity, as O2 can adsorb on the B sites end-on with an optimal binding energy. The eg occupancy theory can be further extended to spinel oxides, although the ORR active sites of spinel are not tetrahedral sites but octahedral sites (Fig. 11(h)) [183].Amorphization to modulate the atomic scale arrangement, and thus increase the catalytic performance, is another research hotspot [184–189]. Short-range atomic arrangements of amorphous phases are beneficial for increasing the density of active sites [190–194]. As early as 1995, Weber et al. [190] investigated the structural units of the amorphous compound MoS3 and found that all molybdenum is present in the Mo4+ oxidation state, while sulfur atoms occur in two different types of coordination: S2– and S2 2–. Merki et al. [191] and Benck et al. [192] then confirmed that the amorphous MoS2 is more active in catalyzing HER. Structural measurements demonstrated that an amorphous MoS x film is extremely rough in surface and sulfur-rich in composition, resulting in a large active area and intensive active sites for HER catalysis. Benck et al. [192] further revealed that the increase in the HER activity of amorphous molybdenum sulfide is contributed to by the large number of active sites caused by the amorphous structure and rough, nanostructured morphology, as the activity scales with the electrochemically active surface area. Meanwhile, Li et al. [195] and Li et al. [196] systematically studied the origin of the catalytic activity of the amorphous MoS2 in terms of the composition and crystallinity. Interestingly, the experimental results revealed that the crystallinity is crucial for determining the catalytic performance, whereas the composition is not particularly significant.In addition to the HER catalysis, Smith et al. [185] demonstrated that amorphous materials are more active than the comparable crystalline materials for OER catalysis, based on a study of the mixed-metal oxides of iron, nickel (Ni), and cobalt (Co). Due to the amorphous structure, the distribution of the metals in the amorphous films is homogeneous and their compositions can be accurately controlled. Modulated a-Fe100- y - z Co y Ni z O x with an optimal element content exhibited excellent catalytic property that was even comparable to that of commercial noble metal oxide catalysts. Thanks to the controllable composition of amorphous materials, the effect of metal composition on electrocatalytic performance can be further studied along with the effect of amorphization. Smith et al. [184] prepared 21 complex metal oxide films for electrocatalytic water oxidation, and demonstrated the excellent stoichiometric concentrations of Fe, Co, and Ni in each sample. Structural characterization and electrochemical measurement confirmed that iron content is important for lowering the Tafel slope, and that cobalt or nickel are beneficial in reducing the overpotential (Fig. 12 ). For scale-up production, Kuai et al. [186] proposed an aerosol-spray-assisted method by which amorphous mixed-metal oxides can be sustainably obtained, which is very suitable for industrial applications. The obtained Fe6Ni10O x exhibited a low overpotential of 0.286 V for driving 10 mA·cm−2 and a small Tafel slope of 48 mV·decade−1 for the electrochemical OER, exceeding the best catalytic performance of all investigated Fe–Ni–O x series.Although the active sites of amorphous catalysts can be greatly enhanced by amorphous engineering, the electrical conductivity of the amorphous materials will be decreased due to short-range disorder in the crystal structure. Coupling these low-conductive materials with highly conductive materials is an effective route to guarantee the excellent electrocatalytic performance of amorphous catalysts. For example, Lee et al. [197] synthesized amorphous MnO x nanowires supported by Ketjenblack (KB) carbon as highly efficient ORR electrodes. The low-cost and highly conductive KB acts as a supporting matrix for the catalyst, greatly accelerating the electron transfer during the electrocatalytic processes. Many other amorphous/conductive composite materials, such as amorphous MoS x /carbon composite catalyst [198], amorphous MoS x /polypyrrole copolymer film (PPy/MoS x ) [199], and amorphous MoS x /N-doped CNT (NCNT) forest hybrid catalyst [200], have also been reported. The highly conductive skeletons in these composite materials can overcome the barriers induced by the low electrical conductivity of amorphous catalysts, leading to a remarkable increase in catalytic activity (Fig. 13 ). Porous metal nanostructures, such as Ni foam [201] and nanoporous gold [202], are also used as conductive substrates to support an amorphous MoS x catalyst, of which the HER activity can be significantly enhanced.Defects exist widely in nanomaterials. It has been realized that the surface of catalysts with defects always exhibit higher reactivity than the defect-free sites [203–206]. Accordingly, defect engineering has gradually developed as an effective method to tune the electronic and surface properties of nanomaterials [207–209]. Cheng et al. [210] synthesized tetragonal or cubic M x Mn3– x O4 spinels by reducing the amorphous MnO2 in aqueous M2+ solution under ambient conditions. Due to its highly active area and abundant defects, nanocrystalline Co x Mn3– x O4 is endowed with considerable catalytic activity for both ORR and OER. Similarly, Ma et al. [211] synthesized oxygen vacancy (OV) defect-rich mesoporous MnCo2O4 materials, and found that their stability and methanol tolerance ability even exceeded those of a Pt/C catalyst. In order to obtain insight into the effect of defects in catalytic performance, our group performed a DFT + U calculation of OV concentration on the electronic structure of β-MnO2 catalysts and their catalytic performance for ORR [212]. As shown in Fig. 14 , a moderate concentration of bulk OVs will greatly increase the electric conductivity of MnO2, while excessive OVs will hinder the ORR process. Such a curvilinear relationship between the electronic structure and OV concentration suggests that the conductivity and ORR catalytic activity of β-MnO2 can be modulated by the OV concentration. Defect engineering can also be applied to increase the density of active sites of nanomaterials for electrocatalysis [203,213,214]. Xie et al. [203] designed a reaction with a high concentration of precursors and different amounts of thiourea, thus realizing controllable defect modulation in as-formed MoS2 ultrathin nanosheets. Due to the defect-rich structure, many tiny cracks formed on the basal surfaces, resulting in 13 times more active sites for the defect-rich MoS2 ultrathin nanosheets than for the defect-free MoS2.Similar to element vacancy in metal compounds, intrinsic defects in carbon-based electrocatalysts are universal but have been ignored for a long time [209]. Defects easily form after heteroatom doping, and act as the active sites favoring electrocatalysis [215,216]. However, the electrocatalytic reactivity of carbon-based materials has mainly been ascribed to the induced changes of heteroatoms doping. As time goes on, some research has found that the catalytic activity of carbon electrocatalysts with intrinsic defects is even better than that of heteroatoms-doped carbon materials [217,218]. For example, Jiang et al. [219] found that defective carbon nanocages (CNC) possess a high ORR activity that exceeds that of B-doped carbon nanotubes. In this case, defect-rich CNC were successfully synthesized with many typical defect locations, but without any dopants (Fig. 15 (a)). The electrochemical results indicated that the resultant CNC material with the highest defect density showed the best electrochemical ORR activity (Fig. 15(c)). The DFT results further indicated that the high ORR activities of these defect materials could be attributed to the pentagon and zigzag edge defects (Fig. 15(d)). Zhao et al. [220] used first principles calculations to predict that a type of 585 defect on graphene would be even more active than the N-doped sites for ORR, and obtained strong support for this theoretical prediction through experimental investigations. With the defect mechanism in mind, Zhao et al. [221] prepared a porous carbon (PC) material lacking any elemental doping by carbonizing Zn-MOF at 950 °C. With the benefit of the removal of zinc (Zn) atoms, defects could be formed on the PC catalyst, endowing the PC catalyst not only with excellent ORR activity, but also with a stability comparable to that of a commercial Pt/C catalyst. Furthermore, with the exception of the ORR process, the individual electrocatalytic activities for the other three electrochemical reactions in the energy conversion from water to water—that is, HOR, OER, and HER—were demonstrated to be particularly sensitive to the types of defects derived by the removal of heteroatoms from graphene [222].Atomic doping is the most widely used strategy for modulating the properties of catalytic materials. By reasonably introducing one or more metallic or nonmetallic elements into the lattice of the material, the electron structure of the original material can be adjusted, thus effectively improving the catalytic performance of the material [223–232]. Taking MoS2 as an example, many metallic elements such as Ni, Co, Fe, vanadium (V), lithium (Li), and copper (Cu) have been reported to be successfully doped into its crystal structure, positively affecting the physical and chemical properties [171,233–237]. Among these doped metallic elements, Ni and Co tend to locate around the S in MoS2, which will decrease the hydrogen adsorption energy at the S edge and increase the density of the active sites in MoS2 [233–235]. Unlike Ni- and Co-doped MoS2, V doping cannot increase the number of active sites, but will enhance the conductivity of MoS2 [236]. Interestingly, our group explored the influence of the Ni-doping of molybdenum carbide on its surface electronic structure and its relationship with HER performance by combining experimental and theoretical evidence [97]. As shown in Figs. 16 (a)–(d), one-dimensional (1D) NiMo2C nanowire arrays were directly constructed onto conductive 3D Ni foam (NiMo2C/NF) via a facile and controllable strategy combined with hydrothermal and post-carburization treatment. The binder-free integrated NiMo2C/NF electrode showed superb HER catalytic activity in comparison with Mo2C and Ni catalysts (Fig. 16(e)). The DFT calculations clearly demonstrated that the incorporation of Ni into the Mo2C lattice brought about changes in the charge distribution on the catalyst, which resulted in a synergistic effect of Ni and Mo2C that decreased the hydrogen binding energy (Figs. 16(f) and (g)).Apart from metallic elements, research on doping with nonmetallic elements is also very active. Xie et al. [238] successfully synthesized oxygen-doped MoS2 ultrathin nanosheets, on which the synergistic modulations of both the active sites and the conductivity could be rationally realized. According to the DFT calculations, the smaller differential binding free energy of the oxygen-incorporated MoS2 revealed a lower energy barrier for driving the HER process. Our group further proposed a partial phosphorization of metal oxide precursors to construct oxygen-incorporated NiMoP2 with enhanced HER activity [239]. As illustrated in Figs. 16(h)–(i), the H adsorption energy on the NiMoP2 surface was optimized by the oxygen incorporation, as the ΔG H* of the O–NiMoP2 is much closer to zero than its undoped equivalent. In addition, the Ni and Mo in O–NiMoP2 possessed more positive charge, which was beneficial for adsorbing and activating water molecules, greatly accelerating the water dissociation in the alkaline HER.Aside from these metal compounds, carbon-based materials have been intensively used as doping objects, greatly enlarging the scope of catalyst research [240–245]. Early in 2013, our group reported a phosphorus (P)-doped graphene with an ORR catalytic performance comparable to that of commercial Pt/C [246]. Furthermore, N, P dual-doped graphene/carbon materials were prepared as electrocatalysts for both ORR and OER, and their catalytic activities exceeded those of the benchmark Pt/C catalyst [247]. In order to reveal the underlying reasons for the high activity of the heteroatom-doped carbon, our group conducted a comprehensive DFT calculation on graphene doped by a series of different heteroatoms for ORR [248]. The DFT results indicated that there was a triple effect of the carbon sites—namely, the charge, spin density, and ligand effect—determining the intrinsic catalytic activity of the doped carbon catalysts and their ORR mechanism (Fig. 17 ). When the carbon materials are doped by a single heteroatom, the carbon sites around the doped atom can only be activated by the triple effect separately. This causes the ORR to proceed via the associative mechanism, and there is a limitation with an intrinsic overpotential of 0.44 V. However, when carbon materials are doped by metal or dual-heteroatoms, the ORR follows the dissociative mechanism, as double carbon sites can be activated by the triple effect. Thus, the activity limitation of the associative mechanism will no longer be in effect, leading to enhanced ORR activity. Our group also synthesized graphene co-doped with metallic and nonmetallic elements, and revealed the roles of nitrogen configuration in N-doped graphene as well as that of the trace atomic Ni in HER [249]. We found that quaternary nitrogen (N) is the most active site of the three N types in HER, whereas when doping with trace atomic cobalt, the planar (pyridine and pyrrolic) N becomes the most active. In contrast, when trace atomic Co was replaced by Ni, the planar (pyridine and pyrrolic) N exhibited depressed HER activity.Hybrid nanomaterials have an interface located at the boundary of two components [250]. It is extremely important for a heterogeneous catalyst to have a proper interfacial structure because the interface region always presents unique physical and chemical properties [251]. These unique properties can facilitate the capability of the resultant material to bind, convert, and transport the surface species (e.g., adsorbents, electrons, and intermediates), greatly promoting the catalytic reactions occurring on the surface [252–255]. In recent years, abundant research studies have reported the design and synthesis of electrocatalysts for water–hydrogen electroconversion with the assistance of interface engineering. In general, depending on the relative locations of the components, hybrid materials can be classified as supported structures, heterostructures, or core–shell structures [256]. The special character of a supported structure is that the support component is much larger than the other components; in contrast, the components in a heterostructured material have similar sizes. In a core–shell structure, one component is covered by another component, with interfaces existing at the boundary between the two components. These three types of hybrid materials with different interface structures have been reported as a result of research assembling metals, metal oxides, nonoxides, and so forth. For example, Zhang et al. [257] reported on a MoNi4 electrocatalyst immobilized on MoO2 cuboids (MoNi4/MoO2@Ni) with a supported structure that was made by controlling the outward diffusion of nickel atoms during calcination. By heating the NiMoO4 precursor under a reduction atmosphere, MoNi4 nanoparticles (20–100 nm) supported by MoO2 cuboids (~1 μm) were synthesized, and the supported hybrid catalyst was found to exhibit excellent HER activity in alkaline solution (Figs. 18 (a) and (b)). Heterostructured Co/CoP nanoparticles were prepared by Xue et al. [258] via the gradual phosphidation of Co metal into CoP components. By changing the weight ratios of the NaH2PO2 and Co species, the CoP contents in Janus Co/CoP nanoparticles could be controllably modulated, affecting the interface zone in the Co/CoP catalyst (Figs. 18(c)–(f)). As illustrated in Figs. 18(g) and (h), Xu et al. [259] fabricated NiCo-based porous microrod arrays composed of carbon-confined NiCo@NiCoO2 core@shell nanoparticles (NiCo@NiCoO2/C PMRAs (porous microrod arrays)) by the reductive carbonization of bimetallic (Ni, Co) metal–organic framework microrod arrays and subsequent controlled oxidative calcination. The obtained NiCo@NiCoO2/C PMRAs combined several desirable qualities for OER, including a large surface area, good conductivity, and multiple electrocatalytic active sites.For the purpose of rational catalyst design, the origin of the enhanced catalytic performance of these hybrid materials with an abundant interface has been deeply explored. The modulation of the electronic structure in the interface of hybrid materials has been well proven [260–267]. An observed electron transfer was evidenced by Yu et al. [268] using X-ray absorption near-edge spectroscopy (XANES) for a Ni(OH)2/Pt catalyst. As pointed out by the authors, the pre-edge and main absorption edge of Ni for both the α- and β-Ni(OH)2/Pt electrode are shifted to lower energies, and the absorption intensities decrease in comparison with their counterparts, demonstrating that the electrons transfer from the Pt substrate to the hydroxide. Han et al. [269] revealed a strong metal-support interaction (SMSI) effect between CoS2 and Pt in the Pt/CoS2 hybrid system for highly active water splitting. A downward shift in the Pt d-band structure was proved by DFT calculation and was further supported by X-ray photoelectron spectroscopy (XPS) and X-ray absorption fine structure (XAFS) analyses. Although the electron transfer has been well demonstrated in these metal/metal compound catalysts, the underlying mechanisms for the high catalytic activity are ambiguous. In order to uncover this mystery, our group studied the chemical properties of the metal/metal oxide interface, and found an interface-induced synergistic effect—the “chimney effect” (Figs. 19 (a) and (b)) [270]. As revealed by the DFT results (Figs. 19(c) and (d)), the neighboring sites around the interface are immune to H2O* and OH* species but selectively adsorb H*, which prevents the poison effect of H2O* and OH* on active sites. Meanwhile, the active sites on the interface have good capability for the adsorption and desorption of the H* reactant for the H2 product, as its ΔG H* is close to 0. Due to these features, the hydrogen evolution on metal oxide/metal catalysts behaves similarly to a chimney for generating hydrogen productive continuously. The strong positive correlation between the HER activity of the metal oxide/metal composites and the interfacial metal atoms (Fig. 19(e)) further confirmed that the hydrogen generation can be greatly accelerated by increasing the amount of active site. With the exception of metal/metal compound catalysts, the origin of the performance enhancement of metal compound hybrid catalysts was further investigated by our group through the casting of monometallic NiO–Ni3S2 heteronanosheets [271]. The electron transfer from Ni–S to Ni–O that occurs at the interface of NiO–Ni3S2 results in a superior activity of overall water splitting that is even superior to the benchmark Pt/C and RuO2 assembly (Fig. 19(g)). DFT calculation results further demonstrated that the activation barrier of the hydrogen (or oxygen-containing) intermediates on the NiO–Ni3S2 interfaces are significantly lowered, which is beneficial for the OER and HER processes (Figs. 19(h) and (i)).Vacancy generation is another important factor in developing interfaced materials with improved catalytic performances [272–277]. A typical example was reported by Li et al. [278], who prepared FeS2/CoS2 hybrid nanosheets (NSs) for overall water electrolysis. In the synthesis process, CoFe2O4 nanoparticles were first synthesized by a coprecipitation method, and then transformed into the FeS2 and CoS2 phases. During the latter sulfuration process, interfaces with defect sites were created in the FeS2/CoS2 NSs. Electron paramagnetic resonance (EPR) spectra revealed a stronger EPR signal at g = 2.007 recorded with the FeS2/CoS2 NSs composite, indicating abundant S vacancies. Extended X-ray absorption fine structure (EXAFS) was used to further investigate the local structure of the obtained samples, and a distinct decrease in peak intensity in Fe K-edge EXAFS was detected with FeS2/CoS2 NSs, demonstrating the coordination deficiency from Fe. Similar results were reported by Gao et al. [279] in CeO2/NiO with an embedded configuration (Ce–NiO–E) or the surface-loaded configuration (Ce–NiO–L). An increasing trend of Ni3+ and oxygen defects for NiO (Ni3+ 62%, O defect 24%), Ce–NiO–L (Ni3+ 69%, O defect 26%), and Ce–NiO–E (Ni3+ 71%, O defect 32%) correlated well with the activity trend, indicating that these vacancy sites at the interface region contribute greatly to the enhanced activity. In fact, electron modulation and vacancy do not exist independently, and the catalytic performances of the interfaced materials may be influenced by the two factors simultaneously.Alloying is the diffusion penetration of the metal atoms of two or more metals, or the addition of nonmetallic elements into metals through melting, sintering, or vapor deposition processes. Metal alloying is an effective strategy for improving the performance of metal catalysts. It can not only refine the grain size to improve the mechanical strength and surface area of catalysts, but also selectively reduce the amount of a single component (e.g., Pt, Au, Ru, and other precious metals) in order to reduce the catalyst cost. Moreover, the catalytic activity and selectivity of metals can be changed by incorporating other metals to form alloys, due to the synergistic effect between metal components [280–284] According to the Brewer–Engel valence bond theory [285], alloying a metal with an unfilled d orbital and a metal with internally paired d-electrons can modulate the hydrogen adsorption energy on the alloy surface, thus improving the hydrogen evolution activity. Raj [286] prepared a series of Ni-based binary composites by means of electrodeposition techniques. The trend in their HER activities in alkaline water followed this order: Ni–Mo > Ni–Zn > Ni–Co > Ni–W > Ni–Fe > Ni–Cr. Due to its excellent activity, the Ni–Mo alloy has been considered to be the most promising HER catalytic material. Zhang et al. [287] successfully constructed a layer of Ni–Mo alloy nanorods with uniform size and uniform element distribution on the surface of a nickel substrate by means of magnetron sputtering technology, and found that the HER exchange current density was almost ten times higher than that of single-metal Ni or Mo catalysts. This research showed that the excellent catalytic activity of the Ni–Mo alloy electrode mainly comes from two factors: ① There is an increase in the specific surface, caused by the grain refinement of the two-component metal in the growth process; and ② the electronegativity difference between the elements Ni and Mo causes electrons to concentrate around Mo, thus forming a synergy for electrocatalysis.In order to reduce the usage of precious metals, many non-noble metals (i.e., Co, Ni, Fe, Cu, V, Cr, Mn, Zn, etc.) have been incorporated with noble metals to form alloys as electrocatalysts [206,288]. Studies of ORR performances on PtM alloys revealed the following activity order [289–291]: PtFe/C > PtCo/C > PtV/C > PtNi/C > Pt/C. The following stability order was also determined [292]: Pt3Ir(111) > Pt3Co(111) > Pt3Ni(111) > Pt3Fe(111). In addition, Greeley et al. [293] found a classic volcano-shaped dependency based on the ORR activity of Pt alloys versus the d-band center position of 3d metals (Fig. 20 ). According to that study, the ORR mechanism on Pt3M catalysts was either O2 dissociation or proton/electron transfer to molecular O2, and the optimal ORR catalyst should have a weaker oxygen-binding energy than Pt. Bampos et al. [294] further synthesized a series of carbon-supported Pd–M (where M = Ag, Co, Cu, Fe, Ni, or Zn) bimetallic catalysts as ORR electrocatalysts in acidic solution. The ORR activity of these Pd-based alloys was found to descend as follows: PdZn/C > PdNi/C > Pt/C > PdAg/C ≥ PdCo/C > PdFe/C > PdCu/C > Pd/C. Notably, PdZn/C exhibited a specific activity three times higher than Pt/C at a potential between 0.35 and 0.5 V versus Ag/AgCl.In addition to metallic elements, the introduction of nonmetallic elements can improve the catalytic performance of alloys [295–299]. For example, Kiran et al. [296] prepared few-layer MoS2(1– x )Se2 x alloys that possessed higher HER activity than pristine MoS2 and MoSe2. A systematic structure-activity relationship was revealed by tuning the Se/S incorporation in MoS2(1– x )Se2 x , with MoS1.0Se1.0 having an Se/S ratio of 1:1 showing the highest HER activity. Similar results regarding Mo–S–Se alloys were reported by the groups of Gong et al. [299] and Xu et al. [297]. Xu et al. [298] successfully controlled the composition of sulfur (S) and selenium (Se) in ternary WS2(1– x )Se2 x nanotubes on carbon fibers. The disordered atom arrangement introduced in the alloyed structure resulted in WS2(1– x )Se2 x having excellent electrocatalytic properties for HER. A ternary pyrite-type CoPS was further obtained by Cabán-Acevedo et al. [300] for photo/electrochemical hydrogen evolution. Because of the higher electron-donating character of the P2– ligands in CoPS, this ternary pyrite CoPS was endowed with a more thermoneutral hydrogen adsorption, resulting in a higher HER activity than CoS2.The majority of the energy dissipation in electrochemical hydrogen–water conversion is induced by the activation energy driving the electrochemical reactions involved in the energy system. Materials are at the core of a high-efficiency energy system, as the activation energy of the reactions involved in the system is strongly influenced by the catalytic materials. It is urgent to accelerate the development of active and robust electrocatalysts with acceptable cost in order to ensure that an efficient and sustainable energy system becomes a reality. In general, a catalyst’s performance depends on two main factors: ① the number of active sites in a given area; and ② the intrinsic activity of each active site. Tuning the geometric structure (i.e., the nanoarchitecture and facet engineering) makes it possible to realize enhancement of the active sites, because when the catalyst size is reduced to the nanoscale, the catalyst possesses a high surface area with increased exposure of the catalytically active planes. Defect engineering and amorphization can also be applied to enrich active sites by exposing thermodynamically unstable active sites or active unsaturated atoms. Increasing the intrinsic activity of the active site is a more fundamental yet difficult strategy for activity improvement, which involves accurate modulation of the electronic structures of catalytic materials. Optimization of the intrinsic activity can be achieved based on a fundamental understanding of the particularities of the reactions and insight into the design of catalysts with targeted functionalities. The featured examples (their catalytic performances are summarized in Tables 2–5 ) discussed herein present successful modulation of the intrinsic activity via element doping, interface engineering, polymorph engineering, alloying, and so forth, with the aid of synergistic interaction between theoretical and experimental studies.At present, with the rapid development of chemical synthesis techniques, many catalytic materials have been reported with outstanding electrocatalytic performance. Meanwhile, developments in physical characterization methods and theoretical calculation are providing more evidence and guidance for researchers, allowing us to better understand the performance-improvement mechanism of catalysts. However, the new-generation electrode catalysts are mainly being produced through traditional trial-and-error processes rather than by means of rational design. This research model not only wastes social resources and the energy of scientific researchers, but also greatly delays scientific development. Thus, the development of a performance-oriented design strategy for electrocatalyst design from the most fundamental level to a practical application level is urgent. This target requires computational chemistry, electrocatalytic chemistry, and synthetic chemistry working in concert; unfortunately, however, all three fields require improvement at present. Current theoretical calculation models are still imperfect, as most do not take into account the influence of ionic strength, the double-layer effect, the solvation effect, and so forth, making it difficult to accurately reflect the actual reaction process of the catalytic material surface. Furthermore, although the reaction mechanisms of these hydrogen- and oxygen-involving reactions have been widely investigated, the actual mechanisms on different catalyst surfaces remain shrouded in mystery. In fact, it is also difficult to identify the actual active sites, as the surface structure of most catalysts varies during the electrocatalytic process. New progress will come from the merging of advanced theoretical calculations and experimental characterization (e.g., in situ, ex situ, and operando techniques), which will allow us to promote our understanding of the electrochemical reaction mechanism and the dynamic evolution of the catalysts involved at the molecular level. Finally, function-oriented catalyst preparation remains a major challenge. Existing material synthesis technologies are able to regulate the physical and chemical properties of specific catalysts to some extent at the nanoscale, but the use of controlled synthesis technology to modulate the electronic structure of catalysts at the atomic scale remains immature. The development of controllable synthesis methods with strong applicability and large-scale production is another focus of future research.We gratefully acknowledge financial support from the National Natural Science Foundation of China (21576032 and 51772037), the Key Program of the National Natural Science Foundation of China (21436003), the Major Research Plan of the National Natural Science Foundation of China (91534205), and the National Program on Key Basic Research Project of China (2016YFB0101202).Lishan Peng and Zidong Wei declare that they have no conflict of interest or financial conflicts to disclose. c reactant surface concentration, mol·m−3 F Faraday constant, ∼96 485 C·mol−1 G Gibbs free energy, J·mol−1 f decay rate to products and reactants, dimensionless i current, A j current density, A·cm−2 j 0 exchange current density, A·cm−2 n number of electrons transferred in the reaction, dimensionless R resistance, Ω T temperature, K U 0 reaction equilibrium potential, V V voltage, V α charge transfer coefficient, dimensionless μ overvoltage, V ΔG act activation energy barrier, J·mol−1 ΔG H* hydrogen adsorption free energy ΔG max gibbs free-energy change, J·mol−1 ΔG O* oxygen adsorption strength reactant surface concentration, mol·m−3 Faraday constant, ∼96 485 C·mol−1 Gibbs free energy, J·mol−1 decay rate to products and reactants, dimensionlesscurrent, Acurrent density, A·cm−2 exchange current density, A·cm−2 number of electrons transferred in the reaction, dimensionlessresistance, Ωtemperature, Kreaction equilibrium potential, Vvoltage, Vcharge transfer coefficient, dimensionlessovervoltage, Vactivation energy barrier, J·mol−1 hydrogen adsorption free energygibbs free-energy change, J·mol−1 oxygen adsorption strength

In the context of the current serious problems related to energy demand and climate change, substantial progress has been made in developing a sustainable energy system. Electrochemical hydrogen–water conversion is an ideal energy system that can produce fuels via sustainable, fossil-free pathways. However, the energy conversion efficiency of two functioning technologies in this energy system—namely, water electrolysis and the fuel cell—still has great scope for improvement. This review analyzes the energy dissipation of water electrolysis and the fuel cell in the hydrogen–water energy system and discusses the key barriers in the hydrogen- and oxygen-involving reactions that occur on the catalyst surface. By means of the scaling relations between reactive intermediates and their apparent catalytic performance, this article summarizes the frameworks of the catalytic activity trends, providing insights into the design of highly active electrocatalysts for the involved reactions. A series of structural engineering methodologies (including nanoarchitecture, facet engineering, polymorph engineering, amorphization, defect engineering, element doping, interface engineering, and alloying) and their applications based on catalytic performance are then introduced, with an emphasis on the rational guidance from previous theoretical and experimental studies. The key scientific problems in the electrochemical hydrogen–water conversion system are outlined, and future directions are proposed for developing advanced catalysts for technologies with high energy-conversion efficiency.

With increasing concerns of the fossil fuel–related environmental crisis and global warming, there is an imperative demand for developing alternative green and sustainable energy conversion and storage technologies, such as batteries, fuel cells and water electrolysis [1–7]. As a crucial reaction of secondary metal-air batteries and electrochemical water splitting, oxygen evolution reaction (OER) plays an important role in the efficiency and operational stability of such systems [8–12]. High-performance electrochemical catalysts are thus urgently required to promote sluggish OER reaction kinetics, to provide low overpotentials (η) and excellent catalytic stability. Considering the rarity and high cost of noble metal OER catalysts (e.g. Ir- or Ru-based materials) [13,14], earth-abundant and cost-effective transition metal based electrocatalysts are clearly desired.Metal-organic frameworks (MOFs) are a class of porous materials composed of organic linkers and metal nodes with coordination bonding [15]. MOFs and derived materials have been used in a wide range of fields, e.g. gas storage and separation, batteries and catalysis etc., benefiting from their high specific surface area, tuneable porosity and abundant active metal sites [16–21]. Two-dimensional (2D) MOFs based materials have attracted growing attention for OER catalysts with unique physicochemical features. The 2D structure enables hydroxide units in the electrolyte to easily reach the active site and fast dissociation of generated O2, as well as shortening the electron transfer pathway through the thin film to the conductive support [22–24]. 2D MOFs can also be engineered to possess a large number of coordinatively unsaturated metal atoms exposed as the active sites [25,26]. The atomic surface structure and bonding arrangement can be elaborately controlled to facilitate the interaction between the active site and the reaction intermediates for superior OER electrocatalysts [27,28]. However, 2D MOFs have a high tendency to aggregate [29,30], leading to a decreased effective surface area during operation. Avoiding aggregation with high active surface area and improving the integral structural stability for superior OER catalysts are thus essential.To tackle these issues, there are increasing reports demonstrating that the introduction of functional nanoscale components (nanosheets or nanoparticles etc.) in a MOF composite could prevent the aggregation and enhance the integral structural stability during operation [30,31]. Meanwhile, the electronic structure of the metal units in the MOFs can be optimized by incorporation of heterogeneous metal-containing groups for superior OER performance [30–35]. For example, Qiao and coworkers have synthesized 2D Ni-BDC/Ni(OH)2 heterostructure, exhibiting a lower η of 320 mV at 10 mA cm−2 than that of Ni-BDC nanosheets, and a good catalytic durability of 20 h. The improved OER performance is attributed to the rational design of the composition and structure of the composite, and to the mitigation of aggregation of Ni-BDC by coupling with Ni(OH)2 [30]. Similarly, Qin et al. have reported hybrids of Fe-Co polyphenolic network–wrapped Fe3O4 nanocatalysts for enhanced OER with an η of 260 mV at 10 mA cm−2 and a durability of over 24 h, taking advantage of strong metal-polyphenolic ligand complexation that ensures robust metal-polyphenolic shells for prolonged operations [31]. Inspired by these reports, 2D Ni-based MOFs could be promising candidates for constructing hybrid electrocatalysts due to an excellent surface structure and physicochemical features. Meanwhile, considering Fe3O4 nanoparticles with good electrical conductivity (>100 S cm−1) and fast electron transfer between Fe2+ and Fe3+ in the crystals [31,36], the incorporation of Fe3O4 nanoparticles on the surface of 2D Ni-based MOFs could be promising candidate to be used for OER.Herein, we have successfully prepared ultrasmall Fe3O4 nanoparticles that are uniformly immobilized on 2D Ni-based MOFs (Fe3O4/Ni-BDC). The functionalized Fe3O4 nanoparticles (6 ± 2 nm) with abundant surface hydroxide groups are initially synthesized, and then either added directly during the synthesis of 2D Ni-BDC layers (4 ± 1 nm) or alternatively mixed with pre-synthesized 2D Ni-BDC. We have investigated the resulting morphology changes and electronic structure modulation to tackle the aggregation issue for OER via tuning the amount of Fe3O4 immobilized on the 2D Ni-BDC layers. The optimized Fe3O4/Ni-BDC-4 composite demonstrates significantly enhanced OER performance with an η of 295 mV at 10 mA cm−2, a Tafel slope of 47.8 mV dec-1 and an excellent catalytic durability over 40 h. DFT calculations are further conducted to identify the active site and help to understand how the valance states of the transition metals affect the OER performance.Iron (III) chloride hexahydrate (FeCl3·6H2O, 97%), sodium bicarbonate (NaHCO3, ≥99.7%), L-ascorbic acid, terephthalic acid (1, 4-BDC, 98%), N,N-dimethylformamide (DMF, 99.8%), Nafion solution (10 wt%) were bought from Sigma-Aldrich. Triethylamine (TEA, 99%) was purchased from Merck (Germany). Nickel (II) chloride hexahydrate (NiCl2·6H2O, 98%) was from BDH Chemicals Ltd Poole England. All chemicals were used as received without further purification. Ultrapure water (≥18.25 MΩ·cm, Sartorius arium® pro, Germany) was used to prepare all the aqueous solutions.Water-dispersible Fe3O4 nanoparticles were synthesized according to a previous report [37]. Briefly, a 20 mL aqueous solution of 1 mM L-ascorbic acid was added into a 60 mL aqueous mixture of 6 mM FeCl3·6H2O and 18 mM NaHCO3, under stirring for 20 min. The mixture was transferred to a 150 mL Teflon-lined stainless-steel autoclave, which was kept at 150 °C for 6 h. The product was separated using a magnet, washed with ultrapure water more than three times, leading to Fe3O4 nanoparticles that could be re-dispersed in water for further using.Ni-BDC was grown with or without the presence of water-dispersible Fe3O4 for Fe3O4/Ni-BDC composites. 63 mg of 1,4-BDC was first dissolved in a mixed solvent of DMF (15 mL), ethanol (1 mL), and ultrapure water (1 mL), into which 90 mg of NiCl2·6H2O and a certain volume of water or Fe3O4 dispersion were added subsequently and under ultrasonication for 10 min, followed by a quick injection of 0.50 mL TEA. To optimize the ratio of Fe3O4 and Ni-BDC in the composite, different volumes of Fe3O4 dispersion (12 mg mL−1; 1, 2, 3, 4 or 5 mL) were used. The above mixture was sealed and continuously ultra-sonicated for 6 h at room temperature. Finally, the precipitates were centrifuged and washed with ethanol three times, followed by drying in a vacuum oven at 60 °C for 12 h. The obtained composites were collected and labelled as Fe3O4/Ni-BDC-n (n = 1, 2, 3, 4 and 5), where n is the volume in mL of Fe3O4 dispersion used.The crystallinity of the synthesized materials was characterized by X-ray diffraction (XRD, D8 Advance X-Ray diffractometer (Huber)). X-ray photoelectron spectroscopy (XPS) was performed by drop-casting samples onto silicon substrates with a Thermo-Scientific system (Al-Kα radiation, 1484.6 eV). Fourier transform infrared spectroscopy (FTIR) measurements were performed on an Alpha-P FTIR spectrometer (Bruker) in the range of 4,000–400 cm−1 with a resolution of 2 cm−1. The specific surface area was estimated by a surface area and pore size analyzer (ASAP 2020, Micromeritics). Elemental analysis was performed by inductively-coupled plasma optical emission spectrometry (ICP-OES). Micro-, nanostructure and composition characterization were conducted with scanning electron microscopy (SEM, Quanta FEG 200 ESEM, 15 kV), atomic-force microscopy (AFM, Agilent Technology 5500, tapping mode, a mica sheet as the substrate), and transmission electron microscopy (TEM, Tecnai G2 T20, 200 kV).In order to prepare catalyst inks, 4 mg of active materials (Fe3O4/Ni-BDC-n, Ni-BDC or Fe3O4), 4 mg of carbon black (Alfa Aesar™) and 25 μL of Nafion solution (10 wt%) were mixed with 0.75 mL of 2-propanol and 0.25 mL of ultrapure water for a uniform ink after 1 h sonication. Prior to use, rotating disk electrodes (RDE) using glassy carbon electrodes (GCEs, d = 5.0 mm, A = 0.19625 cm2) and rotating ring disk electrodes (RRDEs, d disk = 5.61 mm, A disk = 0.2472 cm2, d ring (inner) = 6.25 mm, d ring (outer) = 7.92 mm, A ring = 0.1859 cm2) with a GCE disk and a Pt ring were polished on a polishing pad with Al2O3 slurries with decreasing particle diameters (1.0, 0.3 and 0.05 µm). Afterwards, 10 µL of the catalyst ink was drop-cast onto the surface of the GCEs, leading to a mass loading of 0.398 mg cm−2, and dried under room temperature. Electrochemical tests were carried out in a typical three-electrode setup with 1.0 M KOH solution as the electrolyte on an electrochemical workstation (Autolab PGSTAT12) with a graphite rod as the counter electrode and a Ag/AgCl (sat. KCl) as the reference electrode. Rotation of the RDE and RRDE were controlled on a Pine Instruments rotating system. The applied potentials were compensated for the solution resistance R s and current I via: E Ag/AgCl-corr = E Ag/AgCl - IR s [38], where the uncompensated Ohmic solution resistance (R s) in the high-frequency region was measured by electrochemical impedance spectroscopy (EIS) in a frequency range from 100 kHz to 0.1 Hz at 1.525 V vs. RHE. All measured potentials were calibrated to reversible hydrogen electrode (RHE) potential according to the following equation: E RHE = E Ag/AgCl-corr + 0.197 + 0.059 × pH. To maintain the O2/H2O equilibrium at 1.23 V versus RHE, oxygen gas (O2 ≥ 99.995%) flow was kept in the electrolyte during the test. The η for OER was defined as: η = E RHE −1.23 V. For OER tests, working electrodes were initially scanned for 10 cycles using cyclic voltammetry (CV) to obtain stable signals. Then, linear sweep voltammetry (LSV) curves were obtained at a slow scan rate of 5 mV s−1 at a rotational speed of 1,600 rpm to decrease capacitive currents and interference from generated gas bubbles. The electrode durability was evaluated by chronopotentiometry at a current density of 10 mA cm−2. The Tafel slope (b) was calculated based on the Tafel equation [39,40]: (2.1) η = a + b ∙ log j and compared to b = 2.303RT/αnF, where j is current density of samples from the LSV test, α is the charge transfer coefficient, n is the number of transferred electrons during the redox reaction, F is the Faraday constant (96485 C mol−1), R is the gas constant (8.314 J mol−1 K−1), and T is the temperature (K).The electrochemical double-layer capacitance (C dl) was tested using CVs in a narrow potential range of 1.223–1.323 V vs. RHE, with scan rates of 40, 60, 80, 100, and 120 mV s−1. The plot of Δj = (j a − j c), where j a and j c are the anodic and cathodic current, respectively, at 1.24 V vs. RHE (no faradaic reaction occurring) against the scan rate had a linear relationship, whose slope was twice of C dl. The electrochemically active surface area (ECSA) relative to GCE and GCE-normalized current density were calculated according to the equations [41]: (2.2) ECSA = C d l _ s a m p l e s C d l _ G C E A geo (2.3) j E C S A _ n o r m a l i z e d = j ECSA Herein, C dl-GCE is the specific capacitance for a plane surface in the range of 20–60 μF cm−2, and C dl-GCE = 40 μF cm−2 was used [32]. A geo is the geometric area of the GCE. To investigate the reaction mechanism for OER, RRDE voltammograms were recorded to determine the OER reaction pathway by measuring the HO2 − formation, with the ring potential held at 1.50 V vs. RHE at 1,600 rpm, and linearly scanning the potential of the disk in O2-saturated 1.0 M KOH solution.Spin-polarized density functional theory (DFT) calculations were performed using the Vienna ab-initio simulation package (VASP) [42,43]. The Perdew − Burke − Ernzerhof (PBE) functional within the generalized gradient approximations (GGA) was used to account for exchange correlation effects. The projector-augmented wave method was used to represent the core-valence electron interaction. The valence electronic states were expanded in plane-wave basis sets with energy cutoff at 450 eV, and the force convergence criterion in the structure was set to be 0.05 eV/Å. The transition states were searched using a constrained optimization scheme [44-48]. The Hubbard U approach was used to correct Ni and Fe 3d orbitals with U eff = 3.8 eV [49-52]. The zero-point energies (ZPE) and entropy corrections to the free energies at room temperature (298.15 K) were applied. DFT-D3 method with Becke-Johnson (BJ) damping was employed to describe Van der Waals interaction [53,54]. The free energy of gas-phase O2 molecule is discussed in Supporting Information.On the basis of our XRD characterization (Fig. 1 a and Fig. S1) and previous report [52], the (200) surface of Ni-BDC was constructed and used to study the interactions between catalysts and adsorbates. A p(2 × 1) Ni-BDC catalytic surface system was modelled with two metal oxide layers separated by one BDC layer (Fig. S22). A (5 × 5 × 1) k-point was used for geometry optimization with quasi-Newton algorithm. The (311) plane of Fe3O4, as an example of a curved surface of the ultrasmall nanoparticles, was used for constructing a model of the active surface, exposing five-coordinated Fe sites, two-coordinated and three-coordinated O sites (Fig. S23). Four stoichiometric layers of Fe3O4(311) on the Ni-BDC(200) surface was created to model a Fe3O4/Ni-BDC composited structure. A (1 × 5 × 2) k-mesh was used for geometry optimization. While there could be residual surface groups from the hydroxide groups and DHAA, the models used for DFT focus on the possible interaction in this Fe3O4/Ni-BDC composited structure.Fe3O4/Ni-BDC composites are synthesized through a two-step procedure as illustrated in Scheme 1 . Water-dispersible Fe3O4 nanoparticles are initially prepared with a modified hydrothermal method [37]. Dehydroascorbic acid (DHAA) is oxidized from the ascorbic acid, serving as a stabilizer and capping ligand on surfaces of Fe3O4 nanoparticles interacted by carbonyl groups, ensuring a good dispersibility of Fe3O4 nanoparticles in aqueous solution [55,56]. During the subsequent sonication process, functionalized Fe3O4 nanoparticles are homogenously dispersed in an alkaline solution (pH ~ 10), immobilized on 2D Ni-BDC. Triethylamine (TEA) serves as the deprotonating agent and the modulator to promote the nucleation of pristine Ni-BDC for monodisperse film with the controllable orientation [57]. Introducing the Fe3O4 nanoparticles solution into the synthetic process of Ni-BDC leads to lower TEA concentration, restraining the continuous nucleation of Ni-BDC layers. As a consequence, smaller Ni-BDC layers with more defects and edges are obtained, and with Fe3O4 nanoparticles immobilized, leading to Fe3O4/Ni-BDC composites.The water-dispersible Fe3O4 nanoparticles (Fig. 1a) were characterized by XRD analysis, showing peaks matching well with the standard Fe3O4 phase (JCPDS no. 89–0688), with peaks at 18.3°, 30.0°, 35.4°, 43.0°, 56.9° and 62.5° fitting well with (111), (220), (311), (400), (511) and (440) planes of magnetite, respectively. The diffraction peaks of pure Ni-BDC (Fig. 1a) can be assigned based on Ni-BDC composites in literature [27,52]. The main diffraction peaks at 8.8°, 15.7° and 17.9° are ascribed to the (200), (201) and (-201) planes of Ni-BDC, respectively. Fe3O4/Ni-BDC-n composites with various amount of Fe3O4 are successfully synthesized, and XRD patterns of Fe3O4/Ni-BDC-4 (Fig. 1a) confirm the co-existence of the crystalline phases of Fe3O4 and Ni-BDC. Similar diffraction peaks are also observed on those of Fe3O4/Ni-BDC-1, 2, 3 or 5 (Fig. S1). It can be observed that the diffraction peaks of Ni-BDC become weaker with the increase of Fe3O4 content, a phenomenon due to the process that Fe3O4 nanoparticles limit the growth of Ni-BDC layers. In addition, FTIR spectra of Fe3O4, Ni-BDC and Fe3O4/Ni-BDC-4 are shown in Fig. 1b. The band at 553 cm−1 is assigned to the Fe-O stretching vibration of Fe3O4 [58], with the bands in the range of 600–1300 cm−1 attributed to out-of-plane vibrations of the BDC linker [59]. The strong bands at 1373 and 1564 cm−1 are ascribed to the stretching modes of the coordinated carboxylate (–COO-) of the terephthalate linker of Ni-BDC, indicative of the presence of both Fe3O4 and Ni-BDC [59]. The absorption band at 1647 cm−1 is regarded as the coordination between O atom of the carbonyl group (C = O) from DHAA and Fe units from the surfaces of Fe3O4 nanoparticles [37]. The broad band at 3315 cm−1 is attributed to the strong stretching mode of hydroxyl groups (–OH) [37,59,60].The chemical bonding states of Fe3O4, Ni-BDC and Fe3O4/Ni-BDC composites are investigated by XPS. All signals originating from expected elements (Ni, Fe, O or C) are recorded in the survey spectra (Fig. S2). The high-resolution Ni 2p spectra (Fig. 1c) are deconvoluted into two satellite peaks and a couple of peaks for Ni2+ (854.8/872.5 eV for Fe3O4/Ni-BDC-4 and 860.2/878.5 eV for Ni-BDC) [27]. In addition, in high-resolution Fe 2p spectra (Fig. 1d), a spin–orbit doublet at 709.2/722.6 eV and 709.4/723.0 eV is assigned to the Fe2+ in Fe3O4/Ni-BDC and Fe3O4, respectively [61]. Peaks at 712.2/725.9 eV and 712.2/725.9 eV are belong to the Fe3+ in Fe3O4/Ni-BDC and Fe3O4, respectively [34]. High-resolution Ni 2p (Fig. 1c) and Fe 2p (Fig. 1d) spectra indicate the Ni 2p3/2 (856.7 eV for Ni-BDC) and Fe 2p3/2 (710.4 eV for Fe3O4) shift to lower and higher binding energies (Fig. 1c, d and S3; Ni 2p3/2 and Fe 2p3/2 of Fe3O4/Ni-BDC-4), respectively [52,62]. The detailed binding energy data are summarized in Table S1. Binding energy level of Ni 2p3/2 decreases and while Fe 2p3/2 increases slightly with the amount of Fe3O4 nanoparticles among the five composites. This indicates that there is change in the bond strength of both Fe and Ni to varying degree in the different Fe3O4/Ni-BDC composites. The typical O 1 s spectrum of Fe3O4/Ni-BDC-4 (Fig. 1e) indicates peaks at 533.4, 532.4 and 531.6 eV, that are assigned to O–H, O = C-O and Ni-O bonding, respectively, originating from the terephthalate linker and NiO6 octahedra in Ni-BDC [27]. Other peaks at 531.2, 530.7 and 529.8 eV are assigned to O–H, O = C and Fe-O bonding, respectively, attributed to the surface hydroxyl and carbonyl groups from the DHAA and internal Fe-O units of the water-dispersible Fe3O4 [61,63]. To quantify the specific surface area and pore sizes of Fe3O4/Ni-BDC-4, the typical adsorption − desorption isotherm is recorded using nitrogen adsorption − desorption measurements (Fig. 1f). The specific surface area is determined to be 136 m2 g−1, which is attributed to the slit-like structure formed by aggregation of Fe3O4/Ni-BDC-4 [52]. The pore size distributions of the composite are mainly centered at 40 nm. Besides, the pore size distribution indicates the presence of micropores (<2 nm), mesopores (2–50 nm) and macropores (>50 nm), confirming a hierarchical porous structure for Fe3O4/Ni-BDC-4, which is beneficial for mass transport for OER. Finally, ICP-OES is conducted to check the metallic components of various Fe3O4/Ni-BDC composites (Table S2). It verified that proportional Fe in the composite increases with the added amount of Fe3O4. Fe3O4/Ni-BDC-4 is composed of 24.7 at% Ni and 75.3 at% Fe.The morphologies of Ni-BDC, Fe3O4 and Fe3O4/Ni-BDC are characterized by SEM and TEM. SEM images of pure Ni-BDC (Fig. 2 a, Fig. S4a and b) show a hierarchical-layer structure comprised of aggregated 2D nanosheets. After immobilizing different amounts of Fe3O4 nanoparticles (Fig. 2b, Fig. S4c and d), the generated Fe3O4/Ni-BDC composites show changes in morphology and microstructures. With increasing the ratio of Fe3O4, the Fe3O4/Ni-BDC (Fig. 2c, Fig. S4e and f, Fig. S5) turns from a layered structure to a smoother structure with smaller Ni-BDC grains. Likely caused by Fe3O4 nanoparticles anchoring Ni-BDC on their surface hindering extended growth of Ni-BDC layers. As a control to assess if any leaching Fe from could cause MOF formation, samples following the same synthetic route of Fe3O4/Ni-BDC-4 without the addition of NiCl2·6H2O are also fabricated, but were difficult to purify and separate from solution. SEM images and the corresponding EDS spectrum (Fig. S6) indicate that the control sample is the water-dispersible Fe3O4 nanoparticles with no formation of Fe based MOF. This further confirms the importance of Ni source (NiCl2·6H2O) in the formation of Fe3O4/Ni-BDC composites. TEM image of pure Ni-BDC (Fig. 2d) clearly demonstrates a two-dimensional hierarchical-layer structure. The AFM image (Fig. S7) of partial Ni-BDC samples further indicates the thickness of Ni-BDC nanosheets is 5 ± 1 nm. TEM image of Fe3O4 (Fig. 2e) obviously shows ultrafine nanoparticles with a particle size of 6 ± 2 nm. The interplanar spacing of the lattice (Fig. 2f) is measured to be 0.485 nm, matching well with the (111) plane of magnetite Fe3O4 (JCPDS no. 89–0688) [64]. When the mass ratio of Fe3O4 is low, the high-resolution TEM image of Fe3O4/Ni-BDC-1 (Fig. S8) clearly shows the boundary between Fe3O4 and Ni-BDC. Fe3O4 nanoparticles are tightly anchored on the Ni-BDC layers, originating from the strong coupling effects between them. With a higher mass ratio of Fe3O4, the typical TEM images of Fe3O4/Ni-BDC-4 (Fig. 2g-h and S9) show that ultrafine Fe3O4 nanoparticles are homogenously immobilized on the Ni-BDC layers. The high-resolution TEM image of Fe3O4/Ni-BDC-4 (Fig. 2i) demonstrates a crystalline interplanar spacing of 0.297 nm, in accordance with the (220) plane of magnetite Fe3O4 (JCPDS no. 89–0688) [65]. The STEM-EDS elemental mapping images corresponding to a fragment of Fe3O4/Ni-BDC-4 (Fig. 2j-n) suggests the homogeneous distribution of nickel (cyan), iron (red), oxygen (green) and carbon (purple) elements, confirming that Fe3O4 nanoparticles are uniformly distributed on Ni-BDC sheets. Linear elemental distribution of Fe3O4/Ni-BDC-4 composite (Fig. S10) further verifies that ultrasmall Fe3O4 nanoparticles distributes on the surface of Ni-BDC.OER performance of the proposed electrocatalysts was investigated in a conventional three-electrode cell containing O2-saturated 1.0 M KOH solution by LSV at a scan rate of 5 mV s−1. As controls, the OER activities of pristine Ni-BDC, Fe3O4 and commercial RuO2 with the same mass loading on GCE are examined. Catalytic performance of Fe3O4/Ni-BDC-n is tested for screening the optimal Fe ratio (Fig. 3 a and Fig. S11). The best OER performance is obtained with the Fe3O4/Ni-BDC-4 (Ni 24.7 %, Fe 75.3 %), exhibiting the lowest η of 295 mV at a current density of 10 mA cm−2. In comparison, large η of 369, 465 and 339 mV (Fig. 3b) is registered for pristine Ni-BDC, Fe3O4 and commercial RuO2, respectively. It is noteworthy that introducing Fe3O4 nanoparticles, although themselves being poor OER catalysts, radically improves the overall water oxidation ability of Ni-BDC, decreasing η with as much as 170 mV. The enhanced OER performance of Fe3O4/Ni-BDC-4 is attributed to the optimal electronic structure of transition metals and hierarchical-layer structure, which are confirmed by XPS and TEM results (Fig. 1c and d, Fig. 2g-i). To illustrate the role of electronic structure change upon OER performance, the η of 337 mV at 10 mA cm−2 of physical mixture of Fe3O4 and Ni-BDC (Fig. S12a) is significantly larger than that of Fe3O4/Ni-BDC-4. High-resolution Ni 2p and Fe 2p spectra (Fig. S12b and c) of physical mixed samples shows no shifts from the individual samples. In situ growth of Ni-BDC in the presence of Fe3O4 nanoparticles causes the binding energy changes of Ni 2p and Fe 2p in the composites (Table S1), optimizing the integral electronic structure of Fe3O4/Ni-BDC-4 composite for high OER performance. Tafel curves of Fe3O4/Ni-BDC-4, Ni-BDC, Fe3O4 and commercial RuO2 are shown in Fig. 3c. Tafel slope of Fe3O4/Ni-BDC-4 (47.8 mV dec-1) is considerably smaller than those of Ni-BDC (60.5 mV dec-1), Fe3O4 (148.1 mV dec-1) and commercial RuO2 (83.5 mV dec-1), revealing the significantly improved catalytic reaction kinetics on Fe3O4/Ni-BDC-4 [66]. The change of Tafel slope is related to the potential-determining step (PDS) of the electrochemical reaction, indicating the PDS of Fe3O4/Ni-BDC-4 is the second step for electron transfer (formation of O*) [40,67]. In addition, stability is also a critical parameter to evaluate the catalyst. The chronopotentiometric curve of Fe3O4/Ni-BDC-4 is shown in Fig. 3d. Compared with previously reported Fe3O4 or Ni-BDC based catalysts (Table S3), Fe3O4/Ni-BDC-4 demonstrates a superior durability over 40 h with a stable OER activity at a constant current density of 10 mA cm−2. As a control, the OER catalytic stability of pure Fe3O4 (Fig. S13a) shows a sharp decay after 8 h, and the pristine Ni-BDC (Fig. S13b) exhibits a weak catalytic stability with a lifetime of less than 5 h. The good catalytic stability of Fe3O4/Ni-BDC-4 implies that the active sites continuously interact with the reaction intermediates for OER during operation. It’s apparent that stability is achieved with the intermixing of Fe3O4 nanoparticles with Ni-BDC. The coupling effects between Fe3O4 nanoparticles and Ni-BDC layers could support the structural stability during OER process by efficiently preventing the aggregation of Ni-BDC. Furthermore, the hierarchical structure of the composite could offer abundant defects and edges.The key parameter involved the number of transferred electrons during OER is further investigated by a RRDE. A much lower current density (Fig. S14) on the ring electrode at 1.50 V compared with that on the disk electrode during OER process was recorded. It indicates a desirable four-electron reaction path (4OH-→ O2 + 2H2O + 4e-) occurs on Fe3O4/Ni-BDC-4 with negligible generation of hydrogen peroxide during OER [52,68]. This further confirms the observed disk current density results from water oxidation rather than other side reactions (Fig. S14). The above results validate that Fe3O4/Ni-BDC-4 is an efficient OER catalyst.Electrochemical behaviour of the samples was characterized by CV in O2-saturated 1.0 M KOH solution in a potential window of 1.123–1.573 V vs. RHE, a region without OER and pre-oxidation peaks are generally observed. Such pre-oxidation peaks are usually related to the oxidation of transition metals (from 2+ to 3+), which are involved in the OER process [25,52]. Ni-BDC (Fig. 4 a) and Fe3O4 (Fig. S15) show the anodic peak potential (E pa) at 1.406 and 1.460 V vs. RHE, respectively [69,70]. The oxidation peak area ratio of Fe3O4/Ni-BDC-4 (Fig. 4a) normalized on the basis of Ni-BDC in CV curves is larger over others synthetic samples, and such an increased oxidation peak area is believed to be significant for enhanced OER catalytic ability [71,72], in good agreement with the above measured η data. As summarized in Table S4 and Fig. 4b, E pa of Fe3O4/Ni-BDC-n has a slight positive shift trend with increasing n (n = 1, 2, 3, 4), but the trend reverses when n = 5, revealing that the amount of Fe3O4 nanoparticles could affect E pa. As a control, the main E pa (1.411 V vs. RHE) in the CV of physical mixture of Fe3O4 and Ni-BDC (Fig. S16) shows a negligible change compared with the E pa of pristine Ni-BDC, furtherly revealing that directly prepared Fe3O4/Ni-BDC composites modify the integral electronic structure for higher oxidation-state situations. With the increasing amount of Fe3O4 nanoparticles in the Fe3O4/Ni-BDC-n (n = 1, 2, 3, 4 and 5), more positive E pa is observed than that of pristine Ni-BDC (Fig. 4a). The strong interaction between Fe3O4 and Ni-BDC suggests that the corresponded E pa peaks from the synergistic effects of oxidation of Ni and Fe species, which shifts positively. The more positive E pa peaks indicate the higher oxidation state of active sites accounting for good OER catalytic performance [25]. Most importantly, there is likely a link between E pa and onset potential of OER (E onset). The optimal Fe3O4/Ni-BDC-4 composite demonstrates the highest E pa (1.431 V vs. RHE) and the lowest E onset (i.e. smallest η) among Fe3O4/Ni-BDC-n. Previous reports [25,73] indicate this could be due to higher-oxidation-state metal originating from the coupling effects of Ni and Fe in the composite is responsible for an enhanced OER performance. Overall, we successfully demonstrate that the modulation of the oxidation state of elemental Ni and Fe of 2D Ni-BDC by incorporating with Fe3O4, which is revealed by XPS, leads to a high E pa and a small η.ECSA is another crucial parameter, which is correlated to the number of active sites and has been determined via C dl measurement (Fig. S17) [32,41,74]. As displayed in Fig. 4c, the C dl of Fe3O4/Ni-BDC-4 is 478 μF cm−2, much higher than those of Fe3O4 (277 μF cm−2) and Ni-BDC (283 μF cm−2). Meanwhile, Fe3O4/Ni-BDC-n (n = 1, 2, 3, 5) exhibit C dl values of 249, 299, 360 and 407 μF cm−2, respectively (Fig. S18). Fe3O4/Ni-BDC-4 shows the highest C dl value, mainly attributed to assumption that the introduction of Fe3O4 nanoparticles on the surface of Ni-BDC layers can lead to the formation of hierarchical structure, offering abundant defects and edges. Besides, the coupling effects between Fe3O4 and Ni-BDC could effectively optimize the electronic structure modulation. These effects are favorable for the improvement of active sites, likely to be related to ECSA. Although ECSA can assess the number of active sites, it is hard to ensure all active sites measured by ECSA are catalytically active [41], we adopted ECSA value for normalizing the current density of LSV in Fig. 3a. Fe3O4/Ni-BDC-4 (Fig. S19) demonstrates the lower η than those of pristine Ni-BDC and Fe3O4 after normalization. The normalized current density of Fe3O4/Ni-BDC-4 is considerably large, for example, reaching 2.6 mA cm−2 at 1.55 V vs. RHE, in comparison to 0.19 and 0.05 mA cm−2 for pristine Ni-BDC and Fe3O4, respectively. This result strongly indicates that the incorporation of Ni-BDC and Fe3O4 effectively promotes the catalytic activity. Further, EIS helps to understand charge transfer kinetics at the electrolyte/electrode interface. Nyquist plots of Fe3O4/Ni-BDC-4, Ni-BDC and Fe3O4 are shown in Fig. 4d. Diameter of the semicircles in high-middle frequency region corresponds to the charge-transfer resistance (Rct) [75]. Rct (10 Ω) of Fe3O4/Ni-BDC-4 during OER is significantly lower than those of pristine Ni-BDC (82 Ω) and Fe3O4 (746 Ω), implying a rapid charge transfer process on Fe3O4/Ni-BDC-4.The morphology and crystalline changes of Fe3O4/Ni-BDC-4 after duration test have been evaluated. XRD pattern of Fe3O4/Ni-BDC-4 (Fig. S20a) after LSV shows the disappearance of Fe3O4 peaks, indicating the possible amorphous transformation of Fe3O4. This observation may be attributed to the oxidation of ultra-small Fe3O4 nanoparticles (6 ± 2 nm) during OER, leading to the formation of oxy-hydroxides. Besides, TEM images (Fig. S20b and c) show iron oxides nanoparticles are still tightly anchored on the surface of Ni-BDC layers. While it is hard to obtain the clear crystalline interplanar of Fe3O4 nanoparticles in the high-resolution TEM image (Fig. S20d), further suggesting the amorphous transformation of Fe3O4 during the OER process. STEM-EDS mapping (Fig. S20e-i) demonstrates the uniform distribution of elemental Ni, Fe, O and C. Meanwhile, XPS results of Fe3O4/Ni-BDC-4 (Fig. S21) after long-term test indicate the partial transformation of metal units in Fe3O4/Ni-BDC-4 to high-oxidation state (Ni3+, Fe3+) due to the partial oxidation during OER process. In comparison to the pristine Fe3O4/Ni-BDC-4, larger satellite peaks in the XPS spectra (Fig. S21) after long-term test are observed, correlated with the oxidation of metal units in Fe3O4/Ni-BDC-4 during OER [27,76].DFT calculations have been performed to uncover the nature of Ni-BDC and Fe3O4/Ni-BDC catalyst and reveal their different performances on the oxygen evolution electrocatalytic process. The (200) surface of Ni-BDC was studied according to the XRD data (Fig. 1a), which is exposed with five-coordinated Ni atoms (Ni5c) and two-coordinated O atoms (O2c) (Fig. S26). The Bader charge analysis shows that the average charge of surface Ni in the Ni-BDC(200) system is + 1.345 |e| (Fig. 5 e). After incorporation with Fe3O4, the XRD results (Fig. 1a and Fig. S1) show the gradual disappearance of the (200) main peak of Ni-BDC and emergence of (311) main peak of Fe3O4 in the Fe3O4/Ni-BDC composites. The Fe3O4(311) surface, therefore, has been considered as the possible representative active surface to investigate in the Fe3O4(311)/Ni-BDC(200) system (Fig. 5b and Fig. S26b). There are five-coordinated Fe atoms (Fe5c), two-coordinated O atoms (O2c) and three-coordinated O atoms (O3c) exposed on the Fe3O4(311) surface (Fig. S26b). Interestingly, the average charge of surface Ni in Fe3O4(311)/Ni-BDC(200) is slightly reduced to + 1.341 |e|. The average charge of surface Fe is + 1.867 |e| in Fe3O4(311)/Ni-BDC(200), higher than that in pristine Fe3O4(311) for + 1.522 |e|(Fig. 5e). The valence states of surface Ni sites are reduced while the surface Fe sites become oxidized on the Fe3O4(311)/Ni-BDC(200) surface. Those agree well with our XPS characterizations that the Ni 2p peak shifts to negative and Fe 2p peak shifts to positive (Fig. 1c and d, Fig. S3 and Table S1). It is safe to conclude that the valence states change of surface Ni and Fe sites are correlated to the improved OER performance. Furthermore, relative to the density of state (DOS) of the pristine Fe3O4 system [51], the density of states of Fe3O4(311)/Ni-BDC(200) system in Fig. 5d shows the Fermi level slightly left-shifting, indicating the electron donator role of the Fe3O4 in the composite system. The electrostatic potential analysis (Fig. S25) illustrates that the electrostatic potential of surface Ni layer is lower than that of Fe3O4 slab, indicating the partial electrons transferring from Fe3O4 to Ni-BDC in Fe3O4(311)/Ni-BDC(200) system, thereby leading a higher oxidation state of surface Fe sites and a reduced oxidation state of surface Ni sites. Therefore, the interaction mechanism of Fe3O4 and Ni-BDC from Fe3O4/Ni-BDC could be proposed. The local electronic environment of Ni nodes in the Ni-BDC is changed after coupling with Fe3O4, partial Ni nodes may be interacted or replaced by Fe3O4 nanoparticles. As BDC ligands are good electron acceptors [30], it suggests that Fe sites in the Fe3O4/Ni-BDC composites provide more electrons with a higher oxidation state in comparison to that of the pure Fe3O4, ensuring the successful formation of composites. Meanwhile, Ni sites in the Fe3O4/Ni-BDC composites will share the extra electrons from the Fe3O4, thus maintaining a reduced oxidation state in comparison to that of pristine Ni-BDC. The optimal electronic structureof Ni and Fe in the Fe3O4/Ni-BDC composites benefits the OER catalytic performance.To further understand the difference in OER electrocatalytic activity of Fe3O4/Ni-BDC and pristine Ni-BDC systems, we adopt the electrochemistry model developed by Nøskov et al. and investigated the thermodynamics of four-electron reactive paths for OER from the free energy landscape (T = 298.15 K) [77,78]. The elementary steps are shown as follows, with * denoted as the catalytic active sites or the adsorbed species: (3.1) ∗ + H 2 O ( l ) → O H ∗ + H + + e - (3.2) O H ∗ → O ∗ + H + + e - (3.3) O ∗ + H 2 O ( l ) → O O H ∗ + H + + e - (3.4) O O H ∗ → O 2 ( g ) + H + + e - Fig. 5a and b show the free energy diagrams of OER on both Ni-BDC(200) and Fe3O4(311)/Ni-BDC(200) surfaces. The free energy diagrams show the step (3.2) referring to the formation of adsorbed O* species is the PDS for both two system (blue line at 0 V in Fig. 5a and b). Specifically, the free energy change of step (3.2) is 1.39 eV at the surface F5c site of the Ni-BDC(200) surface. The overpotential (η) is 0.16 V. In contrast, the free energy change for that step becomes 1.22 eV on the Fe3O4(311)/Ni-BDC(200) surface, 0.17 eV lower than that on the Ni-BDC(200) surface. This implies that the adsorbed O* species are more stabilized at the Fe3O4(311)/Ni-BDC(200) surface, thus enhancing the OER performance. The energy barriers for OH dissociation were also calculated in the two systems. The dissociation barrier of adsorbed OH to generate the adsorbed O is 0.13 eV on the Fe3O4(311)/Ni-BDC(200) surface, lower than that (0.21 eV) on the Ni-BDC(200) surface (see Fig. 5f). OH dissociation is more favourable on the Fe3O4(311)/Ni-BDC(200) surface, consistent with the free energy diagrams in this study. There are also four-coordinated Fe4c sites exposed on the Fe3O4(311)/Ni-BDC(200) surface and was also investigated to compare with Fe5c site. The step (3.1) was found to be the PDS (Fig. S27) with reaction energy of 1.67 eV. It is much higher than that at Fe5c site (0.96 eV) with an overpotential of 1.44 V. Thus the surface Fe5c site is more active towards OER on the Fe3O4(311)/Ni-BDC(200) surface. The Bader charge analysis shows that the adsorbed O* species on the surface Fe5c site possesses a charge of −0.60 |e|, while the charge of adsorbed oxygen at the surface Ni site on the pristine Ni-BDC(200) surface is −0.48 |e|. This indicates the stronger electronic interaction of Fe-O relative to that of Ni-O, stabilizing oxygen adsorption and lowering the free energy for step (3.2) in Fe3O4(311)/Ni-BDC(200) system. By applying a potential of 1.5 V (red lines of Fig. 5a and Fig. 5b), the free energy diagrams for both systems go down, showing the favourable thermodynamics for OER. The step 3.2 still has a lower free energy change in the Fe3O4(311)/Ni-BDC(200) system than the Ni-BDC(200) system (Fig. 5b), consistent with the better OER performance observed in the experiments. These theoretical findings propose the possible structure of Fe3O4/Ni-BDC system and reveal the Fe3O4/Ni-BDC catalyst favours the step of OH dissociation into adsorbed O species that boosts the OER performance. The introduction of balanced amount of Fe3O4 nanoparticles in the composite effectively modulates the electronic structure that lows the potential required for PDS, enhancing OER catalytic activity [25,73]. Besides, the 2D hierarchical-layer structure created by ultrafine Fe3O4 nanoparticles immobilized on 2D Ni-BDC layers provides a large surface area and promotes fast mass transport of the electrolyte to the reactive sites and the liberation of the generated oxygen gas. Meanwhile, the hierarchical composite structure could efficiently prevent the aggregation present in pure Ni-BDC layers that have poor stability (Fig. S13), maintaining the structural integrity during OER for a superior catalytic stability. The detailed XPS binding energy results (Table S1) differ from those in pristine Ni-BDC and Fe3O4, i.e. binding energy level of Ni 2p3/2 decreases, while the binding energy of Fe 2p3/2 increases slightly with the amount of Fe3O4 nanoparticles. This indicates that the Ni atoms in the composites possess higher electron densities than that of pristine Ni-BDC with the increasing amount of Fe3O4 nanoparticles. While Fe atoms in the composites have lower electron densities comparing with that of pure Fe3O4 and then trend to the stable electron densities. Initially, when the mass ratios of Fe3O4 in the composites are relatively low (Fe3O4/Ni-BDC-1, 2 and 3). The electronic densities of Fe and Ni atoms in the composites are tuned, offering improved OER performance. Further, Fe3O4/Ni-BDC-4 possesses the optimal electronic structure of Fe and Ni atoms, showing the best OER catalytic activity among five composites. When the Fe3O4 nanoparticles are further added to form Fe3O4/Ni-BDC-5, the electronic densities of Fe and Ni atoms in the composite is hardly changed in comparison to that of Fe3O4/Ni-BDC-4. Reversely, the further addition of Fe3O4 limits the OER performance with a reduced catalytic activity. All above results ensure that Fe3O4/Ni-BDC-4 could be a promising and stable OER electrocatalyst.Ultrafine Fe3O4 nanoparticles homogeneously immobilized on 2D Ni based MOFs (Fe3O4/Ni-BDC) were synthesized. The functionalized Fe3O4 nanoparticles (Ø 6 ± 2 nm) with abundant surface hydroxide groups are produced by a hydrothermal method, and then mixed into 2D Ni-BDC layers during synthesis (thickness: 4 ± 1 nm) creating strong interactions, which are not achieved by physically mixing the two components. Introduction of Fe3O4 modifies the integral electronic structure for reduced overpotential and prevents the aggregation of 2D Ni-BDC layers for enhanced OER catalytic stability. Different atom ratios of (Ni/Fe) in Fe3O4/Ni-BDC are tested for OER. Fe3O4/Ni-BDC-4 demonstrates the optimized OER performance with an η of 295 mV at 10 mA cm−2, a Tafel slope of 47.8 mV dec-1 and superior catalytic durability (40 h). DFT calculations further identify the active sites for Fe3O4/Ni-BDC as mainly contributed by Fe species with a higher oxidation state, and the PDS is the formation of the adsorbed O* species, which are facilitated in the Fe rich composite. The persistent stability during cycling (Fig. 3d) and the TEM images show that agglomeration is not occurring, indicating that this typically performance reducing effect can be handled. Such structure design methodologies for electronic structure and adsorbate modulation will inspire further development of promising catalysts for OER.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.Finance support from the Chinese Scholarship Council (201706220080) for W.H., the Natural Science Foundation of Hunan Province (2019JJ50526) for C. P., The Danish Council for Independent Research for the YDUN project (DFF 4093-00297) to J.Z., Villum Experiment (grant No. 35844) for X. X. is greatly acknowledged.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jechem.2021.05.030.The following are the Supplementary data to this article: Supplementary data 1

Two-dimensional (2D) metal organic frameworks (MOFs) are emerging as low-cost oxygen evolution reaction (OER) electrocatalysts, however, suffering aggregation and poor operation stability. Herein, ultrafine Fe3O4 nanoparticles (diameter: 6 ± 2 nm) are homogeneously immobilized on 2D Ni based MOFs (Ni-BDC, thickness: 5 ± 1 nm) to improve the OER stability. Electronic structure modulation for enhanced catalytic activity is studied via adjusting the amount of Fe3O4 nanoparticles on Ni-BDC. The optimal Fe3O4/Ni-BDC achieves the best OER performance with an overpotential of 295 mV at 10 mA cm−2, a Tafel slope of 47.8 mV dec-1 and a considerable catalytic durability of more than 40 h (less than 5 h for Ni-BDC alone). DFT calculations confirm that the active sites for Fe3O4/Ni-BDC are mainly contributed by Fe species with a higher oxidation state, and the potential-determining step (PDS) is the formation of the adsorbed O* species, which are facilitated in the composite.

Data will be made available on request.Schiff base ligands, which can form stable coordination bonds with metal ions, have received considerable attention in recent years [1]. They have played an important role in the development of coordination chemistry due to their preparative accessibility and structural variety [2]. The chemistry of coordination compounds containing metal-nitrogen bonds has been of particular interest to researchers in recent years due to the extraordinary properties of many complexes [3]. Indeed, N4 and N2O2 [4,5] M(II) nitrogen ligands have been used as catalysts in the reduction of organic substrates with nitro, olefinic, acetylenic and aldehyde groups under mild reaction conditions, as well as catalysts in the reduction of nitrogen. Other studies have shown that transition metal complexes based on Schiff NNNN base ligands function as excellent homogeneous and heterogeneous phase catalysts [6], precursors of electrocatalytic processes [7] and chemical sensors [8]. In particular, Schiff-based ligands are of considerable interest due to their structural properties of being potential models for several biological systems due to the chelation property of the ligands with most proteins [9]. These compounds also present biological, anticancer, antibacterial, antifungal activities. On the other hand, these ligands have an interesting application in the field of corrosion. Thus, in our laboratory, studies have shown that the L1 and L2 ligands are corrosion inhibitors of carbon steel in a HCl solution [10,11]. Indeed we have extended this study on the chelating behavior and the antimicrobial and antioxidant power of Cu(II),Ni(II),Co(II) and Zn(II) complexes based on the tetradentate ligand N4 [(N1Z, N2Z)-N1,N2-bis((1H-pyrrol-2-yl)methylene)ethane-1,2-diamine] (L) derived from 1H-pyrrole-2-carbaldehyde with ethylenediamine and to compare this potency with that of the free ligand. The structural study of the complexes was performed by chemical ionization mass spectrometry, UV–visible spectroscopy, Fourier transform infrared spectroscopy (FT-IR) and nuclear magnetic resonance (NMR, 1H,13C). Schiff's base and its corresponding complexes were examined for their antibacterial and antioxidant activities against Gram (+) and Gram (-) bacteria, for their antioxidant activity against DPPH radical for their antimitotic and hemolytic biological activity.All chemicals used in this work were analytical reagent (AR) grade and of the highest purity. The reagents for the synthesis of the Schiff base ligand (L) were purchased from Alfa Aesar. Melting points were recorded in open capillaries by a Stuart Melting Point apparatus, SMP10. IR spectra of the compounds were recorded by FT-IR Fourier transform infrared FTIR TENSOR27 spectrometer using KBr pellets. The electronic absorption spectra were obtained by a Perkin-Elmer Lambda 35 UV–vis spectrophotometer. Magnetic susceptibility of the metal complexes in the solid state was measured by the Gouy balance calibrated with mercuric tetrathiocyanatocobaltate (II). 1H NMR spectra of Schiff bases in DMSO‑d 6 were recorded on a JEOL 500 MHz FT NMR System JNM-ECZ500R/S1 spectrometer. Thermal analyses (DTA and TG) were performed on a DTG-60H thermal analyzer from temperature 20 °C to 1000 °C at a heating rate of 20 °C/min. High resolution mass spectra (HRMS) were acquired by the electron boiling ionization (ESI) technique using a Bruker APEX-2.Ligand (L) was prepared by condensation of (1 g; 2 mmoL) 1H-pyrrole-2-carbaldehyde and ethylenediamine (0.31 g; 1 mmoL) in ethanol (30 mL) at reflux for 3 h. The progress of the reaction was monitored by TLC. The resulting pink crystalline precipitate was washed with ethanol and ether and dried under vacuum (Scheme 1 ).(L):R: 89 %; m.p. 182 °C; IR (KBr, cm−1) ν: 3157(NH) 3088, 2940–2836C-Haleph, 1577 (CC), 1641 (NC), 1315(CN), 1315; 1H NMR (500 MHz,DMSO‑d 6,ppm) δ: 11,39 (1H, s, NH), 8.02 (1H,s,H-CN), 6.81–6.80 (1H,d,J = 5HZ), 6.38–6.37 (1H,m,Pyrrol) 6.05–6.04 (1H,m,Pyrrol) 2.46–2.45 (2H,t,J = 5HZ, ethylene); 13C NMR (125 MHz, DMSO) δ:153,254 (CN), 130.48–109.34 (Pyrrol-C), 61.97 (CH2– ethylene); MS [m/z]+=215.12;UV–vis (DMSO): max (nm) = 327,95; Solubility: DMSO, MeOH, EtOH, DMF, CHCl3.A solution of MCl2 (M = Zn, Cu, Ni and Co) dissolved in 15 mL of ethanol was added dropwise to an ethanolic solution (30 mL) of ligand L under stirring and reflux for 4 h. The precipitates obtained were filtered and purified by washing with ethanol and then dried under vacuum (Scheme 2 ).ZnL: solid black; R: 63 %; m.p °C:309.12; MS; conductivityΛ(Scm2mole-1): 7; IR (KBr, cm−1) ν: 2929–2864 (CH), 1602 (NC), 1559 (CC), 1286 (CN), 584 (M−N); 1H NMR (500 MHz, DMSO‑d 6, ppm) δ:9.42 (1H,s,H-CN), 7.17–6.23 (3H,m,Pyrrol), 2,452 (2H,ethylene); μeff (BM): dia; [m/z]+=277.04; UV–vis (DMSO): max (nm) = 330, 524; Solubility: DMSO, DMF, CHCl3.[CuL]Cl22H2O:solid black; Yield: 62 %; m.p°: 279,34; Molar conductivityΛ (Scm2mole-1):57; Analysis Calculated for; IR (KBr, cm−1) ν:3450 (H2O), 3302–3232(NH), 2937–2881(CH), 1635 (NC),1568 (CC), 1278 (CN), 528 (M−N); 1H NMR (500 MHz, DMSO‑d 6,ppm) δ: 9.40 (1H,s,H-CN), 7.15–6.21 (3H,m,Pyrrol), 2.24 (2H,ethylene) 13C NMR (125 MHz, DMSO) δ: (CN) 179.69, 140.29–111,10 (Pyrrol-C) 59.92 (CH2-ethylene); μeff (BM): dia;MS [m/z]+=278.07; UV–vis (DMSO): max (nm) = 330,400; Solubility: DMSO, DMF,CHCl3.[CoLCl2]H2O:Solid brown; Yield: 68 %; m.p:283,93; Molar conductivityΛ (S cm2mole-1): 35.,9; Analysis Calculated for; IR (KBr, cm−1) ν:3551 (H2O), 3216–3128 (NH),(CH)2959,1624 (NC), 1578 (CC), 1302 (CN), 525(M−N); μeff (BM):4.84 MS [m/z]+=344.01; UV–vis (DMSO): max (nm) = 365,640,708; Solubility: DMSO, DMF, THF.NiL: solid orange; Yield: 58 %; m.p: 345.12; Molar conductivityΛ(S cm2mole-1): 4.2; Analysis Calculated for; IR (KBr, cm−1) ν:2844–2915C-Haromatic, 1574 (NC), 1524(CC), 1303 (CN), 530 (M−N); 1H NMR (500 MHz,DMSO‑d 6,ppm) δ: 7.57 (1H,s,H-CN), 6.68–5.89 (3H,m,Pyrrol), 2.45 (2H,ethylene) 13C NMR (125 MHz, DMSO) δ: (CN) 160.63, 143.56–111.30 (Pyrrol-C) 55.89 (CH2-ethylene); μeff (BM): dia; MS [m/z]+=271,01; UV–vis (DMSO): max (nm) = 357,429,454,522; Solubility: DMSO, DMF, CHCl3.The 2,2-diphenyl-1-picrylhydrazyl (DPPH) is a stable free radical that accepts an electron or hydrogen radical to become a stable diamagnetic molecule [12] which is a stable scavenging activity in chemical analysis [13]. The antioxidant capacity of the tested products was estimated by comparison with a synthesized antioxidant (ascorbic acid). DPPH radical reduction capacity was determined by the decrease in its absorbance at 517 nm induced by the antioxidants. The absorption maximum of a stable DPPH radical in ethanol was at 517 nm. By using this method, it is possible to determine the free radical scavenging capacity of an antioxidant by measuring the decrease in absorbance of DPPH at 517 nm. Resulting from a color change from purple to yellow, the absorbance decreases when DPPH is trapped by an antioxidant, by hydrogen donation to form a stable DPPH molecule. The DPPH radical scavenging activity of the complexes was measured as indicated according to the protocol described by Lopes-Lutz et al [14]: each 1 mL of solution of the ligand and its complexes at different concentrations prepared in DMF, 2.5 mL of DPPH solution (2.4 mg/100 mL) was added. At the same time, a negative control was prepared by mixing 2.5 mL of DPPH with 1 mL of the methanol solution. After incubation in the dark for 30 min and at room temperature. Absorbance readings are taken at 517 nm using a spectrophotometer against a blank (1 mL of products at different concentrations + 2.5 mL of methanol). The ability to trap the DPPH radical was calculated using the following equation (1) % R S A = Ac - A s Ac × 100 Ac: absorbance of the control (DPPH solution in the absence of the tested compound).As: absorbance in the presence of the tested compound.The microdilution test was used to determine the Minimum Inhibitory Concentration (MIC), in a96-well microplate. The ligand solution and its complexes were examined against two Gram-positive bacteria: Bacillus subtilis ILP1428B and Staphylococcus aureus CIP543154 (Pasteur Institute Collection), as well as two Gram-negative bacteria: Pseudomonas aeruginosaATCC27653 and Escherichia coli CIP5412 (American Type Culture Collection). Mueller HintonBroth was supplemented with the emulsifier (1 % (v/v) DMSO). Then, 50 μL of bacterial (106 CFU/mL) was deposed. Finally, bacterial growth was revealed by turning resazurin from purple to pink. The lowest inhibitory concentration of the ligand solution and its complexes corresponded to the lowest concentration that inhibited the reduction of blue resazurin dye into pink resorufin. MBC was determined by sub-culturing the contents of wells with greater concentrations than the MIC values on LB agar plates and incubating them at 37 °C for further 24 h. Experiments were repeated three times [15].The species used in this work is the Alenois Cress (Lepidium sativum). Seeds are washed with pure distilled water to eliminate any impurity and are tested for a germination rate higher than 95 % [16]. Seeds are sown in 50 mm diameter Petri dishes, lined with a layer of Whatman type filter paper, impregnated by 5 mL of aqueous solution of each ligand at a concentration of 1 mg/mL. The control box is impregnated by 5 mL of distilled water. Hydration is done once at the beginning of the test. All plates are placed in the oven at 25 °C and in the dark. The germination process and elongation of the seed radicles are observed directly in the Petri dishes every 24 h for one week. Three control replicates and three replicates of each of the products are used in Petri dishes containing 20 seeds each of average size i.e. (3 × 20) seeds for each test. Data are expressed as the average radicle elongation and results are reported in mm.Calculation and expression of germination results. • Determination of the germination capacity of seeds of Lepidium sativum: Seed germination capacity is determined by calculating the seed germination rate expressed as a percentage: Number o f s e e d s g e r m i n a t e d / T o t a l n u m b e r o f s e e d s × 100 [Number of seeds germinated/Total number of seeds] × 100 • Determination of the germination inhibition index (GI): Determination of the germination capacity of seeds of Lepidium sativum:Determination of the germination inhibition index (GI):This method is developed to determine the phytotoxicity of the ligand and its synthesized complexes. The monitoring of seed germination is determined every 24 h. The number of germinated seeds is noted and the percentage of germination inhibition is calculated as follows: IG % = PGte - P G t r / P G t e × 100 PGte: germination percentage of the control lot.PGtr: percentage of germination of the batch treated with the synthesized product. • Vigor of Lepidium sativum seedlings Vigor of Lepidium sativum seedlingsAfter determining the germination rate for seven days for each of the replicas, the length of the radicle was measured. This quantity is expressed as the average of the radicle elongation and the results are reported in mm. Seedling v i g o r = G e r m i n a t i o n p e r c e n t a g e × s e e d l i n g l e n g t h Fresh human blood was collected in a sterile tube containing sodium citrate as an anticoagulant at a ratio of one volume to four volumes of blood. After decantation, the pellet was washed four times with sterile physiological water. In hemolysis tubes, 50 µL of the pellet is added to 1 mL of sterile physiological water mixed with 1 mg/ml of the component to be tested. The control tube is devoid of product. The preparations are then incubated in the dark for 1 h at room temperature and centrifuged at 1500 rpm for 3 min. Observation of hemolysis is performed directly with the naked eye. Each test is repeated three times.In test tubes, 1 mL of fresh human blood, collected on EDTA, is mixed with 1 mg of the test product. The control tube is devoid of product. The preparations are incubated in the dark for 1 h at room temperature. Thin smears are made, after homogenization of the blood, on clean slides. After drying, the smears are subjected to the May-Grünwald-Giemsa (MGG) staining. An observation is made under the optical microscope at a magnification of 1000x to illustrate the behavior of leukocytes towards the test products. Each experiment is repeated three times.All the synthesized metal complexes are solids and are in the form of colored powder. These products are stable in air and soluble in dimthylformamide (DMF) and dimethylsulfoxide (DMSO). The main physicochemical properties of the ligand (L) and the corresponding complexes are summarized in Table 1 . Elemental analysis as well as FT-IR and UV–vis measurements demonstrate that all the obtained complexes have a metal–ligand stoichiometry (1:1) and present a composition in good agreement with the proposed formulas. The chelates were found to be stable under atmospheric conditions, whether stored in solution or as pure solids.The results of the conductimetric study of the 10-3 M concentration solutions in DMF of the ligand and its complexes are recorded in Table 1. The molar conductivities of the freshly prepared solutions of the complexes are between 4.2 and 57 Scm2mol−1.These values being low show that the complexes are not electrolytes [17]. The molar conductivity site of [CuL]Cl22H2O complex of 57 Scm2mol−1, respectively, referring to the electrolytic behavior of this complex. This result is proposed with the studies of Greenwood [18].The electronic spectra of the ligand (L) and its complexes (Fig. 1 , Table 2 ) were recorded in DMF solutions (10-5 M) at room temperature. In the UV–vis spectra of the parent ligand, an intense absorption band at about 327.95 nm has the n → π* transition of the azomethine group. This band underwent a bathochromic shift for the complexes (Table 2), suggesting that Cu(II), Ni(II), Zn(II) and Co(II) are indeed coordinated [19]. Furthermore the electronic spectrum of the Co(II) complex shows the presence of two d-d bands at 640 and 708 nm attributed to the 4T1g→4A2g et 4T1g→4T1g(P) transitions respectively which suggests an octahedral geometry of the cobalt [20,21]. The value of magnetic moment for cobalt (II) complex was 4.84B.M. which was consistent with high spin octahedral geometry for Co(II) complex [22]. The electronic spectrum of complex Ni (II) shows two spin-allowed bands at. These absorption bands may be assigned to the 1A1g→1A2g and 1A1g→1Eg transitions, respectively, and reflect d8 ions in a square planar geometrical environment [23]. As for the copper complex the presence of an absorption band at 400 nm corresponds to the2Eg ←2B1gtransition indicates the square planar geometry for Cu(II) complexes which is in perfect agreement with the literature data [21,24]. For the zinc complex, it does not show a d-d electronic transition due to the completely daughter d10 orbital. The appearance of the new transition absorption band at 524 nm is attributed to ligand–metal charge transfer (LMCT) [20]. The complexes of Cu (II), Zn (II), and Ni (II) were also subjected to magnetic susceptibility test. They were diamagnetic, and supported by distinct signals in the 1H NMR spectra [21].IR spectra provide valuable information regarding the nature of the functional group. These studies are of great importance in the evaluation of important characteristic frequencies necessary for the comparative interpretation of the nature of the binding of Schiff base ligands and their metal complexes, molecular symmetry, electron distribution, and stability of the complexes formed. The main IR bands of the free ligand and their metal complexes are listed in Table 3 .The IR spectra of the ligand and the Cu(II) and Co(II) complexes (Fig. 2 ) showed the presence of a characteristic band due to ν(NH) around 3157–3088 cm−1, for the ligand and around [(3216–3128);(3302–3232) cm−1] for the Co(II)and Cu(II) complexes [25] the absence of this band for the zinc-nickel complex indicates deportation of the pyrolic nitrogen after coordination. An absorption band is systematically observed in the range 2862–2940 cm−1 for either the ligand or the complexes. This band is due to the deformation absorption of aliphatic CH The intense absorption band characteristic of the stretching vibrations of the azomethine group (-CN-) was observed at 1641 cm−1 in the spectrum of the Schiff base ligand. However, the band due to -CN- shifted to lower wavenumbers (1574–1636 cm−1) in the metal complex spectra [24]. This hypsochromic shift of the order of 5–67 cm−1 indicating the participation of the imine nitrogen in coordination. Also the vibrations of the double bond (CC) of the aromatic nuclei are characterized and maintained by a relatively intense and sharp band at 1577.97 cm−1 in the case of the ligand L and is shifted from about 1 to 53 cm−1 for the complexes of this same ligand [26,29]. In addition, new bands appear from 522 to 584 cm−1 in the spectra of the metal complexes which are attributed to ν(M−N) [27], confirming the coordination of copper(II), cobalt(II) zinc(II) and nickel(II) with the ligand [26]. After complexation, of the ligand with the copper and cobalt metal salts, a broad absorption band was observed around 3450–3551 cm−1 which is attributed to the presence of water of hydration molecules in the crystal lattice of Cu(II) and Co(II) complexes. Table 4 shows the 1H NMR spectral data of the ligand (L) (Scheme3 ) and the corresponding Ni(II), Cu(II), Co(II) and Zn(II) complexes.The 1H NMR spectrum of the Schiff base (L) (Fig. S1), shows a singlet at 11.34 ppm which corresponds to the proton NH1 [31].The signal at 8.018 ppm is attributed to the azomethine protons (-H5C = N) In the region 6.81 and 6.036 ppm, multiplets were observed, and can be attributed to the protons of the pyrrolic rings. The methylene protons (N-(CH2)6) appear at 2.46 ppm. Concerning the spectra of the Zn(II) and Ni(II) complexes (Figs. S2 and S3), also the absence of the NH proton of the pyrolic ring was observed thus suggesting the involvement of the ring nitrogen in the coordination with the metal ion. A shift of the signal of the azomethine protons (H5C = N) of the complexes with respect to those of the free ligand towards the strong fields [(9.42; 9.41; 7.57) ppm] for the Zn(II), Cu(II) and Ni(II) complexes which confirms the coordination of the azomethine group with the metal ion (data not shown).On the other hand, it is observed after coordination that the aromatic protons of the pyrrolic ring appear in the regions [(7.17 ppm-6.23 ppm); (7.15 ppm-6.21 ppm); (6.67–5.89 ppm)] for Zinc, Copper and Nickel. The aliphatic protons (CH2) of the complexes were detected at [2.45 ppm, 2.24 ppm and 2.45 ppm] for Zinc, Copper and Nickel (Table 4).The 13C NMR spectral data of the base ligand of Schiff L, shows that the peak appearing at 153.25 ppm is assignable to the imine carbon atoms. On the other hand, the signals observed in the region 130.48–109.37 ppm are assigned to the signals of the pyrrolic ring carbons of the ligand. The aliphatic N-CH2 carbon peaks of the ligand was detected at 61.97 ppm (Fig. S4). The same signals are present in the 13C NMR spectra of the complexes (Figs. S5 and S6), but they are shifted to higher values invoking the coordination of the ligand to the metal ions by its azomethine group. The signals of the pyrrolic ring carbons appear in the regions [(140.29–111.10); (143.56–111.30) ppm] for the Cu(II) and Ni(II) complexes respectively. The aliphatic carbon (CH2) peaks are detected at 59.92 and 55.89 ppm for the Cu(II) and Ni(II) complexes (Table 5 ).The mass spectra of the ligand and its complexes were recorded on HPLC-MS equipment using the electrostatic spray ionization (ESI) technique. The mass spectrum of ligand (L) revealed a molecular ion peak at [m/z] = 215 which confirms the proposed formula. The mass spectra of the Co(II), Ni(II), Cu(II) and Zn(II) complexes, were recorded and all spectra show peaks at molecular ions (M+). The proposed molecular formula of these complexes was confirmed by comparing their molecular formula weights with the m/z values. The molecular ion (M+) peaks obtained for the different complexes are (1) m/z = 344 (Cobalt(II) complex), (2) m/z = 271 (Nickel(II) complex) (3) m/z = 278 (Copper(II) complex) (4) m/z = 277 (Zinc(II) complex). These data are in good agreement with the proposed molecular formula for these complexes, i.e. [MLCl2] respectively [where M = Co(II), Ni(II)], and [ML] [where M = Cu(II) and Zn(II)], L = ligand. This confirms the formation of the Schiff base complex framework.The thermal behavior of transition metal complexes was studied to establish the decomposition process. To perform the measurements of this analysis, the temperature was increased from 20 °C to 1000 °C at a rate of 20C min−1 under air atmosphere. The thermograms obtained are shown in Figs. S7–S10 and the thermal data of the complexes are shown in Table 6 .The ATG curve indicates that the cobalt complex is decomposed in four main steps, the first is observed in the temperature range of 39.5–200C° with a weight loss of 4.67 %, associated with the loss of one molecule of water of hydration [28]. The DTA curve coincides exactly with the observed weight loss and presents a first endothermic reaction at 100.04 °C. The second decomposition step in the temperature range 201.09–290.05 °C corresponds to the loss of a chlorine atom, with a mass loss of 9.28 % [29]. his is also highlighted by an exothermic peak at 288.03 °C. The third step c was attributed to the elimination of the organic part C12H12N4 with a weight loss of 51.77 accompanied by an exothermic peak observed at 478.5 °C. The fourth decomposition step in the temperature range 556.72–616.92C corresponds to the loss of 9.33 % and was assigned to the elimination of a second chlorine atom in the presence of the ATD peak at 605.18 °C to finally give CoO as the residue. The thermogram of the Zn(II) complex showed a single decomposition step in the temperature range 296.21–716.2 °C corresponds to the loss of one molecule of the ligand C12H12N4, with a mass loss of 79.14 %. The exothermic change between 306.73 and 569.71 °C in the presence of the ATD peak at 568.1 °C is related to the decomposition of the ligand. The TG plot of the Ni(II) complex showed a decomposition pattern with two well-defined decomposition steps The first decomposition step in the temperature range 237–439 °C corresponds to the loss of C4H3N molecule with a mass loss of 18.83 % (Calc. 19.3 %) accompanied by an exothermic peak at 347.53 °C. The second decomposition step corresponds to the loss of one molecule C8H9N3 with a mass loss of 53.95 %, and this is confirmed by the presence of an exothermic peak at 496.86 °C on the DTA curve. The final product of the decomposition of the Ni(II) complex corresponds to nickel oxide. The first stage of decomposition of the Cu(II) complex is accompanied by a loss of mass of 11.93 % attributed to the departure of two molecules of water of hydration this is confirmed by the presence of an endothermic peak at 100.67 °C on the DTA curve. The second decomposition step in the temperature range 227.47–525.47 °C gives a loss of mass of 67.26 % corresponding to the loss of C12H12N4 accompanied by an endothermic peak at 456.6 °C.The antioxidant activity of ligand (L) and its complexes Ni(II), Co(II), Zn(II) and Cu(II) was evaluated using the DPPH radical assay (Table 7 ) [30–32]. In the present study the analysis of the results obtained showed that the DPPH free radical scavenging activity of the complexes is higher than that of the free ligand. Zinc and copper complexes showed, for low concentrations, a very high activity compared to the ligand, as well as compared to nickel and cobalt complexes. This difference in activity can be attributed to the coordination environment and the redox properties which depend on several factors such as the size of the chelate ring, the axial ligation and the degree of unsaturation in the chelate ring [33]. Moreover, the low antioxidant activity of the Co(II) and Ni(II) complexes compared to the Cu(II) and Zn(II) complex could be due to the steric hindrance induced by the geometrical structure, preventing the approach of the DPPH radical towards the active centers of the complex [34]. The antioxidant activity of all the complexes studied is significantly higher than that of ascorbic acid. We can conclude that the complexation of this ligand promoted the antioxidant power which can be translated due to the electron withdrawing effect of M(II) ion (M = Zn(II),Cu(II),CoII) and Ni(II)) facilitates the release of hydrogen as a free radical in the presence of DPPH [35,36].The results of antibacterial effect are presented in Table 8 . A careful observation of the results indicated a moderate antibacterial power against all tested strains. In addition, [CoLCl2]H2O and ZnL were more efficient against S. aureus, whichvalue of MIC 0.312 mg/mL, which in accordance with the results obtained by Kargar et al [31]. Furthermore, similar studies were performed to assess whether ligand and its complexes also inhibited α-amylase, as a carbohydrate-hydrolyzing enzyme. The results revealed that Zn possessed the highest inhibitory activity as compared to other complexes, whereas ligand had the lowest inhibitory effect [37]. On the other hand, B. subtilis and E. coli were more sensitive to [CoLCl2]H2Owith MIC value (0.019 mg/mL). These results showed that the chelation of the ligand increase the antibacterial effect, which was also confirmed by Polo-Cerón [38]. It might be related to a modification in cation polarity caused by the hybridization of ligand filled orbitals with empty “d” orbitals of the metals. Generally, ligands and their respective metal complexes are moderately more potent inhibitors of Gram-positive rather than Gram-negative bacterial strains. This is due to the fact that Gram-positive bacteria have a thick peptidoglycan layer but no outside lipid membrane, whereas Gram-negative bacteria have a thin peptidoglycan layer but an outer lipid membrane [39]. • Germination capacity of Lepidium sativum seeds: Germination capacity of Lepidium sativum seeds: The germination capacity of Lepidium sativum seeds after hydration with distilled water in the control batches was 95 % (Fig. 3 ).After treatment with the different products at a concentration of 1 mg/mL the germination percentages, after seven days of incubation, ranged from 0 % observed in the [CuL]Cl2H2O ligand batch to 60 % for that of [CoLCl2]H2O and ZnL (Fig. 4 ). • Germination inhibition capacity of Lepidium sativum seeds Germination inhibition capacity of Lepidium sativum seeds Fig. S11 shows the inhibition rate of the ligand and its complexes during the seven days of treatment and thus confirms the inhibitory effect of the five products from the first day of incubation. The results again show that [CuL]2H2O exerts a complete inhibition of germination on all seeds (100 % inhibition rate). The inhibition rate of the other components is 78 % for the L ligand, 57 % for the NiL products and 36 % for the [CoLCl2]H2O and [ZnL]Cl2. • Vigor of Lepidium sativum plantlet Vigor of Lepidium sativum plantlet The vigor of Lepidium sativum seedlings gives us information on the mitotic capacity of the radicles in the presence of the synthesized products compared to the control test, the greater their length the greater the speed of growth and multiplication of the cells and vice versa. Fig. 5 shows that the length of Lepidium sativum radicles after 7 days of treatment is decreased by 100 % in the presence of [CuL]Cl22H2O, 91 % for ligand L, 66 % for ZnL and 86 % for [CoLCl2]H2O andNiL.These results of phytotoxicity tests indicate that the ligand and its corresponding metal complexes exhibit antimitotic activity with a higher capacity of this activity for [CuL]Cl22H2O and less effective for ZnL. These results are in agreement with those found by Abdou Saad El-Tabl et all whose copper-containing metal complex had antitumor activity on hepatic carcinoma cells (Hep-G2) [40], and also with the study of Tudor et all which showed that copper-based complexes showed an antimitotic effect on cervical cancer cell lines “HeLa”[35], as well as the study of Ceyhan et all who tested their copper complexes on the ovarian cancer cell line “A2780″, and proved that their synthetic substances induced a loss of deviability of the tested tumor cells [36]. • Hemolysis test Hemolysis test This test is carried out in order to observe the in vitro effect of the ligand and its synthesized complexes on the behavior of human red blood cells in order to show the presence or absence of toxicity of these different tested components. Blood cells have a vital and important role on human health. They are very sensitive to changes in the composition and osmotic pressure of the surrounding environment. The results obtained show a hemolytic effect of [CoLCl2]H2O complexes following a lysis of the red blood cells releasing the hemoglobin, the tube presents a red pigmented supernatant (Figure S12). However, the other components have no hemolytic effect, the supernatant remains clear and therefore retains its hemoglobin content. • Effect of the ligand and its corresponding complexes on leukocytes Effect of the ligand and its corresponding complexes on leukocytes This test is carried out with the aim of concretizing the in vitro effect of the ligand and its synthesized complexes on the leukocytes which are the nucleated cells and whose role is important on the defense and maintenance of the immune system in man. Fig. 6 illustrates the observation under the optical microscope of blood smears made from human blood treated with the different synthesized products. The control smear is made from fresh untreated blood. The results confirm the hemolytic effect of the [CoLCl2]H2O component, the red blood cells appear pale and emptied of their hemoglobin and the leukocytes are totally deformed, altered and have a defragmented nucleus. This complex acts not only on the integrity of the cell membranes but also on the genetic material of the cell, which confers to this product a genotoxic and lytic character for the blood cells. Microscopic observation of other smears taken from blood treated with the different products (L, [CuL]Cl22H2O, ZnL and NiL) showed no destructive effect on blood components. This tolerance test of eukaryotic and human cells towards the products synthesized for this study shows the destructive and thus toxic effect of complex [CoLCl2]H2O.The synthesis and structural study of the ligand [(N1Z,N2Z)-N1,N2-bis((1H-pyrrol-2-yl)methylene)ethane-1,2-diamine] (L) and its corresponding complexes of Cu(II), Co(II) and Zn(II) were carried out by IR, UV–Visible, (1H,13C)NMR and Differential Thermal Analysis (DTA, TGA) spectrographic techniques and mass spectrometry. These last ones with the study of the conductivity showed that these complexes are not electrolytes on the one hand, and on the other hand that the ligand is coordinated to the metal by the four nitrogen atoms. The electronic spectral data of the complexes suggest an octahedral geometry for the cobalt and nickel complexes, tetrahedral for the zinc complex and square plane for the copper complex. The antioxidant activity of the ligand and its metal complexes was studied by DPPH free radical scavenging methods and showed that the ligand and its complexes have very good radical scavenging activity compared to standard ascorbic acid and that the zinc (II) complex has a higher antioxidant activity with an IC50 = 0.0047 g/mL. Biological studies of these complexes show better activity compared to the ligand and the Zn(II) and Co(II) complexes have good activity against all bacterial strains. The toxicity study concludes that the ligand and its complexes have significant biological activity and inhibitory power on promising use in cancer treatment after further study on animal cells. Ibtissam Elaaraj: Conceptualization, Methodology, Validation, Investigation, Data curation, Writing – original draft. Najia Moukrad: Methodology, Validation, Investigation, Data curation. Aziz Bouymajane: Investigation, Writing – review & editing. Safae Er Raouan: Investigation. Asmae Nakkabi: Investigation. Oumayma Oulidi: Investigation. Fouzia Rhazi Filai: Conceptualization, Resources, Supervision, Funding acquisition. Ibnsouda Koraichi Saad: Investigation. Francesco Cacciola: Resources, Writing – review & editing, Supervision, Project administration. Noureddine El Moualij: Investigation. Mohammed Fahim: Conceptualization, Methodology, Validation, Data curation, 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.Supplementary data to this article can be found online at https://doi.org/10.1016/j.rechem.2023.100787.The following are the Supplementary data to this article: Supplementary data 1

In this contribution new Cobalt(II), Nickel(II), Copper(II) and Zinc(II) complexes based on the tetradentate ligand [(N1Z,N2Z)-N1,N2-bis((1H-pyrrol-2-yl)methylene)ethane-1,2-diamine] were synthesized and characterized by spectral techniques such as UV–Visible, infrared, nuclear magnetic resonance (NMR, 1H,13C), thermal analysis (DTA and TGA) and mass spectrometry. Results showed that Co (II) and Ni (II) complexes have an octahedral geometry, Zn (II) complex has a tetrahedral structure, and Copper (II) complex has a square planar geometry. Furthermore, based on the molar conductivity values, these complexes were considered as non-electrolytes except the copper complex. The synthesized ligand and their complexes were assessed for their antioxidant and antibacterial activities. All complexes showed greater antioxidant and antibacterial activities with respect to the ligand. In addition, the ligand and its complexes were also tested biologically first on the germination of Lepidium sativum seeds, which is a phytotoxic and antimitotic test, and on the hemolytic and genotoxic behavior of human blood cells. A total inhibition on germination and radicle growth of Lepidium sativum seeds treated with [CuL]Cl22H2O solution was recorded. However, the [CoLCl2]H2O, caused a complete hemolysis of red blood cells and total alteration of nuclei and membranes of leukocytes.

Air pollution contamination resulting from volatile organic compounds (VOCs) is an internationally researched topic (Andelman, 1985; Harkov et al., 1985; Abumaizar et al., 1998; Clarke et al., 2008; Cho et al., 2018; Kim and Lee, 2018). VOCs are carbon-based volatile chemicals emitted into the atmosphere by industry and automotive exhaust causing different environmental and health problems (Bernstein et al., 2008; Yu and Kim, 2012; Yang et al., 2019). In recent years, there has been a dramatic increase in the rate of VOC emissions, directing research efforts to find an efficient and cost-effective method to reduce pollutants (Khan and Ghoshal, 2000; Li et al., 2009; Jafari et al., 2018; Piazzoli and Antonelli, 2018; Lyu et al., 2020). Piumetti et al. (2015) showed that the VOC emissions have a direct effect on the formation of ozone and smog in the troposphere as well as ozone depletion in the stratosphere. Different VOCs such as formaldehyde, naphthalene, chloroform paradichlorobenzene (1,4-dichlorobenzene), acetaldehyde, benzene and toluene are considered toxic and carcinogenic to humans (OEHHA, 2005; Batterman et al., 2012; Hakim et al., 2012; Chin et al., 2013; Louie et al., 2013; Chen et al., 2017; Latif et al., 2019; Nair et al., 2019). Rezaee et al. (2008) showed that the inhalation of toluene affects the nervous system by decreasing the ability to focus and think, memory loss, muscular deficiencies, and vision problems. Therefore, there is an urgent need for effective processes to remove and reduce VOCs from the environment, eliminating their effect on human health and improving environmental quality.Among VOCs, toluene (TOL) is considered a highly representative model as it contains aromatic hydrocarbons that resist biodegradation under normal conditions (Momani and Jarrah, 2009; Dole et al., 2013). Different epidemiological studies have confirmed that exposure to TOL can cause severe health problems including effects on the nervous system, memory, and muscles (Djurendic-Brenesel et al., 2016; Jafari et al., 2019). Other studies have shown that the inhalation of TOL as a suspected factor of causing cancer (Dees et al., 1996; Lee et al., 2006, 2008). Toluene has been detected in different industrial processes such as printing, paint pressing and petrochemical industries (Jenck et al., 2004; Kamal et al., 2012; Dole et al., 2013). Toluene (TOL) has a long half-life (Labeau et al., 2003), and it also has important photochemical properties leading to the creation of tropospheric ozone (Derwent et al., 1996; Zheng et al., 2009; Nair et al., 2019).The removal of VOCs via biotic (e.g., biodegradation) and abiotic processes (incineration, adsorption, micro-filtration) have been widely employed (Vandenbroucke et al., 2011; Gil et al., 2014; Li et al., 2014). Biotic biodegradation using aerobic processes has been recognized as a reduction alternative for only biodegradable VOCs (Urase and Kikuta, 2005; Onesios et al., 2009; Lahti and Oikari, 2011). Non-biodegradable and/or toxic contaminants are difficult to be reduced by these processes. Abiotic processes, on other hands, have many disadvantages including the production of secondary pollutants, high cost, and energy requirements, difficult operational conditions and low efficiencies under low VOC concentrations (Li et al., 2014). Different studies have focused on developing economically feasible and effective treatment processes for the removal of VOCs (Everaert and Baeyens, 2004; Moulis and Krýsa, 2013; Tejasvi et al., 2015; Qian et al., 2018).Amongst these processes, advanced oxidation technologies (AOTs) showed promising results in removing different VOCs from the environment (Gamal El-Din et al., 2006; Al Momani, 2007; Al Momani and Jarrah, 2010; Hussain et al., 2011; Almomani and Baranova, 2013; Liu et al., 2019). Photo-catalytic oxidation processes (PHCOPs) are an important part of AOTs that can be performed through the exposure of the photo-catalyst to light photons and generate radicals that attack the pollutants leading to their degradation into eco-friendly products (Zhao et al., 2016; Yang et al., 2019). The PHCOPs have many advantages over other treatment processes including high efficiency, low utilization costs, long lifetime of the catalysts, mild operational conditions and the ability for breakdown wide varieties of complex pollutants into simpler compounds.In recent years, AOTs and mainly PHCOPs were operated with the solar irradiation to reduce the operational cost while maintaining high efficiency. In such combinations, titanium dioxide (TiO2) appears as the main photo-catalyst due to its low cost and high physio-chemical characteristics (chemical and thermal stability, non-toxic, and high photo-catalytic activity) (Fernández-García et al., 2004; Khalifa, 2005; Chen and Zhang, 2008; Wu et al., 2013; Tejasvi et al., 2015; Nomura et al., 2020; Zeng et al., 2020). The use of TiO2 in AOTs showed high removals of TOL and ethylbenzene under UV irradiation (Chen and Zhang, 2008). However, due to high UV-driven activity triggered by TiO2’s wide bandgap (≈3.2 eV) and the rapid recombination of electron-hole pairs, its efficiency with the sunlight irradiation is very low. As such, the enhancement of the photo-catalytic ability of TiO2 under visible light is of great importance. Toward this aim, rigorous efforts have been given to the addition of dopant elements into the lattice of TiO2 to permit its use under visible light. Such enhancement will permit the use of sustainable energy resources (e.g. solar irradiation) rather than the need for the production of the expensive UV light (Labeau et al., 2003; Almomani et al., 2016, 2018a, 2018b).Different research works have reported that doping TiO2 with other metals could increase its service life and enhance its TiO2 activity under solar irradiation (Zhou et al., 2006; Ni et al., 2007; Devi et al., 2009, 2010; Christoforidis et al., 2012). The doped metal improves the photo-catalytic activity by shifting the adsorption spectrum to the visible light range, decreasing the bandgap energy, improving the production of electron-hole pairs, rising the speed of photon transfer to the surface of the catalyst and reducing the recombination of the electron-hole pairs. Moreover, the presence of metals with TiO2 can contribute to electron-hole separation and thus enhance the photo-catalytic activity of the process (Litter and Navio, 1996; Zhang et al., 1998; Adán et al., 2007; Li et al., 2014). Kundu et al. (2014) reported a 70% loxacin (25 ppm) photo-degradation using 1 g/L of Ni-doped on TiO2 prepared by hydrothermal method. Rahimi et al. (2012) showed that the photocatalytic degradation performance of methylene blue using N- and S-co-doped TiO2 was 16% higher than naked TiO2. The higher photo-catalytic activities were related to the reduction in the band-gap of the co-doped TiO2. Similarly, Wen et al. (2009) reported significant improvement in the photo-catalytic degradation of methylene blue using I–F-co-doped TiO2. The improved photo-catalytic activity was related mostly to the significant increase in the surface area and stronger absorbance in the visible light range after doping with I and F (Almomani et al., 2019). enhanced the solar photo-reduction of CO2 by adding Cu into the lattice of TiO2.Different procedures were proposed for the preparation of metal-doped TiO2 including precipitation (Dvoranova et al., 2002), hydrothermal, solvothermal (Zhu et al., 2006), chemical vapor deposition (Wu et al., 2007) and electrospinning (Patil et al., 2003). Among all methods, sol-gel method is advantageous for the synthesis of nano-powders due to the production of homogenous and high purity powders under controlled stoichiometry and ambient temperature (Akpan and Hameed, 2010; Bhosale et al., 2016; Catauro et al., 2017; Elsellami et al., 2018; Xiao et al., 2018; Ji et al., 2019).Although different studies reported doping TiO2 with metal such as V, Cr, Fe, N, C, and S, up to our knowledge there is no study investigated the effect of doping of Cobalt (Co) into the TiO2 lattice. Moreover, a well-established rule concerning the optimum Co–TiO2 composition to TOL removal efficiency has not been reported yet. There still a lack of knowledge of how doped metal affects the mechanism of solar PHCOPs. Discrepancies also exist regarding the optimal doping content in the TiO2 lattice for maximum VOC removal. Accordingly, the present work presents the preparation of a co-doped Co–TiO2 photo-catalyst using a modified sol-gel method. A multi-characterization technique (UV–Vis, N2 isotherms, XPS XRD, TEM) was used to examine the structural and electronic properties of the synthesized photo-catalysts. The as-synthesized catalyst was tested against sunlight toward the photo-oxidation of TOL. The catalyst selectivity and oxidation products were identified and the main mechanism of photo-catalytic oxidation of TOL under solar irradiation was presented.Analytical grades of Cobalt (II) acetate (CH₃COO)₂Co·4H₂O, Merck, CAS#: 6147-53-1), mono-ethanolamine (MEA) (NH2CH2CH2OH, Merck, CAS#: 141-43-5), isopropanol alcohol (C3H8O, Merck, CAS #: 200-661-7), titanium butoxide (Ti(OCH2CH2CH2CH3)2, Merck, CAS #: 5593-70-4), ethyl alcohol (C2H5OH, CAS #: 64-17-5.) and toluene (C₆H₅CH₃, Merck, CAS #: 108-88-3) were used in this study.A sol-gel method was used to prepare the Co–TiO2 photo-catalyst with a mass fraction of Co of 1, 2, 5 and 10 wt%. The preparation process is similar to the procedure proposed by (Bhatia et al., 2016) with some modification. For each photo-catalyst, a specific amount of (CH₃COO)₂Co·4H₂O, calculated based on the required mass fraction, was added stepwise to a mixture of 0.91 g NH2CH2CH2OH, 10 mL deionized water, and 15 mL C3H8O producing solution I. In a separate bottle, 10 mL of (Ti(OCH2CH2CH2CH3)2 was dispersed in 40 mL C2H5OH and sonicated for 20 min producing solution II. Solution II was added into the solution I in a stepwise fashion, mixed at 120 rpm and room temperature, and a stable sol was finally obtained after stirring for 2 h. The resulted sole was left reacting for another 6 h at room temperature, dried with dry air at 75 °C for 40 h, calcined at 500 °C for 2.5 h and used in tests.The composition and morphology of the photo-catalyst were examined using scanning electron microscopy (SEM–EDS, Quanta 600 model). Tests were performed at an electron beam of 20 eV. The structure of the photo-catalyst was tested using XRD (Hiltonbrooks). A Brunauer–Emmett–Teller (BET) analyzer was utilized to estimate the surface area of the nano-catalyst. An XPS (Kratos Axis Ultra) was used to determine the chemical composition and electronic states of the photo-catalyst. A Porosimetry analyzer (Micromeritics Autopore IV 9500 V1.05) was used to determine the pore size and porosity of Co–TiO2 under mercury (Hg) pressure in the range of 0.1–20,000 psia. The band-gap and visible light absorbance were measured by UV spectroscopy (UV–Vis, Cary 300).Catalytic decomposition of TOL was carried out in a flow-type solar pilot plant (SPP) as shown in Fig. 1 . The SPP consists of synthetic air (20 vol % O2/N2), TOL tanks (Aldrich, purity 99.8%), a humidifier, a gas mixer, solar photo-reactor (SPHR) and a gas chromatograph. The SPHR contains four quartz tubes (L = 40 cm and ID = 1.6 cm) making a 45° with the horizontal line. The photo-elements are hosted within a compound parabolic collector (CPC) to allow all the solar radiation (direct and reflected) arriving the solar platform to be available for TOL oxidation. The as-prepared photo-catalysis were attached to quartz tubes following the procedure presented in our previous work (Almomani et al., 2016). The sun’s movement was tracked automatically and a radiometer (Macam Q102 PAR) was used to determine the light intensity available for solar oxidation. The effluent line is connected to a humidity meter and Gas Chromatograph (GC) for off-gas analysis. The mass flow rate in the inlet line was controlled using a mass flow controller (MFC- SFC5300, USA). The temperatures of the TOL and humidifier were controlled at 25 ± 2 °C. The concentrations of inlet and outlet streams were determined by a GC-TCD (PerkinElmer Clarus 500).The photo-catalytic oxidation of TOL over TiO2 and CO–TiO2 with different Co mass fractions was conducted at atmospheric pressure. The gas-phase inlet mixture (flow rate of 27.5–82.5 100 L/min) was prepared by mixing TOL vapor with a wet air stream producing a gas stream with the required inlet concentration of TOL. The mixture was circulated through the reactor for 80 min h in dark. Then, the reactor cover was removed, and the photo-catalyst was left to react under natural solar irradiation. The inlet and outlet streams concentrations were analyzed by an online GC-TCD using several columns (Shin Carbon ST, Q PLOT, 2 OV101, and molecular sieve) to quantify TOL, O2, CO2, CO, Benzene and benzaldehyde. The detection limit for all these gases was determined to be in the range of 0.2 ± 0.02 ppmv. The rate of TOL photo-catalytic oxidation was followed under steady-state conditions, typically accomplished after 140 h of irradiation.Measured inlet and outlet concentrations of TOL and the concentration of effluent CO2 gases were used to calculate TOL conversion (%TNConv.) as in Eq. (1) and the degree of Mineralization (%Min) as in Eq. (2): (1) % T N C o n v . = [ T O L [ i n − [ T O L ] o u t [ T O L ] i n ∗ 100 % (2) % M i n = [ C O 2 ] o u t l e t ∗ 100 7 ∗ [ T O L ] i n i t i a l ∗ T N c o n v . where [ T O L ] i n a n d [ T O L ] o u t are the initial and outlet concentrations of TOL, [ C O 2 ] o u t l e t is the concentration of CO2 in the effluent gas, Q is the gas flow rate.The adsorption of TOL on TiO2 and Co– TiO2 was estimated following (3): (3) Q e = ( [ T O L ] i n − [ T O L ] e ) V M , where [ T O L ] e is TOL equilibrium concentrations of (mg/L). V is the reactor volume of (L) and M is photo-catalyst the mass (mg). The experimental data were fitted to Langmuir isotherm (Eq. (4)) to estimate the adsorption constant, Ka: (4) Q e = Q m a x K a [ T O L ] e ( 1 + K i [ T O L ] e ) , where Qe (mg/gcat) is the amounts of TOL adsorbed on the photo-catalysis, Qmax (mg/gcat) is the maximum amount of TOL adsorbed and Ka (mg−1·L) is the adsorption constant. The kinetic of the solar photo-catalytic oxidation of TOL on TiO2 or Co–TiO2 was tested against Langmuir–Hinshelwood (L–H) expression (Eq. (5), integrated form) (Momani and Jarrah, 2009). (5) ln { [ T O L ] i n [ T O L ] o u t } ( [ T O L ] i n − [ T O L ] e ) = Q m a x k ( V V ˙ ) ( [ T O L ] i n − [ T O L ] e ) − K i where k is the constant of reaction, Ki is the adsorption equilibrium constant, and V ˙ is the gas flow rate. The simultaneous adsorption and oxidation of TOL were proposed as in Eq. (6), linear form: (6) − d [ T O L ] t d t = { K P H + Q m a x K a 1 + a + K [ O H ] [ T O L ] t where KPH and K are the reaction rate constant to photon-energy and hydroxyl radical, respectively. [ O H ] t and [ T O L ] t are the concentration of the hydroxyl radical and TOL in the gas phase. Fig. 2 a shows the XRD pattern of TiO2 doped with Co at different weight percentages (0, 1.0, 2.0, 5 and 10.0 wt%). The diffraction peaks of the Co–TiO2 structure can be indexed to the tetragonal anatase phase of TiO2 as confirmed by comparing the obtained diffraction power with standard card (JCPDS #21–1772). The peaks of Co in the mixture are mixed with the peaks of TiO2 as shown in the XRD of the Co–TiO2 structure under all doping ratios. The obtained trends can be attributed to the high dispersion of Co within the TiO2 structure and the close ionic radius of Co (0.72 A) to with Ti4+(0.68A), conditions that make it difficult differentiate between the diffraction powers of TiO2 and Co in the lattice. The obtained trends agree with the findings of (Hamadanian et al., 2010) and (Huang et al., 2006), who showed no identified diffraction peaks of Co during the doping process of Co in TiO2. Barakat et al. (2005) reported that the TiO2 anatase peaks occur for TiO2 structures calcined in the temperature range 723–873 K. Co–TiO2 structure calcined at higher temperature showed diffraction peaks for Co and TiO2. The crystallite size of Co–TiO2 was calculated using Scherre equation to be in the range 10.03 ± 0.06 to 12.09 ± 0.06 nm with general trends showing higher crystallite size at a lower mass fraction of Co. Increasing the wt% of Co in the structure led to a noticeable decrease in the crystallite size. The specific surface area (Asp) of the Co–TiO2 was determined to be in the range 75.31 ± 0.05 to 85.40 ± 0.05 m2/g. It was also observed that Asp increased by increasing the wt% of Co in the Co–TiO2 structure. Besides, it was observed the structure has homogenous spherical characteristics with a diameter in the range 7–25 nm as estimated by TEM measurements (Fig. 2b). The EDS measurements showed that Co is dispersed on the surface of TiO2 and the morphological structure was not changed during the doping process. Fig. 2c–e shows the chemical composition and the oxidation state of Co 2P, Ti 2P O 1s of Co–TiO2 with 5 wt% of Co. The binding energy of Co 2p1/2 core at 796.4 eV and 2P3/2 at 780.4 eV belongs to Co(ll) oxide as confirmed by (Wagner, 2007). The satellite peak of Co 2P3/2 and Co 2P1/2 represent Co+2 at a binding energy of 786–10 and 802-5 eV, respectively. The obtained results show that Co exists within the TiO2 lattice in the form of Co+2. These observations were confirmed by (Huang et al., 2006) and (Shifu et al., 2008). The peaks of Co 2P before after the photo-catalytic reaction show no change. The binding energy of Ti 2P3/2 core level at 459 eV and Ti 2P1/2 at 465 eV represent Ti4+ in TiO2 structure. (Fig. 2d). The XPS spectrum of O 1s region illustrates that O2 exists in two forms in the structure with a binding energy of 530 and 532 eV (Fig. 2e). The peak at 530 is related to O in the bulk of TiO2, and the peak at 532 eV is related to O2 on the surface or oxygen within hydroxyl species (Peng et al., 2012). Fig. 2f presents the absorption spectrum of TiO2 and Co–TiO2 at different mass fractions of Co. Adding Co toTiO2 showed a redshift in the absorption spectrum toward the visible light region. The incorporated Co in TiO2 matrix is the main reason for this shift due to the charge transfer transition from the 3d orbitals of Co to the conduction band. The band-gap energies of the Co–TiO2 catalyst were estimated to be in the range 2.51–3.04 eV with the largest shift observed at higher Co content. Shifu et al. (2008) observed the same shift in the absorption spectrum upon adding Co to TiO2.The chemisorption measurements showed that the Co dispersion and surface area increased by decreasing the mass fraction of Co up to 5 wt%. Higher mass fraction decreased the Asp. Co–TiO2 doped with l, 5 and 10 wt% of Co has Asp of 60.7, 85.6 and 63.4 m2/g, and a degree of dispersion of 90, 93 and 91%, respectively. The obtained results show that the higher the mass fraction in the structure, the more uniform and the higher dispersion on the TiO2 surface. However, the high load of Co might block the pore space in the photo-catalyst leading to lower Asp and decrease in the mean dispersion. Fig. 3 presents the pore size distribution and N2 adsorption-desorption isotherms of Co–TiO2 calculated by the Barrett, Joyner and Halenda (BJH) method. All Co–TiO2 exhibits H1-type hysteresis attributed to the mesoporous structure of these photo-catalysts (Yu et al., 2007). Increasing the CO mass fraction decreased the pore size diameter being 8.8, 8.0, 7.6 6.1 and 5.8 nm for the Co–TiO2 with Co of 0, 1, 2, 5 and 10 wt%, respectively. The decrease in pore size diameter can be attributed to the increase in the amount of Co deposition on the TiO2, blocking the pores and reducing the pore diameters.Dark adsorption plays a key controlling step in determining solar oxidation efficiency. The results gathered from this step reflect the ability of the catalyst to adsorb the TOL and/or oxidize it. Results showed that the adsorption capacity of TOL decreased by increasing the Co content of the catalyst suggesting a beneficial enhancement in the treatment TOL, where most of the processed TOL undergoes solar oxidation. Fig. 4 a shows the evolution of TOL concentration under dark, first 80min, and solar irradiation for an additional 60 min using TiO2 and 5 wt% Co–TiO2. Tests were conducted with an inlet TOL concentration of 150 ppmv, the flow rate of 42 NL/min, a relative humidity of 50% and a solar light intensity of 0.1 W/cm2. The results revealed that the oxidation of TOL in dark is negligible. The concentration of TOL in the reactor effluent gradually increased to reach up to 95.5 and 90% of the inlet concentration for tests carried out under dark with TiO2 and 5 wt% Co–TiO2, respectively. The observed 4.5 and 10% decrease in the effluent concentration of TOL is due to adsorption on the catalyst. The results also suggest that both photo-catalysts do not affect TOL oxidation in dark. No by-products were detected in the effluent confirming that no degradation nor mineralization occur for TOL. As the TiO2 and Co–TiO2 catalysts were exposed to humidity, a reaction took place between H2O and the cation on the surface of the photo-catalyst leading to the formation of hydroxyl group (i.e. Ti–OH or Co–OH bonds) from the surface of TiO2 or Co–TiO2 (Wen et al., 2009; Almomani et al., 2019). The hydroxyl groups after that bond to π-electron from the TOL aromatic ring and absorb TOL on the surface. In addition, direct electrostatic attraction between the aromatic ring in TOL and Co cations can increase the adsorption of TOL (Dvoranova et al., 2002; Takeuchi et al., 2012; Pham and Lee, 2015). Fig. 4a also shows that the Co–TiO2 achieved saturation when the concentration of TOL in the inlet stream was almost equal to the outlet stream after 71 min, while TiO2 took 76 min to reach the same.Under solar light irradiation, the concentration of TOL in the effluent gas stream showed an instantaneous increase, exceeding the inlet concentration due to the desorption of the TOL by the scrubbing effect of CO2 produced from the solar oxidation on TiO2 or Co–TiO2 surfaces. The results achieved with Co–TiO2 compared to naked TiO2 represent a significant enhancement in the photo-catalytic activity of Co–TiO2, which is related to the presence of Co. As the addition of Co to TiO2 shifted the adsorption spectrum toward visible light and improved the solar light absorption, it is expected that the absorbed photon increases the photo-catalytic activity. It is expected that the bandgap energy of the photo-catalyst will decrease, the production of electron-hole pairs will enhance, and the movement of photons within the catalyst will be faster which will reduce the recombination of the electron-hole pairs. Consequently, the produced electrons and holes react with oxygen or water to produce different radicals including hydroxyl radicals (•OH) and superoxide radicals (• O 2 − ) (Rezaee et al., 2008), attacking TOL and causing its decomposition into different byproducts including CO2, benzene, benzaldehyde and H2O. this result was confirmed by the increase in the concentration of these compounds in the effluent stream as shown in the second part of Fig. 4b and c. The oxidation reaction of TOL over the Co–TiO2 photo-catalyst can be represented by reactions 1-5 C o − T i O 2 → s o l a r i r r a d i a t i o n e C B − + h V B + ( Re a c t i o n   1 ) 2 h V B + + 2 H 2 O → 2 H + + H 2 O 2 ( Re a c t i o n   2 ) H 2 O 2 → 2 O H · ( Reaction  3 ) ( O 2 ) a d s + e C B − → O 2 · − ( Re a c t i o n   4 ) O H · + C 7 H 8 → +   C O 2 + H 2 O + b y p r o d u c t s ( benzene ,  benzaldehyde ) ( Reaction  5 ) The low attachment of the generated CO2 in addition to its scrubbing effect enhanced the desorption of TOL adsorbed during the dark period, showing a sudden increase in the effluent TOL concentration. Then TOL underwent fast photo-catalytic oxidation over Co–TiO2 resulting in a fast decrease in its effluent concentration and an increase in the production of CO2 (Fig. 4b). Once the adsorbed TOL was detached from the surface of Co–TiO2, a sharp decrease was observed. The concentration of TOL in the effluent stream reached a steady-state value of 4.5 and 60% of the initial TOL concentration after 120 min for tests carried out with 5 wt% CO–TiO2 and TiO2, corresponding to %TNconv of 96.5% and 28.5%, respectively. The corresponding CO2 concentration in the effluent stream reached 962 and 140 ppm, respectively. Since the inlet concentration of TOL was kept constant at 150 ppm, TOL underwent continuous photo-catalytic oxidation on the surface of Co–TiO2 and TiO2 producing a steady-state amount of CO2 of 962 and 124 ppm, corresponding to 91.3 and 33.2% of mineralization, respectively. The concentration of other byproducts (benzene and benzaldehyde) showed similar trends to CO2. The steady-state concentrations of benzene and benzaldehyde were 392 and 99 ppm, respectively. Fig. 4b and c shows that the effluent concentrations of CO2, benzene, and benzaldehyde produced over Co–TiO2 were 5, 9 and 12-fold higher than the same compounds produced over TiO2, confirming the superior photo-catalytic activity of the first photo-catalyst.Another set of experiments were carried out with solar irradiation started from the beginning of the experiment, eliminating the dark adsorption period, as shown in Fig. 5 . The effluent TOL showed a decrease in the concentration from the inlet concentration of 150 ppm to stable values of 3.1% and 39% of the initial concentration after 30 min of solar oxidation for the tests carried out with 5 wt% of Co–TiO2 and TiO2, respectively. Initially, TOL was removed by both adsorption and photo-catalytic oxidation, leading to a decrease in the TOL effluent concentration. As the adsorption of TOL was disturbed by the produced CO2, the main byproduct of the photo-catalytic oxidation, the high % TNconv of TOL over Co–TiO2 compared to TiO2 is due to the improvement in the photo-catalytic activity by doping with Co. The stable effluent of TOL after 50 min is related to continuous oxidation over the surface of Co–TiO2. The effluent concentration of TOL was stabilized at 3.1 and 39% for tests carried out with 5 wt% of Co–TiO2 and TiO2 corresponding to oxidation of 145.4 and 58.8 ppm of out the initial TOL (150 ppm). The CO2 production was increased from 0 at the beginning of the reaction to 1122 ± 20 ppm (%Min≈ 93.4) after 50 min confirming the conversion of TOL to CO2 over the 5 wt% Co–TiO2 catalyst. Other by-products such as benzaldehyde (99 ppm) and benzene (396 ppm) were detected in the effluent stream (Blount and Falconer, 2002). showed that the photo-catalytic oxidation of TOL over TiO2 and Pt/TiO2 was fast producing benzaldehyde and benzene as intermediates which react further to oxidized products.Considering the high inlet concentration of TOL (150 ppm) used in the present test, the effluent concentration 4.6 ppm is considered an excellent removal efficiency compared with other studied processes. European Union set the time-weighted average (TWA) of ambient air quality standard for toluene at 20 ppm, while the 8-h TWA in the workplace in Quebec, Canada was set at 100 ppm (Masih et al., 2017; Golbabaei et al., 2018).The photo-catalytic activity of Co–TiO2 for the oxidation of TOL was tested under different initial concentrations of TOL in the range 20–150 ppm and hydraulic retention time (HRT) in the range 50–150 s at natural pH of 6.8. Fig. 6 shows that the inlet concentration of TOL had a minor impact on TOL degradation for tests conducted as high HRT ≥100 s and mass fraction of Co in the lattice ≥5 wt%. Decreasing the HRT resulted in a decrease in TOL degradation under all the studied inlet concentrations. The percentage degradation of TOL at an initial concentration of 38 ppm was decreased from 62.2% at HRT of 150 s to 41.6% at 50 s for tests performed with 1 wt% Co–TiO2. Tests carried out with higher concentrations (e.g. 100 ppm) showed a decrease in the percentage degradation of TOL by 15% as the HRT reduced from 100 to 50 s, over the same catalyst. The decrease in percentage degradation of TOL is attributed to low residence time which impacted the time required to achieve complete oxidation. Fig. 7 a shows that increasing the mass fraction of Co in the photo-catalyst from 1 to 5 wt% increased the percentage degradation of TOL. However, a further increase in the mass fraction of Co to 10 wt% resulted in a reduction in the degradation efficiency. Up to 95.6% of photo-catalytic degradation of TOL was achieved after 50 min of irradiation with photo-catalyst with a Co mass fraction of 5 wt%. Tests carried out with 1 wt % photo-catalyst showed 69.7% TOL degradation. The relationship between TOL oxidation and the Co mass fraction is due to a large number of active sites on Co–TiO2 accessible for the photo-catalytic reaction. The Co–TiO2 catalysts showed significant catalytic activity toward TOL oxidation compared with TiO2 alone. Tests performed in the dark revealed that the observed oxidation activities were fully attributable to solar photo-induced processes. The observed results also confirm that the change in surface morphology by adding Co to TiO2 lattice enhanced the solar photo-catalytic activities toward TOL oxidation. Moreover, the obtained results show solar oxidation of TOL is independent of the excitation wavelength. Nonetheless, the surface structure and electron generation ability are mainly responsible for the enhanced activity. The decrease in the photo-catalyst activity at a higher mass fraction of Co can be related to the shielding effect of Co, which decreases the solar light penetration to the catalyst.It is known that OH• radical play a key role in the photo-catalytic degradation of TOL (Fuerte et al., 2002; Sleiman et al., 2009). Adding Co to TiO2 shifted the absorption spectrum of the photo-catalyst toward visible light, improved electron-hole separation and thus enhanced the photo-activity. Moreover, the presence of Co in the structure increased the specific surface area and this improved oxidation efficiency. Fig. 7b illustrates that the naked TiO2 did not display significant activity for the solar photo-catalytic oxidation of TOL. It is known that TiO2 absorbs solar light with energy greater than or equal to its band-gap, transferring electrons from valence conduction bands and enhancing the production of electronic vacancies in the valence band (Reaction 6). The transfer electrons and the produced holes contribute to a series of reactions generating hydroxyl radicals that oxidize TOL (Reactions 4 and 7). The %TN conv reported for TiO2 suggest a limited ability of TiO2 to work under solar irradiation toward the solar oxidation of TOL. T i O 2 → h ν ≤ 390 n m e C B − + h V B + ( Re a c t i o n   6 ) O H − + h V B + → H O ( Re a c t i o n   7 ) On the other hand, the addition of Co to TiO2 structure enhanced visible light absorbance, enhanced the photo-catalytic activities and resulted in a significant tendency for the oxidation and mineralization of TOL to CO2 and other byproducts. As TiO2 has very low optical absorption properties under solar irradiation, the improvement in the solar oxidation activity is due to the presence of Co in the structure. Wang et al. (2015) showed that TiO2 has a considerable absorption limit in the wavelength range 254–370 nm. Adding Co to the TiO2 resulted in a significant enhancement of light absorption in the solar irradiation region (See Fig. 2f). It was observed that the light absorption ability increased by increasing the mass fraction of Co in the Co–TiO2 up to 5 wt%. Further increase in the mass fraction of Co to 10 wt% showed a slight increase in the light absorption. A small portion of the photons absorbed in the Co–TiO2 is used as a source of heat, and the major part is used in exciting the electrons from the valence band to the conduction band in TiO2. In the absence of Co, the TiO2 has limited energy gaps between its valence band and conduction band (band-gap energy). Thus, high amounts of photon energy are required to surpass band-gap energy (∼3.2 eV) and excite the electrons to start the oxidation of TOL. Based on this, naked TiO2 absorbs only UV light to fulfill this energy requirement. Addition of the Co to TiO2 enhanced the excitation of the structure by the formation of Ti3+, Ti4+, and Co2+, providing the structure with the ability to absorb more energy photons and utilize them in transferring electrons from the valence band (O 2p) to the conduction band thus, enhancing the solar oxidation efficiency of TOL. As the mass fraction of Co2+ increases, the ability of Co–TiO2 to absorb solar irradiation increased and the oxidation potential improved. However, a high mass fraction of Co might increase the surface coverage of TiO2 leading to a decrease in the light absorbed and inhibiting the photon energy from reaching the TiO2, resulting in a decrease in the ability to oxidize TOL. Therefore, a 5 wt% of Co was determined to be the optimum mass fraction of Co. The observed results suggest that Ti3+, Ti4+, and Co2+ play an important role in electron generation and transfer and the improvement in the electron-hole separation. The generated electrons increased the surface reaction with oxygen and water contributing more to TOL solar oxidation. The highest TOL removal and mineralization were 96.5 and 93.3% respectively, achieved by Co–TiO2 with a mass fraction of 5 wt%. Although the band energies of Co–TiO2 with a mass fraction of 1 and 2 %wt were 2.76 and 2.70 eV, respectively, were lower than that of 10 wt% (2.83 eV), their TOL photo-catalytic removal and mineralization degree were lower than those of the 10 wt %. This is because a higher mass fraction of Co blocked the lattice and decreased the incidental light absorption from reaching the TiO2 layer. Moreover, the excited electrons from the conduction band of Co would easily fall back to the valence band to recombine with holes before reacting with water or oxygen to produce oxy radicals, resulting in lower %TNconv.Another set of tests were conducted to study the effect of humidity on the solar oxidation of TOL. The summary of results is presented in Table 1 . The %TN conv increased by increasing the percentage relative humidity (%RH) from 10% up to 50%, after which the reported %TN conv slightly decreased. The observed trend was aimed at the contribution of water in the gas phase in the formation of hydroxyl radicals, which will increase the %TN conv as shown in reaction 8 combined with reaction 5. The decrease in %TN conv at high %RH is related to the competitive adsorption between the water and TOL on the photo-catalyst active site resulting in a decrease in the %TN conv (Momani and Jarrah, 2009). H 2 O → h ν H ˙ + O ˙ H ( Reaction  8 ) To identify the main mechanism of photo-catalytic degradation of TOL, scavenger tests were performed to assess the ability of Co–TiO2 to produce active hydroxyl radicals (OH• and • O 2 − . ) and identify the role of each species in TOL oxidation. Tests showed that the concentration of OH• is significantly higher than.   • O 2 − . . It is known that OH• is the most powerful oxidizing agent, that can react with organic matter leading to its degradation (Shawaqfeh and Al Momani, 2010). The concentration of OH• was observed to increase by increasing the Co content in the Co–TiO2 lattice up to 5 wt%, after that a decrease in the radical concentration was observed. The obtained results suggest that the prepared Co–TiO2 photo-catalyst is capable of generating high concentrations of OH• under solar irradiation. The high production of hydroxyl radicals could be attributed to the decrease in the bandgap energy, enhancement in the production of electron-hole pairs and increase the rate of photons movement within the catalyst and decrease the recombination of the electron-hole pairs as results of solar light absorption enhancement (Reactions 1–5). The decrease in the OH• generation at higher Co mass fraction (Co > 10 wt %), can be related to the decrease in photons movement through the coupling interface between Co and TiO2. Electrons can be captured within the interface between Co and TiO2, leading to a sudden decrease in the concentrations of hydroxyl radicals.Based on the above analysis, the possible mechanism of the photo-oxidation of TOL could be proposed as per Fig. 8 . Under solar irradiation, the Co–TiO2 reach excitation state generating electrons and holes ( e C B − and h V B + ) in reaction 1. Electrons react with O2 or water producing radicals ( • O 2 − . ) as per reaction 7, and the h V B + react with water producing hydrogen peroxide that dissociates generating OH• (reactions 2 and 3). The presence of Co metal in the photo-catalyst structure facilitates the transfer of electrons and thus enhance the radical generation. Additional OH• can be generated by reaction with oxygen on the surface of metal On the other hand, photo-generated electrons react with oxygen on titanium dioxide itself producing superoxide radicals (Urase and Kikuta, 2005). All the photo-generated radicals are available to react with TOL and contribute to high %TN conv .The amount of TOL adsorbed on the TiO2 and Co– TiO2 did not exceed 3% and 10% of the inlet concentration respectively, suggesting a limited adsorption profile. Regression analysis of experimental data following Eq. (5) showed poor fitting and the model was rejected. As the adsorption of TOL was very low, Eq. (6) can be reduced to Eq. (7), linear form (Almomani and Baranova, 2013): (7) ln { − d C d t } = ln { K P H } + ln ( K O H C O H ) C Information about the concentration of hydroxyl radicals is required to solve Eq. (7). However, as the hydroxyl radical is self-generated in the solar reactor and highly dependent on the solar energy and gas phase, %RH, Eq. (7) can be rewritten as Eq. (8): (8) ln { − d C d t } = ln { K P H } + ln ( R H ) C Experimental data of TOL oxidation over TiO2 and Co–TiO2 showed a good agreement with Eq. (8). Table 1 presents the kinetic constants of the solar photo-catalytic oxidation of TOL over TiO2 and Co–TiO2. Based on the experiment results, the photo-catalytic oxidation of TOL depends in the %RH;(1) At low %RH (i.e. low water vapor), the oxidation reaction initiated by electron transfer from Co–TiO2 to O2 and water generating radicals causing the decomposition of TOL. As the concentration of water is low, the available oxygen plays a key role in the generation of more radicals and enhancing TOL conversion. The obtained results suggest that under low humidity, the oxidation of TOL can be improved by increasing the O2 content in the gas stream.(2) At high %RH, the available water plays an important role in the direct formation of OH• radicals leading to higher %TN conv . Table 1 also presents important information regarding the effect of the gas flow rate on the process kinetic. The general trends showed a decrease in the kinetic constant by increasing the flow rate. The kinetic constant values were 7.6 ± 0.3, 5.0 ± 0.2 and 3.3 ± 0.3 min−1 for tests carried out with gas flow rate of 27.5, 42.0 and 82.5 L min−1 at RH of 20%. Further investigation will be carried out shortly to study the techno-economic aspects of the process as well as the kinetic data for scale-up purposes.In this study, for the first time, a new photo-catalyst (Co–TiO2) was synthesized with a different mass fraction of Co and tested for the degradation of toluene (TOL). The as-prepared catalyst has improved surface characteristics and visible light absorption, and reduced electron-hole recombination. The enhanced photo-catalytic properties of Co–TiO2 improved the degradation of TOL by reducing bandgap energy and increasing the generation of radicals. The photo-degradation of TOL depends on the mass fraction of Co, inlet TOL concentration, gas flow rate, and relative humidity. The highest toluene conversion (%TN conv ) of 96.5% was obtained using 5 wt% Co, 150 ppm toluene concentration, 27.5 L min−1 flow rate and 50% relative humidity. The co-doped Co–TiO2 catalysts showed high selectivity (>90%) toward partial oxidation of TOL to produce CO2, benzene and benzaldehyde. The obtained results suggest that adding Co metal to TiO2 displayed excellent solar photo-catalytic properties that can be employed to remove toluene from the gas phase stream at an industrial scale.The publication of this article was funded by the Qatar National Library.

Cobalt (Co) co-doped TiO2 photo-catalysis were synthesized, characterized and tested toward solar photocatalytic oxidation of toluene (TOL). A multi-technique approach was used to characterize and relate the photo-catalytic property to photo-oxidation performance. Adding Co to TiO2 significantly changed crystal size and surface morphology (surface area, pore-volume, and pore size), reduced the bandgap energy of TiO2 and improved the solar photo-oxidation of TOL. Up to 96.5% of TOL conversion (%TN conv ) was achieved by using Co–TiO2 compared with 28.5% with naked TiO2. The maximum %TN conv was achieved at high hydraulic retention time (HRT) ≥ 100 s, Co content in the photo-catalyst of 5 wt% and relative humidity (%RH) of 50%. The mechanism of TOL solar oxidation was related to the concentration of OH• and • O 2 − . radicals produced from the generated electrons and holes on the surface of Co–TiO2. The products formed during the photo-catalytic oxidation of TOL were mainly CO2 and water, and minor concentration of benzene and benzaldehyde. Overall, the Co–TiO2 could be used as a potential photo-catalyst for the oxidation of toluene in gas-phase streams on an industrial scale.

Since the first industrial revolution, advanced inventions and technologies have enabled us to enjoy warmth in the winter, cool in the summer, brightness at night, and convenient transportation all over the world. 1 Most of these technologies depend heavily on our ability to exploit fossil sources of energy, resulting in an increasing demand for fossil fuel and excessive emissions of CO2. As projected by the International Energy Agency, the global annual energy demand will increase to 18 billion tons of oil equivalents, and 43 gigatons of CO2 will be released per year by 2035, which will aggravate energy crisis, increase the global average temperature, and acidify the ocean. 1,2 Severe situations have motivated a large number of researchers to pursue reliable and clean energy options. Proton-exchange membrane fuel cells (PEMFCs), especially refueled with hydrogen from renewable energy, are generally considered one of the most promising solutions because of their competitive advantages, such as zero emission, high efficiency, fast refueling, and low upfront cost. 3 In a typical PEMFC, fuel molecules (e.g., hydrogen) are oxidized on the anode, and oxygen gas is reduced on the cathode, outputting electric energy with pure water and heat as the only by-products (Figure 1 A). Unfortunately, the difficulty in O2 activation, O–O bond cleavage, and oxide removal causes sluggish kinetics of the oxygen reduction reaction (ORR) on the cathode, thus demanding stringent requirements to the catalysts. 4 After a long period of experimental exploration, platinum (Pt) and Pt-based catalysts are generally considered to be the most efficient ORR catalysts. Low-temperature PEMFCs currently adopt Pt nanoparticles (NPs) supported on carbon (Pt/C) or other Pt-rich materials as the cathode catalyst. 5 Nevertheless, the high cost of Pt greatly hampers further large-scale adoption of PEMFCs. According to the strategic analysis report, catalyst layers in the PEMFC system amount to US $11.24 kW−1 or over 20% of the total cost, in which over US $10 kW−1 attributes to Pt usage. How to reduce the dosage of Pt or substitute nonprecious metal for Pt without loss of performance is now of the greatest concern. 3,5–7 At present, improving atom utilization and boosting the intrinsically catalytic activity of Pt by reducing Pt nanostructure sizes, alloying, and constructing specific nanostructures with Pt-rich skin are common strategies for reducing the dosage of Pt. 4,5,7 A variety of delicate Pt-based nanostructures have been reported to exhibit significantly enhanced ORR activity. For instance, Li et al. fabricated ultrafine Pt jagged nanowires with diameters less than 5 nm, delivering 33 times the specific activity (catalytic activity normalized by surface area) or 52 times the mass activity (catalytic activity per given mass of Pt) of the commercial Pt/C catalyst. 8 The Adzic group deposited Pt monolayers on PdAu NP surfaces by the galvanic displacement method to optimize Pt utilization. The ultra-low Pt content was found to be enough to achieve high ORR catalytic performance. 9 In another important work by Chen et al., Pt3Ni nanoframes with Pt-skin surfaces were constructed after the interior of polyhedral PtNi3 nanocrystals was dissolved, significantly outperforming the commercial Pt/C catalyst for ORR activity. 10 However, these fine nanostructures typically have a high propensity to agglomerate or deform during the electrochemical process, resulting in an unfavorable deactivation and poor stability during long-term operation. 11 Meanwhile, the complicated synthetic procedures cause the manufacture of catalysts to be costly. These disadvantages make Pt-based catalysts still doubtful in further wide adoption. In consideration of the much lower price of nonprecious metals, such as iron (Fe), cobalt (Co), and nickel (Ni), the cost of PEMFCs could be significantly reduced by substituting nonprecious-metal catalysts for Pt-based catalysts. However, the ORR activity of conventional nonprecious NPs is lower than that of Pt counterparts by almost one order of magnitude, preventing them from directly acting as eligible ORR catalysts. 12,13 Similar to the case of Pt-based electrocatalysts, regulation of morphological and electronic structure of the nonprecious-metal catalysts is a general strategy for improving their ORR activity. Unfortunately, despite the tremendous efforts, few results achieve satisfactory catalytic activity and durability because of the flagrantly low intrinsic ORR catalytic activity of nonprecious metals. 4,5 The size of metal particles is a key factor in determining their catalytic performance given that the specific activity per metal atom generally increases with decreasing size of the particles. Single-atom catalysts (SACs) represent the theoretically ultimate size limit for metal particles, in which metal atoms are dispersed on specific supports and isolated from each other without appreciable interaction between them. Therefore, SACs are supposed to possess relatively high catalytic activity and maximum atom-utilization efficiency. 14–16 In 2000, Heiz et al. prepared a series of Pdn cluster supported on magnesium oxide with the help of mass-selected soft-landing techniques. 17 The single palladium (Pd) atom is surprisingly found to exhibit enough catalytic activity in acetylene cyclotrimerization to benzene. In another research by the Zhang group, atomically dispersed Pt atoms supported on Fe oxide (Pt1/FeOx) were successfully synthesized, carefully characterized, and applied in efficient and durable CO oxidation. 18 In 2016, Liu and co-workers developed a convenient photochemical strategy to fabricate a stable SAC with Pd atoms supported on ultrathin TiO2 nanosheets. 19 Such a Pd SAC exhibited high catalytic activity in hydrogenation of C=C bonds, outperforming commercial Pd/C catalysts and homogeneous H2PdCl4. In the same year, the Li and Wu groups initiated a new method to construct isolated metal atoms anchored on three-dimensional nanostructures, achieving a high ORR activity. 20 Other than this significative research, a series of SACs have been reported and showed surprisingly excellent performance in various catalysis processes, such as catalytic reduction of CO2, 21,22 electrochemical synthesis of ammonia, 23 methane conversion, 24 selective acetylene hydrogenation, 25 and other important chemical reactions. 14,15 Inspired by the powerful achievements by SACs, scientists have made fruitful researches in tuning SACs into active, reliable ORR catalysts as an alternative to expensive Pt-based materials. 15,16 In addition, SACs with uniform catalytically active sites provide us a golden opportunity of exploring the relationship between ORR catalytic performance and catalyst structure in an atomic scale, which could spur further research on the atomically rational design of ORR catalysts. 22 In this review, we first introduce the mechanism and electrochemical evaluation of ORR. Then, we concisely retrospect the development of Pt-based ORR catalysts and demonstrate how scientists optimize their catalytic performance by controlling their component, morphology, size, and facet exposure. After that, we describe the ORR performance of nonprecious-metal-based catalysts and the following three common strategies of improving their performance: (1) increasing the intrinsic activity by composition modulation, (2) confining metal species into carbon shells to preserve metal from corrosion, and (3) increasing the accessibility by constructing porous or other large-area structures. Further, we briefly review the developments of SACs, summarize recent advances in SACs for ORR catalysis, and demonstrate how the ORR performance on SACs is promoted by support construction and regulation of electronic structures. At last, we also present a brief perspective on the remaining challenges and future directions of SACs for ORR.It is generally accepted that the ORR undergoes either a “direct” four-electron pathway to generate O2-species (H2O in acidic solutions or OH− in alkaline solutions) or a “series” two-electron pathway to generate hydrogen peroxide (H2O2). 4,5 The “series” way has been considered one of the competitive strategies of producing H2O2, and it could replace the energy-intensive anthraquinone process. However, the “direct” four-electron oxygen reduction pathway is unanimously recognized as the favorable pathway since H2O2 reduces energy-conversion efficiency and accelerates the degradation of the proton-conducting polymer electrolyte in PEMFCs. 13 Different intermediates, including oxygenated (O*), hydroxyl (OH*), and superhydroxyl (OOH*) species, could be generated during oxygen reaction under common ORR conditions. Several possible transformations between these intermediates, as schematically shown in Figure 1B, make the ORR process more complicated. 26 Despite the tremendous efforts to find the rate-determining step in ORR, there is still no definitive conclusion because the reaction pathway depends, to a great extent, on the catalysts and environmental parameters such as solvent, temperature, and applied electrode potential. In the majority of cases, the overall ORR rate is determined by one of these three steps: (1) the first electron transfer to adsorbed O2 molecule, (2) the hydration of O2, and (3) the final desorption of H2O. 4 In addition, several studies have supported that oxygen coverage plays a critical role in ORR mechanisms. A high oxygen coverage causes O–O cleavage posterior to OOH* formation (so-called associative mechanism), whereas a low oxygen coverage makes O–O cleavage anterior to OH* formation (dissociation mechanism). 27 For a rigorous evaluation, a new ORR catalyst is supposed to be employed in a PEMFC and compared with acknowledged benchmarks, such as commercial Pt/C. Nevertheless, the complicated and costly fabrication of a PEMFC makes this approach impractical. In science labs, benefiting from operability as well as inexpensiveness, the rotating disk electrode (RDE) method has been widely adopted to quickly screen the ORR catalytic performance of new materials. 5,27 Typically, the catalyst sample is first dispersed into a homogenous ink with mixing water, alcohol (isopropyl in some cases), and Nafion by an optimized ratio, which is then deposited on a glassy carbon RDE. For Pt-based catalysts, especially for Pt NPs, the Pt loading is usually controlled below 50 μg/cm2 to avoid mass-transport loss caused by catalyst agglomeration.After electrode preparation, in a typical procedure, a cyclic voltammogram (CV) is first investigated by cyclic voltammetry in inert-gas (N2 or Ar)-saturated acidic solutions (0.1 M HClO4, 0.5 M H2SO4, or 0.1 M KOH). H+ would be reduced and adsorbed on catalytically active sites on the catalyst surface during the cathodic scanning, corresponding to the H adsorption region in current-potential (I-V) curve. In reverse, the regeneration of H+ occurs during anodic scanning and corresponds to the H desorption region in the I-V curve. After mathematic conversion, the electrochemical active surface area (ECSA) can be obtained from the integral of H desorption area in the I-V curve. The ORR polarization curve is measured in an O2-saturated solution (usually 0.1 M HClO4, 0.5 M H2SO4, or 0.1 M KOH) in a potential scanning window between 0.05 and 1.20 V versus reversible hydrogen electrode (RHE). As the mass transfer largely influences the ORR catalytic performance, the RDE is rotating (usually at a speed of 1,600 rpm) to mitigate the mass transfer loss during the ORR evaluation. To minimize the contribution from capacitive current, the scan rate is usually controlled below 20 mV s−1. One should note that current PEMFCs adopt cation-exchange membranes to separate the anode and cathode, making RDE tests in an acidic medium more practically significative.After polarization curve measurements, the kinetic current (jk, catalytic current without the loss caused by mass transfer) can be extracted according to the Levich-Koutechy equation: 1 j = 1 j k + 1 j l , c = 1 j k + 1 0.62 n F A C 0 * D 0 2 / 3 ν − 1 / 6 ω 1 / 2 , where j is the apparent current density (extracted from the polarization curve under different applied potentials directly) and jl,c is the diffusion-limited current density (usually the highest current density under relatively negative potentials). Thus, the jl,c is determined by the average number of electron transferred during ORR (n), the faradic constant (F), the geometric area of electrode (A), the concentration of dissolved O2 in catalysis solution (C0*), the diffusion coefficient of O2 (D0), the kinetic of viscosity of the solution ( ν ), and the RDE rotation speed ( ω ). 5 Based on a series of polarization curves under different rotation speeds and measurements of jl,c, the Levich-Koutechy equation allows us to extract jk (under 0.9 V versus RHE or other specific potentials), which can be further conversed into specific activity and mass activity by normalizing with ECSA and Pt (or other metal) loading. The catalytic capability of a material is generally evaluated by these two parameters. In addition, the half-wave potential (E1/2), which is the required potential to achieve a current that is half that of jl,c, is also widely used to describe the catalytic performance (Figure 1C). A higher E1/2 signifies a lower required overpotential to achieve 1/2 j l,c and thus reflects a higher catalytic activity. It is also noteworthy that the mass activity, especially for Pt-based catalysts, may be more significant to evaluate the ORR activity in consideration of the expensiveness of Pt. Researchers usually benchmark obtained ORR performance against the targets set by the United States Department of Energy (DOE). In these targets, a mass activity of 0.44 A mgPGM –1 (PGM: precious group metal) at 0.900 V in a PEMFC should be achieved before 2020. 28 As a common strategy, the shape and size control has been widely used in numerous catalytic systems to improve the catalytic performance of nanocatalysts. The ORR catalytic performance of different Pt facets, especially low-index planes (planes with low Miller indices), has been extensively studied. In strongly adsorbed electrolyte (such as H2SO4 solution), in which the strong adsorption of anion deactivates the Pt (111) surface dramatically, the ORR activity follows an order of Pt (111) < (100). While in weakly adsorbed electrolyte (such as HClO4 solution), the ORR active increases in the order of Pt (100) ≪ Pt (110) ≈ Pt (111). 27 Several high-index planes (planes with high Miller indices) have been demonstrated to exhibit higher ORR catalytic activity. For example, the Xia group prepared Pt concave nanocubes by a synthetic method with Br− as a capping agent to hinder the growth of the <100> axis. 29 With the help of high-resolution transmission electron microscopy (HRTEM), these nanocubes were confirmed to be enclosed mainly by {720} as well as {510} and {830} facets. A substantially enhanced ORR catalytic activity was observed on these Pt nanocubes. These concave Pt nanocubes exhibited approximately three and two times the specific activity at 0.90 V of the Pt cubes and cuboctahedra (bounded by low-index facets) in ORR, respectively. Unfortunately, despite the achieved high specific activity, the mass activity of Pt concave nanocubes is unsatisfying mainly caused by the relatively large size (>15 nm). In addition, Pt NPs exposing high-index facets with at least one Miller index being larger than 1 have also been found to show relatively high ORR activity; these include but are not limited to tetrahexahedron (hk0), trapexezohedron (hkk), and trisoctahedron (hhk). The enhanced ORR activity is generally ascribed to dense surface steps, edges, and kinks on high-index facets. 5 The main obstacle still lies in the difficulty in stabilization of these thermodynamically unstable shapes during long-term ORR catalysis operation in a more practical and large-scale synthesis.Generally, the activity of heterogeneous catalysts largely depends on the size of the metal particles. Reducing the particle sizes may boost the catalytic performance for multiple reasons. Cutting bulk materials into NPs brings a considerable portion of formerly inner atoms to surfaces where the catalysis reaction occurs. The smaller the NPs are the larger surface ration is. Meanwhile, small particles are more likely to possess dense low-coordinated species such as steps, edges, and kinks, which are more capable of achieving high catalytic performance because of the high surface free energy. Additionally, size reduction enhances metal-support interactions, which may rearrange the electronic structure of metal species and further promote the catalytic process. Thus, size reduction is widely regarded as one of the most effective strategies for improving atomic utilization and catalytic activity. Size effects of Pt in ORR have been deeply explored over the last several decades. 4,5 There is now substantial theoretical and experimental research showing that the Pt mass activity is optimized when the size of particles is reduced into the range of 2–5 nm. 30 When the size is further reduced, the mass activity is found to decrease with the decreasing of particle size. The reason for this unexpected phenomenon, however, is still in dispute. 5 Being blocked by surface atoms, the interior atoms in NPs can be hardly collided by reactants and hence rarely contribute to catalytic activity directly. Therefore, constructing hollow Pt NPs by removing the interior Pt atoms in solid NPs may retain the original catalytic activity and decrease the Pt usage simultaneously. Additionally, a hollow particle offers two sides, the inner surface and outer surface, which may further improve the atom utilization. By a template-removal method, the Adzic group prepared hollow NPs, which were further found to exhibit higher ORR catalytic activity compared with solid Pt NPs with similar sizes. 31 The enhanced catalytic performance, after careful investigation with experiments and density functional theory (DFT) calculation, was ascribed to lattice contraction induced by the hollow structure. A compressive strain is usually regarded to shift the Pt d-band center downward, weaken the adsorption of strongly adsorbed oxygenated intermediates, and finally improve the ORR catalytic activity. 32,33 Further, Zhang et al. prepared Pt nanocages with sub-nanometer thickness and investigated their ORR catalytic performance. 34 They first prepared Pd nanocubes with an edge length of ca. 18 nm and then deposited four atomic layers of Pt on the surface of Pd cubes by reducing Pt salt at 200°C. Finally, the Pd cube templates were selectively removed, leaving Pt cubic nanocages covered by {100} facets (Figure 1D). This method also allows them to prepare Pt octahedral nanocages covered by {111} facets by using Pd octahedral templates. The specific activity and the mass activity on octahedral nanocages were found to be five and eight times higher than those of commercial Pt/C, respectively (Figure 1E). Unlike previous research, this synthetic method provides the possibility of synthesizing Pt hollow structure with specific facets. He et al. prepared icosahedral Pt nanocages via a similar synthetic method, which also achieved outstanding ORR performance. 35 Benefiting from the large surface area and relative better stability than NPs and nanowire, one-dimensional nanocrystal is considered an ideal structure to exhibit outstanding performance in electrocatalysis. Liang et al. prepared a free-standing Pt nanowires membrane through a multistep templating route. 36 As expected, the Pt nanowires exhibited both higher mass activity of ca. 0.016 A/mgPt and better durability than the bulk Pt catalyst and commercial Pt/C. The Duan group prepared ultrafine jagged Pt nanowires with an average diameter of ca. 2 nm via a synthetic method with thermal annealing and electrochemical dealloying (Figure 1F). 8 Such Pt nanowires exhibited an extremely high ORR catalytic activity with E1/2 of 0.935 V versus RHE (more positive than that of Pt/C by ca. 75 mV), a mass activity as high as 13.6 A/mgPt, and an excellent long-term durability. The enhanced activity is ascribed to the atomic stress as well as ORR-favorable rhombus-configuration on the surface (Figures 1F–1H).Tremendous works have demonstrated that alloying two or more metals may empower catalysts with unique properties. As early as 1993, scientists had already realized that the kinetics of ORR could be easily enhanced by at least three times by simply alloying Pt with transition metals, such as Ni, Co, and Mn. 37 The atomic and electronic structures of Pt, as generally believed, can be improved by alloying to boost the ORR performance. Further, Markovic et al. systematically investigated the function of alloying metals by DFT calculation. 38 The specific activities of PtM alloys were found to display a volcano-type relationship with the d-band center (as we discussed before, the d-band center can sever as a descriptor in ORR), elucidating that very strong and very weak oxygen-intermediate adsorption will limit the reaction rate by the removal of surface oxides and electron and proton transfer to adsorbed O2, respectively (Figure 2 A). It is also noteworthy that Pt3Co, Pt3Ni, and Pt3Fe alloys dominate the top of the volcano and thus are supposed to exhibit higher ORR activity than that of other PtM alloys or pure Pt materials. Inspired by the above conclusions, scientists have devoted further efforts to tuning diverse PtM alloys into satisfactory ORR catalysts by modulating the component, shape, and size of PtM nanostructures. 5,10,39,40 For example, the Yang group prepared Pt3Ni truncated-octahedral nanocatalysts that dominantly expose {111} facets. 41 The particles were found to exhibit a 4- and 1.8-fold higher ORR mass activity than commercial Pt/C and normal octahedral Pt3Ni particles, respectively. They then raised a synthetic strategy of preparation of uniform icosahedral nanocrystals of other PtM (M = Au, Ni, and Pd) alloys. Their investigation showed that ORR catalytic activity can be further enhanced by using icosahedral Pt3Ni as a catalyst (Figure 2B). 42 Given that both octahedral and icosahedral Pt3Ni nanocrystals are bound by {111} facets, they speculated that the enhanced ORR performance may be ascribed to elastic strain. By using molecular dynamics simulations, they found a tensile surface strain on icosahedral particles but a compressive surface strain on the octahedral ones. These surface-strain differences, as they claimed, serve a vital role in the regulation of electronic structure of surface atoms and thus explain the enhanced ORR performance (Figure 2C). Since catalytic performance is sensitive to the surface structure, an in-depth investigation of the component, atomic, and electronic structure on the surface of the catalyst is of great concern. The Markovic group performed surface-sensitive techniques, including low-energy electron diffraction, low-energy ion scattering, Auger electron spectroscopy, surface X-ray scattering (SXS), and synchrotron-based high-resolution ultraviolet photoemission spectroscopy, 43 to investigate the Pt3Ni alloy surface. An oscillating structure of Pt3Ni, as shown in Figure 2D, was then proposed: the outermost layer is composed exclusively of Pt; the second layer is Ni enriched (52% of Ni content is larger than 25% of Ni content the bulk), and the third layer possesses a Pt-rich feature (87%). They also found that such a “segregation” surface structure is stable under a potential range from 0.05 to 1.00 V versus RHE, which was confirmed by in situ SXS measurements. With an unambiguous surface structure in mind, the d-band center and specific activities of different Pt3Ni surface morphology were investigated and summarized (Figure 2E), indicating the superior ORR catalytic activity of Pt3Ni (111). On the basis of the above achievement, the Huang group made a successful attempt to optimize ORR activity of Pt3Ni (111) by a surface-doping strategy. 44 They tried a series of transition metals and found that Mo could improve the ORR catalytic performance the best by acting as a doping metal. Their DFT calculation suggests that Mo-doping increases the oxygen-binding energies of the center sites on (111) facet, explaining the enhanced ORR activity (Figure 2F).Another instructive inspiration by the Markovic’s research is that PtM alloy particles may possess a compositional-segregation feature and thus exhibit a different property with a different configuration. Two representatives by the Strasser group have promoted our understanding of segregation on shaped alloy nanocatalysts. In 2013, they followed the morphological and compositional evolution of three octahedral PtxNi1−x alloy NP under electrochemical conditions by employing aberration-corrected scanning transmission electron microscopy (STEM) and electron energy-loss spectroscopy. 45 A leach in their facet centers and concave octahedral structure were observed. This dealloying and morphological evolution implies the complicacy with shape-selective alloying catalysts under operating conditions. In the following year, they did an intensive study on the element-specific anisotropic growth and degradation of PtM alloy nano-octahedra. 46 The results, summarized in Figure 3 A, forebode that a further enhancement in ORR activity may be achieved by rational synthesis of Pt alloy ORR electrocatalysts.Besides the Pt-Ni system, scientists have also investigated other PtM alloy or bimetallic systems. The Adzic group prepared a Pt/C catalyst stabilized by Au nanoclusters via an under-potential deposition method. 47 With in situ X-ray absorption near-edge spectroscopy and voltammetry data, the authors claimed that Au clusters raise the Pt oxidation potential and thus confer the stability during ORR operation. By reducing Pt salt with Pd nanoseeds, the Xia group prepared a Pd-Pt bimetallic nanostructure with Pt branches on a Pd core, which was found to exhibit high ORR activity at both room temperature and 60°C (Figures 3A and 3B). 48 The improved ORR performance was ascribed to the large surface areas and high-index facets exposure of Pt branches. Greeley et al. investigated ORR performance on alloys of Pt and early transition metals such as Y and Sc. 49 Particularly, polycrystalline Pt3Y showed higher ORR activity than pure Pt by a factor of 6–10. The Chorkendorff group synthesized eight Pt-lanthanide and Pt-alkaline earth catalysts and investigated their performance in ORR. 50 A volcano-like relationship between the ORR catalytic activity and the bulk lattice parameter was demonstrated (Figure 3C). Thus, the lanthanide contraction was indicated to be capable of modulating strain effects and enhancing the activity and durability of Pt-based catalysts in ORR. Bu et al. prepared hierarchical Pt-Co nanowires bounded by high-index and Pt-rich facets, achieving an ORR mass activity approximately 30 times higher than that of commercial Pt/C catalysts. 40 Constructing Pt-skin structure is also regarded as one of the promising strategies for simultaneously increasing ORR activity and lowering the usage of Pt. The performance of the Pt-skin may be regulated by controlling the composition, size, and shape of intimal metal. Here, we review some representative examples of Pt-skin catalysts with superior ORR catalytic activity. As we discussed in the above section, the component and morphology of Pt-Ni alloys can be protean during growth and catalysis process. Chen et al. prepared PtNi3 polyhedra with uniform morphology and size in oleylamine. 10 Then, they dispersed the oleylamine-capped PtNi3 polyhedra in nonpolar solvents (e.g., hexane and chloroform) and kept this dispersion at room temperature for 2 weeks. The PiNi3 polyhedra were found to gradually transform into Pt3Ni nanoframes (Figure 3D). The obtained Pt3Ni nanoframes were then dispersed onto a carbon support and heated to ca. 400°C in protective argon gas. Most Pt3Ni nanoframes evolved into smooth nanoframes with Pt-skin surface (denoted as Pt3Ni nanoframe/C). Compared with commercial Pt/C catalyst, the Pt3Ni nanoframes/C delivered a factor of 36 and 22 enhancement in mass activity and in specific activity, respectively. The open structure, the sufficient exposure of Pt(111), and the surface strain of Pt atoms were demonstrated to contribute to the enhanced ORR performance. Niu and co-workers synthesized Pt–Ni rhombic dodecahedra (RD) at a lower synthetic temperature than typically reported. 51 This lower synthetic temperature allows them to track the growth process during the prolonged synthetic period of time. They found the synthetic solution would turn from green to yellow, brown, and black in about 60 min. They collected the products at 3, 10, and 30 min after the solution turned black (denoted as RD-3, RD-10, and RD-30, respectively) and further selectively removed the Ni-rich phase within the collected products by chemical corrosion (denoted as RD-3-cor, RD-10-cor, and RD-30-cor, respectively). Thus, the morphology and dispersion of the Pt-rich phase were tracked. The frame-like morphology of RD-30-cor indicates the Pt migration from internal to external during Pt-Ni alloy growth. The RD-30-cor was also found to exhibit higher ORR activity than RD-3-cor, RD-10-cor, and commercial Pt/C catalyst. The Xia group coated Pd nanocubes with conformal thickness-controllable Pt shell by regulating the injection rate of Pt precursor and synthetic temperature. 52 The thickness of the Pt shell could be adjusted to 1, 2, 3, 4, or 6 atomic layers (Figure 3E). All four materials showed higher ORR activity than Pt/C, and as expected, the nanocube with one Pt atomic layer delivered the highest mass activity. Hunt et al. demonstrated a self-assembly synthetic approach that allows them to control the particle size, surface Pt layer coverage, and heterometallic composition of the final Pt-M core-shell architecture. 53 This synthetic method can be expanded to prepare a variety of core-shell systems. Even though the authors did not investigate ORR performance of the final core-shell particles, this method provides us with a golden opportunity to prepare Pt-skin catalysts with a controllable internal component, Pt-skin thickness and coverage, and doping by other novel metals. Recently, the Huang group reported PtPb/Pt core/shell nanoplates that exhibited superb ORR activity (Figures 3F and 3G). 54 At 0.9 V versus RHE, such a catalyst achieved a specific activity of 7.8 mA cm–2 and a mass activity of 4.3 A/gpt (Figure 3H). As revealed by DFT calculation, the edge-Pt and top Pt (110) facets were under large tensile strains, explaining the optimized Pt–O bond strength and the final enhanced ORR performance. Pd, iridium, ruthenium, and other novel metal-based materials have also been studied as ORR catalysts. 4 Despite tremendous efforts to optimize their performance, the best performance by them is barely equivalent to that of commercial Pt/C catalysts. In consideration of their expensiveness, it makes little sense to substitute Pt with these materials.Despite the high ORR activity, the cost and durability issues surrounding the Pt-based catalysts severely hamper their widespread commercialization. 2,5 The development of low-cost and reliable ORR catalysts is now of great concern. In living organisms, several specific enzymes (e.g., cytochrome C oxidase and laccase) activate oxygen molecules into an electron acceptor that captures the electrons from fuels and thus supplies the energy. Although these enzymes feature nonprecious metals as a catalytic active site, they have demonstrated remarkably reduced overpotential for the oxygen reduction compared with man-made catalysts. This proves that a considerable improvement in ORR activity by nonprecious-metal catalysts is not an illegitimate target. 12 During the past decades, various nonprecious-metal-based catalysts have been investigated for substituting expensive Pt-based catalysts in PEMFCs. Among them, nonprecious-metal-nitrogen-carbon composite (M-N-C), nonprecious-metal oxides, chalcogenides, and oxynitrides have been found to be potential candidates. 4,12 The premier employment of M-N-C composite as ORR catalysts can be traced back to the research by Jasinski in 1964, in which Co phthalocyanine was found to exhibit ORR catalytic activity. 55 After that, tremendous efforts have been devoted to the development of M-N-C ORR catalysts. Generally, there are three common strategies: (1) increasing the intrinsic activity by composition modulation, (2) confining metal species into carbon shells to preserve metal from corrosion (from acidic medium and oxidizing potentials), and (3) increasing the accessibility by constructing porous or other large specific-surface-area structures.Wu et al. prepared nonprecious-metal catalysts with Fe and Co confined in multiple C–N shells by using polyaniline (PANI) as a carbon-nitrogen precursor. 56 Their best catalyst showed further improved ORR activity that was closer to that of commercial Pt/C (Figure 4 A) and an enhanced durability in both long-term RDE and PEMFC tests, which was most likely generated from the protection of C–N shells. In 2009, the Dodelet group reported Fe-based catalysts with enhanced ORR activity. 57 A mixture of ferrous acetate, carbon black, and phenanthroline was pyrolyzed in argon and ammonia successively after ball-mill. The obtained material was applied in PEMFC as a cathode catalyst and was found to exhibit improved initial ORR activity, which is competitive in comparison with Pt-based catalysts but has unsatisfying stability. Later, this group prepared another Fe-based catalyst by using zeolitic imidazolate framework (ZIF)-8 as a microporous host for phenanthroline and ferrous acetate. 58 Compared with the previous one, this catalyst achieved an obvious enhancement of ORR activity in H2-O2 PEMFC with a power density of 0.75 W cm−2 at 0.6 V (Figure 4B). They then converted their data into volumetric activity of cathode to compare the obtained activity with that of the US DOE’s targets. The extrapolation of the Tafel slope (at 0.8 V) demonstrated that the volumetric activity by their catalysts was close to the target proposed by the US DOE (300 A cm−3 for non-PGM-based catalysts; Figure 4C). Unfortunately, an apparent activity decay of 15% after 100 h of operation demonstrated the unqualified durability. The Bao group prepared pea-pod-like carbon nanotubes (CNTs) encapsulating Fe NPs (Figure 4D) and found that such a confinement structure can deliver an enhanced durability during long-term PEMFC test. 59 They suggested that a charge transfer from the encapsulated Fe cluster to the carbon tube turn the carbon atoms near Fe cluster into ORR active species so that ORR can proceed without the immediate contact between O2 and Fe atoms (Figures 4E and 4F). Thus, the enhanced durability can be explained because the carbon shells may preserve the Fe clusters from acid corrosion. The ORR activity of the encapsulated Fe NPs, however, still leaves much to be desired. In a typical H2-O2 PEMFC, the encapsulated Fe NPs yielded a voltage of ca. 0.5 V at a current density of 0.10 A cm−2, which was only ca. 60% of that given by Pt/C. Hu et al. prepared Fe3C NPs encapsulated by graphitic layers through a high-pressure pyrolysis process. In both acidic and alkaline electrolyte, this catalyst showed remarkable stability. 60 The outer graphitic layers were also believed to play a protective role in stabilizing inner particle under corrosive conditions. While in acidic electrolyte, the as-prepared catalysts only achieved E1/2 of ca. 0.73 V, largely underperforming Pt/C.Given that a larger specific surface area usually guarantees a greater accessibility, constructing porous structure is generally considered a remedy for the nonprecious-metal-based catalysts that possess relatively low intrinsic activity. The Qiao group developed CNTs with Fe–N decoration from hierarchically porous carbon. 61 Such a material was believed to possess desired merits that included high activity by Fe–N species, facile transportation from large pores, and adequate active-site exposure from large surface area. As expected, an improved ORR performance comparable to that of Pt/C was achieved in alkaline solution. Liang and co-workers prepared a series of mesoporous nonprecious-metal catalysts. Among them, a Co-based mesoporous catalyst, which is fabricated with vitamin B12 as the Co precursor and ordered mesoporous silica SBA-15 as the template, possesses the largest surface area of 568 m2 g−1 (Figure 4G). 62 This mesoporous catalyst showed an outstanding ORR performance in acidic solution with E1/2 of 0.79 V versus RHE, an electron-transfer number of ca.3.95, and as excellent durability. The large surface area and the homogeneous distribution of abundant Co–Nx were claimed to contribute to the ORR performance.Selectivity is another noteworthy issue existing in ORR on nonprecious-metal catalysts. Several works have demonstrated the non-negligible and undesired side products, such as H2O2 and O2 −, in the ORR process on nonprecious-metal catalysts. Recently, the Gewirth group presented their idea that the kinetics of proton transport in the ORR catalysts, to some extent, determine the product distributions. 63 They fabricated a hybrid bilayer membrane with Au electrode modified with a self-assembled monolayer of Cu-based ORR catalyst and a monolayer of lipid consisting of proton carrier. Such a hybrid bilayer membrane allowed them to quantitatively regulate kinetics of proton transport to the catalyst by modulating the amount of proton carrier in lipid. As a result, an insufficient proton carrier results in proton transport that is too slow, and thus O2 would be reduced by 1 e− to O2 −; an excessive proton carrier results in proton transport that is too fast, and the ORR process would undergo a 2 e− pathway with H2O2 as the products (Figure 4H). Thus, a mismatch between proton and electron transport causes unfavorable products (O2 − or H2O2), and a commensuration between proton transport and O–O bond breaking rate ensures 4 e− ORR pathways with H2O as the product.Several nonprecious-metal oxides have also been found to be catalytically active for ORR. Inspired by the biological catalyst in photosystem, the Jaramillo group synthesized Mn oxide thin film through an electro-deposition method. 64 This Mn oxide exhibited E1/2 of 0.73 V, which is more negative than that of Pt/C by 130 mV in alkaline solution. Later, Cheng et al. found that the ORR activity on MnO2 can be improved by introducing oxygen deficiency generated by high-temperature treatment in air or argon. 65 A modified surface-oxygen interaction and a reduced reaction barrier by oxygen deficiency, as suggested by DFT calculation, were claimed to contribute to ORR performance. Recently, the Dong group reported free-standing tubular monolayer superlattices of hollow Mn3O4 nanocrystal (h-Mn3O4-TMSLs). 66 The obtained h-Mn3O4-TMSLs delivered outstanding ORR performance in alkaline solution with an onset potential of ca. 0.91 V and E1/2 of 0.84 V (versus RHE), which is about 10 mV more negative than that of Pt/C. The mesoscale tubular geometry was believed to enhance mass transport, and the monolayer superlattice structure may be beneficial for molecular accessibility. The Dai group grew Co3O4 nanocrystals on reduced graphene oxide (RGO) and found such a hybrid material can sever as an efficient ORR catalyst in alkaline solution. 67 The E1/2 by this hybrid material was ca. 0.83 V versus RHE, similar to that of Pt/C. Because neither Co3O4 nor graphene could hardly catalyze oxygen reduction, the enhanced ORR performance was ascribed to the synergetic chemical coupling effects of the hybrid structure. They further prepared a cobalt-oxide-carbon-nanotube hybrid, which showed an ORR onset potential of 0.93 V in 1 M KOH solution. 68 This hybrid material was also found to be active and stable in 10 M NaOH at 80°C. Several perovskite materials are also active for ORR. The Shao-Horn group investigated ORR activity of a series of perovskites and found that the ORR activity correlates to σ*-orbital occupation and the extent of covalency between B-site metal and oxygen. 69 The Dai group fabricated Co1–xS-RGO hybrid material through a solution-phase process followed by solid-state annealing treatment. 70 This hybrid material showed an active ORR performance with onset potential of 0.87 V versus RHE. The small size of Co1–xS NPs and the strong electrochemical coupling between RGO and Co1–xS NPs were believed to promote the ORR performance. Recently, the Xu group constructed honeycomb-like porous carbons with nitrogen and sulfur dual doping and Co9S8 NPs immobilized inside. 71 They investigated the ORR activity of a series of such materials after calcination under different temperatures. The best catalyst exhibited an ORR performance with an onset potential of −0.05 V and E1/2 of −0.17 V versus Ag/AgCl. The sufficient accessibility from a honeycomb-like structure and the synergetic interactions between Co9S8 particles and the support may explain the enhanced ORR performance. Cao and co-workers prepared CoxMo1–xOyNz supported on carbon, which exhibited onset potentials of 0.918 and 0.645 V in 0.1 M KOH and 0.1 M HClO4, respectively. 72 Other nonprecious compounds such as MnOOH, TiO2, NbO2, ZrOxNy, TaOxNy, and CoSe2 have also been widely investigated. 4,5 In addition, metal-free catalysts, especially N-doped carbon materials, have been reported active for ORR. 73,74 Unfortunately, these current nonprecious-metal-based and metal-free catalysts still leave much to be desired. 5 In particular, because acidic PEMFC is currently much more prevalent than alkaline counterpart, the susceptibility to acid makes nonprecious-metal-based catalysts, especially the oxides, suffer from poor durability in acid medium and limits their widespread adoption in PEMFCs. 12,13 With the development of highly advanced characterization techniques, there is growing awareness that the single metal sites with M–N coordination in nonprecious-nitrogen-carbon composite, especially the porphyrin-like FeN4C12 moieties, may be capable of catalyzing the 4 e– pathway reduction of oxygen to water. In one important work reported by Zitolo et al., Fe–N–C catalysts quasi-free of crystallographic Fe were synthesized by a thermal treatment in either Ar or NH3. 75 These catalysts had the same Fe-centered moieties but a much higher activity and basicity for NH3-treated Fe–N–C. After a detailed XANES study, two modes of FeN4 porphyrinic different O2 adsorption were identified. The authors noticed that it was difficult to integrate the porphyrinic moieties into graphene sheets, in sharp contrast to Fe-centered species assumed for pyrolyzed Fe–N–C. Such findings not only enlighten bottom-up synthesis strategies of M–N–C SACs but also underline the crucial role of the interaction between single metal atoms and support in ORR catalysis.Meanwhile, SACs have been widely assumed to possess high catalytic activity because of the minimum size of metal species and unique coordination structure. A large number of studies have demonstrated that SACs can exhibit distinctive catalytic performance for a wide variety of catalytic systems. Especially given that SACs may achieve an atomic economy of 100% atom utilization, SACs are logically considered potential ORR catalysts. Though much effort has been devoted, there are still hurdles to overcome before SACs become qualified ORR catalysts in practice. As practical ORR in PEMFC requires high catalytic activity in given geometric area, one challenge is how to improve the metal loading in SACs (typically below 1 wt % for the majority of SACs) or how to promote the accessibility of metal sites to increasing the catalytic activity per given area. Certainly, another challenge is how to improve the intrinsic activity of individual active metal sites. Correspondingly, two strategies are generally used to optimize the ORR performance of SACs. One is to create appropriate supports with a larger specific area to anchor or expose more single-metal sites, with the aim to provide more catalytically active sites. The second one is to modulate the electronic structure of metal sites by tuning the coordination environment or doping heteroatoms with the aim of optimizing the intrinsic catalytic activities. Here, we summarize recent advances in SACs for ORR catalysis and demonstrate how scientists attempt to overcome the above-mentioned issues by constructing rational supports, regulating local coordination environment over single metal sites, or doping heteroatoms.As we discussed before, an ideal support ought to provide dense coordinative sites to anchor sufficient isolated metal atoms, as well as large specific surface areas or porous structures for superior accessibility. Porous ZIFs with intrinsically isolated metal nodes and N-contained organic ligands are thus considered a potential precursor for SACs. In 2016, Yin et al. originally reported a facile and effective approach to prepare Co SAC with an extremely high metal loading over 4 wt % via thermal treatment of bimetallic Zn/Co MOFs (Figure 5 A). 20 One significant breakthrough of this work is the elegant use of Zn-Co ZIF for isolating single-metal atoms. Another noteworthy contribution is the greatly enhanced metal loading of the SACs prepared by this approach, highlighting the future direction for practical applications. After this initial disclosure, a variety of protocols for the synthesis of SACs via ZIFs have been reported. For example, Wu and Shao reported a chemical replacement method for constructing Fe-doped ZIF precursors by using an ionic exchange method. 76 In this ZIF system, Fe ions can partially substitute Zn ions and bond with imidazolate ligands in three-dimensional frameworks, forming FeN4 complexes. The as-prepared Fe catalyst showed an outstanding ORR activity in acidic media, with E1/2 of 0.85 V, as well as an enhanced stability with a decay of only 20 mV in E1/2 after 10,000-cycle operation. Recently, Li et al. reported a two-step synthetic method to prepare Mn-based SACs with highly dense MnN4 sites. In the first step, a partially graphitized carbon host was prepared by carbonizing Mn-doped ZIF-8 precursors. In the second step, additional Mn and N sources were then adsorbed into the obtained microporous carbon host, followed by a thermal activation, to increase the density of Mn sites. Measured by inductively coupled plasma mass spectrometry (ICP-MS), the Mn content was found to be more than 3 wt %. In 0.5 M H2SO4 electrolytes, this catalyst delivered an ORR performance with E1/2 of 0.80 V and encouraging durability (Figure 5B). 77 In 2017, Li et al. developed a cage-encapsulated-precursor pyrolysis strategy to generate a highly stable isolated Fe SAC to endow excellent ORR performance. 78 In this case, ZIF-8 was used as a molecular-scale cage to isolate and encapsulate the metal precursor Fe(acac)3 because of its special pore size and cavity. After pyrolysis treatment, the ZIF-8 was converted into nitrogen-doped porous carbon, and the Fe(acac)3 was reduced, forming isolated Fe sites anchored on N species. This SAC had a high Fe loading of 2.16 wt % and exhibited a highly efficient activity toward ORR in alkaline media with an E1/2 of 0.900 V and an exceptional J k of 37.83 mA cm−o at 0.85 V, superior to that of commercial Pt/C. To improve the electron conductivity, the Lin group employed surface-functionalized multiwalled CNTs as a template during the synthesis of Fe-Zn bimetallic ZIFs. After the pyrolysis process, a network of N-doped carbon with atomically dispersed Fe atoms was prepared. The as-prepared Fe SAC gave an ORR performance with E1/2 of 0.81 V in 0.1 M HClO4, and a power density of 620 mW cm−2 in H2-O2 PEMFC. 79 The use of polymers for the synthesis of SACs has also been explored, as the isolated metal atoms can be effectively stabilized by a coordination effect with the nitrogen atoms of N-doped carbon supports In 2018, Li group has offered a polymer encapsulation strategy to prepare SACs supported by porous nitrogen-doped carbon nanospheres. 80 In brief, metal precursors were initially encapsulated in polymers by simply mixing the metal acetylacetonate complexes with monomers during the polymerization process. Then, a pyrolysis process was performed at a high temperature to give the polymer-derived porous nitrogen-doped carbon nanospheres with single metal atoms dispersed uniformly. This approach was found to be applicable to both noble and nonprecious metals. They noticed that the Co SAC showed the best performance among other SACs prepared by this approach and delivered a comparable ORR activity (E1/2 = 0.838 V) and J k at 0.83 V to Pt/C in alkaline media, a good methanol tolerance, and an exceptional cycling stability even after 5,000 cycles. Similarly, a metal-organic polymer supramolecule strategy was introduced by Li and Guo for the construction of a SAC by “self-locking” between metal ions and a natural polysaccharide, sodium alginate (SA). 81 SA has a great number of hydrophilic groups (–COOH and –OH) in α-L-guluronic (G-block) and β-d-mannuronic (M-block) acid units. It can bond with Fe ions to form a hydrogel, which leads to the formation of atomically dispersed Fe-Nx sites in highly porous, sheet-like structures. The resulting Fe SAC exhibited excellent ORR performance in 0.5 M H2SO4 and 0.1 M KOH, along with an admirable durability.Hollow and two-dimensional materials, as a result of high accessibility, are also considered efficient supports for SACs. The Li group reported an interesting template-assisted pyrolysis treatment to access Co SAC dispersed on hollow N-doped carbon spheres (Figure 5C). 82 The single Co sites and the hollow carbon spheres collectively contributed to the excellent ORR performance in acid media. Later, they fabricated N-, P-, and S-co-doped hollow polyhedron with embedded single Fe atoms through the Kirkendall effect process (Figure 5D). 83 In 0.1 M KOH solution, this Fe SAC achieved an outstanding ORR performance with E1/2 of 0.912 V, J k of 71.9 mV cm−2 at 0.85 V, and a record-level Tafel slope of 36 mV dec−1. In 2017, the Bao group described a simple method to prepare a highly dispersed single Fe catalyst by ball milling of iron phthalocyanine (FePc) and graphene nanosheets (GNs). 84 The resultant FeN4/GN showed a high ORR activity in alkaline electrolyte, similar to that of the commercial Pt/C. Importantly, it had a higher stability and resistance to SOx, NOx, and methanol than Pt/C. DFT calculations demonstrated the excellent ORR performance and stability result from the unsaturated Fe centers confined in the graphene nanosheets via four N atoms. Cao et al. introduced a surfactant-assisted approach for the preparation of an Fe SAC (Figure 5E) by pyrolyzing the layered-like precursor formed by Fe-loaded water-soluble surfactant F127 and g-C3N4. 85 They found that the use of surfactant F127 enabled the uniform dispersion of Fe atoms to form Fe–Nx sites on the support. Additionally, the Fe-doped F127 sheets could strongly anchor on the g-C3N4 so that the Fe clusters could be easily removed by acid treatment. They demonstrated that the ORR activity of Fe SAC stemmed from the Fe-pyrrolic-N4 active sites. Its E1/2 was only 30 mV less than 20% Pt/C in acidic medium. For the H2-O2 PEMFC testing, it produced a current density of 0.85 A cm−2 at 0.6 V and 3.34 A cm−2 at 0.2 V and achieved the maximum power density of 823 mW cm−2. They attribute the good performance of the PEMFC to the accessible and high-density active sites on N-doped carbon nanosheets. Recently, Baek and co-workers synthesized a Cu SAC (up to 20.9 wt %) with isolated Cu atoms distributed in ultrathin nitrogenated carbon nanosheets. 86 In their study, the use of L-glutamic was important for high Cu content in SAC because it could introduce additional N species during the synthesis process and effectively trap the metal atoms on the support. Moreover, dicyandiamide could react with the carboxylic acid groups in a trimesic acid to yield a two-dimensional sheet. Because of the synergetic effect between ultrathin nanosheet and high content Cu atoms, a favorable adsorption of O2 and OOH could be obtained. For the ORR testing, it showed over 54 times higher mass activity than Cu NPs at 0.85 V. In addition, it also delivered a lower Tafel slope (37 mV dec−1), higher methanol and carbon monoxide tolerance, and a longer-term stability than commercial Pt/C. The free energy diagrams of ORR pathway on CuN2 and CuN4 were further studied. They showed that under 0.4 V, the reaction on the CuN2 proceeded through a thermodynamically downslope route, implying an excellent electrochemical ORR activity. For CuN4, there was an upslope of 0.75 eV, meaning that the rate-determining factor is the relatively weak adsorption of O2 for ORR on CuN4.Several biomaterials are instinctively potential supports for SACs as a result of porosity. Dai and co-workers demonstrated that thermal treatment of unsubstituted phthalocyanine-FePc complexes within micropores of cattle bones can give atomically dispersed Fe atoms on hierarchically structured porous carbon frameworks (Figure 5F). 87 The Fe SAC catalyst showed ORR performance comparable to that of the Pt/C in 0.1 M HClO4 (E 1/2 = 0.81 V) and an improved long-term durability (7 mV negative shift after 3,000 cycles). Under alkaline conditions, it outperformed the Pt/C in terms of activity (E 1/2 = 0.89 V) and long-term durability (1 mV negative shift after 3,000 cycles).Modulation of the electronic properties of the metal center has also been demonstrated to be an effective route for improving catalytic performance of M–CN catalysts. 88 Generally, two approaches can affect the electronic properties of the active metal sites: one is regulating the center metal element, the species, and/or the number of the coordination atoms; another is using long-range interactions between metal sites and doped atoms on the support materials to adjust the electronic structures. For example, doping with heteroatoms (e.g., N, S, and P) also helps to tune the electronic structures to substantially improve the catalytic activity. 5 Fe-based SACs are most commonly employed for ORR, and the catalytic properties are strongly dependent on the types of metal−N2/N4 conformation. However, the problem of the existing FeNx-based catalysts is that the FeN4 is relatively stable sites, which might not be the most active sites based on theoretical predictions due to a strong interaction with O2* and OH*. 89 In this regard, Guo and co-workers reported a viable template casting method to access isolated FeN2 species on N-doped ordered mesoporous carbon. 90 One significant improvement of this approach is that the Fe precursor can be anchored on the surface of the template (SBA-15). DFT calculations show that the FeN2 outperforms FeN4 because of its lower interaction with OH* and O2* intermediates, along with improved electron transport. Interestingly, another work by Wang and co-workers demonstrated that FeN4 species can also be remarkably active for ORR by mini-trim on the atomic configurations of Fe-N–C moieties. 91 The authors constructed atomically isolated Fe atoms on three-dimensional hierarchically porous carbon. In spite of merely 0.20 wt % of Fe metal loading, this SAC is highly efficient for ORR with E1/2 of 0.915 V in 0.1 M KOH, exceeding those of Pt/C (E1/2 = 0.85 V) and most M–N–C catalysts. Importantly, the atom-utilization efficiency in this study was superior to most previous reports. The experimental and DFT results demonstrate that the hierarchical carbon pores can effectively tune the electronic structure of FeN4 by modulating the local coordination of pyridine N. This leads to the selective cleavage of C–N bond near Fe centers, giving edge-hosted FeN4 moieties to lower the ORR barriers to obtain exceptional catalytic activity and durability (Figure 6 A). Through modulating the electronic properties of the central metal with chlorine ions, the Li group originally developed a FeCl1N4 site catalyst (Fe loading ∼1.5 wt %) by a thermal-migrating method in 2018. 92 The FeCl1N4/CNS showed a superior E1/2 of 0.921 V in 0.1 M KOH, 79 mV more positive than that of Pt/C (Figures 6B and 6C). Moreover, it had an excellent J k of 41.11 mA cm‒2 at 0.85 V and an admirable cycling stability for 10,000 cycles, superior to those of most of the reported nonprecious-metal electrocatalysts. DFT calculations were employed to investigate the effect of chlorine coordination and sulfur doping on the ORR. The catalyst showed a much lower overpotential of 0.44 V than FeN4/CN but a higher binding energy of O2 (E b = −0.64 eV). They demonstrated a volcano curve showing the relationship between ORR overpotentials and O2 binding energies. That is, a higher O2 binding energy would lead to the difficulty in desorbing OH species, and weaker O2 binding may make the hydrogenation of O2* more sluggish. This suggests that the FeCl1N4/CNS has a moderate charge state and thus shows the highest ORR performance. The results reveal the near-range interaction with chlorine and the long-range interaction with sulfur of the Fe active sites contribute to the modulation of the Fe electronic structure. This work demonstrated the importance of modulation of electronic structure in the design and synthesis of SACs on their catalytic performance, an important future direction for this field.In another case, the Qiao group introduced a variety of graphitic carbon nitride (g-C3N4) coordinated transition metals (M–C3N4) for ORR. The g-C3N4 was used as an efficient support to construct a series of M–C3N4 electrocatalysts. They studied the Co–C3N4 in ORR and oxygen evolution reaction (OER) in alkaline media and found excellent activity stems from CoN2 in the g-C3N4 support, along with an optimal d-band of the catalyst. 93 In addition to Fe, Co, Ni, and Mn SACs, carbides of groups have also been investigated as alternatives for precious electrocatalysts in PEMFCs. However, the low density and stability of these active sites greatly hampered their catalytic performance. To address the problem, Chisholm and co-workers originally prepared a new class of single niobium atoms-based carbide catalyst using an arc-discharge approach (Figure 6D). 94 Single niobium atoms and ultra-small clusters stabilized in graphitic layers are identified as active sites for catalyzing the cathodic ORR. They found this unique structure greatly enhanced the conductivity to accelerate the exchange rate of ions and electrons, and strongly suppressed the chemical and thermal coarsening of the active species. Importantly, the single niobium atoms stabilized within the graphitic layers can redistribute d-band electrons and therefore significantly facilitate O2 adsorption and dissociation.Heteroatom-doped M–N–C has also shown great potential in improving the ORR performance to substitute precious-metal-based catalysts. 95 The S element can be easily introduced to the carbon support by thermal treatment of S-containing species. The relatively large atomic radius of p-block S element may create defects on the support, and the electronegativity of S may tune the electronic properties of M–N sites. In 2018, the Li group described an interesting pyrrole-thiophene copolymer pyrolysis strategy for constructing Fe SAC on S and N-codoped carbon with a loading of 0.947 wt %. 96 The S and N contents could be changed by controlling the precursors. Interestingly, the catalytic efficiency of the SAC displayed a volcano-type curve with increasing amounts of S employed. The E1/2 of Fe-ISA/SNC was shown to be 0.896 V, and J k was 100.7 mA cm−o at 0.85 V, superior to that of Pt/C. X-ray absorption fine structure analysis and DFT showed that the incorporated S could elegantly tune the charges surrounding the Fe active sites. This makes the rate-limiting reductive release of OH* more favorable to give the enhanced ORR activity in alkaline conditions. Overall, this work shows a detailed understanding of the effects of heteroatom doping on the activity of SACs. In another important case, Xie and co-workers achieved isolated Fe–Nx on N- and S-co-decorated hierarchical carbon layers to serve as a bifunctional OER and ORR SAC. 97 In their study, vertically aligned CNTs were employed to stabilize catalytic active sites. The isolated single Fe sites on N- and S-co-decorated hierarchical carbon layers were obtained by coating CNTs with 2,2-bipyridine and Fe salt, followed by pyrolysis and acid-leaching steps. Because of the abundance of active Fe sites, three-dimensional conductive networks, and unique hierarchical structure of the support, the Fe SAC catalyst showed an exceptional electrocatalytic performance for ORR and OER in alkaline conditions. By employing this Fe SAC, the polarization curves of Zn-air exhibited an open circuit voltage of 1.35 V and a charge-discharge voltage gap a little lower than that of Pt/C. The maximum power density was as high as 102.7 mW cm−2, and the cycling stability was excellent.Among the recent research on SAC for ORR, there are three works worth emphasizing. In the first work by the Zelenay Group in 2017, FeCl3, a polymer (PANI), and a simple organic compound (cyanamide, CM) were deliberately employed as precursors to endow the final Fe SAC (denoted as (CM+PANI)–Fe–C) with pore structures and high activity (Figures 6E and 6F). At voltages higher than 0.75 V in the H2-air PEMFC test, this Fe SAC achieved almost the same current densities as those obtained with a Pt cathode with a loading of 0.1 mgpt cm−2, underlying the great potential of SAC in practical PEMFCs. 98 Later, Wang et al. described a double-solvent approach to creating Fe–Co dual sites on an N-doped porous carbon support (Figure 6G). 99 The Fe3+ species were reduced and bonded with the adjacent Co atoms. Aberration-corrected high-resolution TEM, X-ray absorption fine structure spectroscopy, and Mössbauer spectroscopic measurements were performed to confirm the Fe–Co dual sites coordination environment. The as-obtained (Fe,Co)/N–C catalyst possessed excellent ORR performance (2.842 mA cm−2 at 0.9 V) in 0.1 M HClO4 solution, along with comparable onset potential (1.06 V) and E1/2 (0.863 V) and a superior cycling stability of 50,000 cycles. Remarkably, this catalyst reached maximum power density values of ∼0.98 W cm−. in H2/O2 PEMFC and 0.51 W cm−P in H2/air PEMFC. Moreover, constant-current operation testing showed that the working voltage of this catalyst was maintained even after a 100 h of operation. This Fe–Co dual-site catalyst outperformed the previously reported Pt-free catalysts in an H2/air PEMFC. DFT calculation showed that the activation of O–O was favored on the (Fe,Co)/N–C dual sites; therefore, the dual sites could greatly decrease the cleavage barrier of O–O bond to give high ORR activity and selectivity to the 4 e− reduction pathway. This work not only reports an outstanding PEMFC performance by Pt-free catalyst but also demonstrates that the introduction of metal-metal bonds may be a feasible strategy of upgrading SACs into more competitive alternatives to Pt-based catalysts in PEMFCs. The third one reported a convenient synthetic method to prepare SACs from bulk metals. In this approach, ammonia gas first trapped metal atoms in upstream bulk metals by the strong Lewis acid-base interaction. Then, the M(NH3)x species were captured by the defects on the downstream supports (e.g., pyrolyzed ZIF-8), leaving isolated metal sites (Figure 6H). The as-prepared CuN4 SACs exhibited superior ORR performance with an E1/2 of 0.895 V in 0.1 M KOH solution. This synthetic strategy may serve as guidance for efficient preparation of SACs directly from bulk metals and demonstrates the potential for scaling up SACs toward industrial applications. 100 Despite the above achievement we reached, our ultimate goal—a highly active, stable ORR catalyst in PEMFCs with affordable costs—has yet to be achieved. For Pt- and other precious-metal-based catalysts, the primary drawback is still the prohibitively high cost. Although in RDE testing, some advanced catalysts have already achieved high mass activities even outdistancing 0.44 A mgPGM −1 at 0.9 V, 10,44,54 few of them can also deliver such extraordinary performance in a practical PEMFC, leaving the 2020 DOE targets yet to be realized. 101 One recent work by the Liu group provides a silver lining. As reported, a Pt-Co catalyst with ultra-low Pt content reached an excellent performance in H2-O2 PEMFC test with a current of 1.77 A mgPt −1 at 0.9 V, exceeding the DOE 2020 target by approximately four times (Figure 6I). 102 Even though the use of extremely expensive Pt precursor and complicated synthetic process lessen the significance, this work demonstrates the feasibility of Pt-based ORR catalysts. However, for nonprecious-metal-based ORR catalysts, it seems that SACs possess greater competitiveness than conventional metal-particle-based materials mainly because of the relatively high activity and, in some cases, the tolerance to acid (while the intrinsic reason has yet to be revealed). Even so, state-of-the-art ORR SACs can still not meet the DOE targets. Thus, to further promote the activity is still the primary research direction for SACs. To fully realize the potential for ORR in PEMFC applications of SACs, more detailed and in-depth research is essential. Detailed structural studies at the atomic level are crucial for us to understand the factors influencing the properties and performance of SACs. So far, the pivotal characterization techniques for SACs are typically based on aberration-corrected HAADF-STEM and synchrotron radiation facility. These techniques are expensive and not easily available for every researcher in the community. Even so, the information we are able to collect by these advanced techniques still leaves much to be desired. In particular, efficient in situ characterization techniques, which enable us to observe catalysts under practical operation, are now urgently needed. In situ TEM, 103 XAS, 104 and other characterization techniques 105 have already been reported to survey the formation process or catalytic behavior under practical conditions. 106 With the help of these techniques, we may understand the relationship between catalyst structures and performance on a more profound level and thus prepare improved catalysts accordingly. 107 In addition, given that commercialization usually requires low synthetic cost, another future research direction for ORR SACs would be to develop efficient synthetic strategies that allow us facilely prepare SACs on a larger scale. Several works have demonstrated the preparation of SACs directly from metal bulks. 100,108 Further research may focus on developing more facile methods that avoid using high temperature or other energy-intensive treatments.This work was supported by the National Key R&D Program of China 2017YFA (0208300 and 0700104) and the National Natural Science Foundation of China (21522107 and 21671180).All authors devised the concept and built the framework of the review. X.W. and Z.L. wrote the manuscript. X.W., Z.L., T.Y., and W.W. organized the figures. X.W., Z.L., Y.W., and Y.L. edited and reviewed the manuscript. All authors approved the final version of the manuscript. X.W. and Z.L. contributed equally.

Platinum (Pt)-based catalysts have been unanimously considered the most efficient catalysts for the oxygen reduction reaction (ORR) in proton-exchange membrane fuel cells (PEMFCs). Unfortunately, the exorbitant cost of Pt hampers the widespread adoption and development of PEMFCs. Scientists have devoted tremendous efforts to achieving higher catalytic activity with less Pt usage by constructing delicate nanostructures. Substituting Pt with cheaper metals may be a feasible solution but suffers from a relatively low intrinsic activity. Recently, single-atom catalysts (SACs), which possess the highest metal utilization and excellent activity because of the minimum size of metal and unique coordination structure, are developing rapidly and have been regarded as a potential alternative to Pt-based materials. Here, we review the development of conventional Pt- and nonprecious-metal-based ORR catalysts and summarize recent achievement in SACs for the ORR. A brief perspective on the remaining challenges and future directions of SACs is also presented.

The depletion of fossil reserves in the last century and the high volume of emissions generated have led to search and develop more environmentally friendly alternative energy sources. In this sense, biomass is an attractive alternative since it is the only source from which both energy and chemicals can be obtained, making it the only one with enough potential to replace fossil fuels completely [1]. Even when biomass is widely distributed on Earth, the type of biomass must be carefully selected since, in some cases, the biomass used to obtain energy and chemicals can compete with the food supply [2]. This fact could cause an increase in prices due to the reduction of farmland for food and speculation leading to social imbalances. Among the wide variety of biomass sources, lignocellulosic biomass has emerged as a relevant sustainable source since it does not interfere with the food chain, it is very abundant, and it aims to valorize agricultural residues to produce high value-added chemicals [1,3]. Lignocellulose is generally composed of cellulose, hemicellulose, and lignin, as well as some minor non-structural components, such as proteins, chlorophylls, ash, waxes, tannins (in the case of wood), and pectin (in most fibers) [4]. Focusing on hemicellulose, which displays a disordered and branched structure (with short lateral chains) and low molecular weight, this can be relatively easy to hydrolyze in their respective monomers (pentoses), mainly xylose and arabinose [5].Then, these C5 monomers can be dehydrated through acid catalysis to give rise to furfural (FUR) as the main product [6,7]. This molecule is considered one of the most interesting of the sugar platform in biorefineries, with an annual world production of about 430,000 tons [8,9]. FUR possesses numerous industrial applications, such as fungicide, nematicide, specialized adhesive, flavoring compound, and for the recovery of refinery lubricants, although FUR is mainly employed as a precursor for a wide variety of high value-added chemicals that can be obtained through hydrogenation, oxidation, dehydration, condensation or opening-ring reactions [10–12]. Among them, the hydrogenation of furfural is the reaction that possesses the greatest commercial interest. Thus, it has been reported that about 62% of FUR is employed to synthesize furfuryl alcohol (FOL) by hydrogenation [12]. Commercially, this chemical transformation has been carried out by using copper chromite as the catalyst for more than 80 years [13–16]. However, the stringent environmental regulations related to Cr species, together with the high susceptibility of these catalysts to undergo deactivation [17], have led researchers to develop highly active and selective Cr-free catalysts [12]. In this sense, several authors have demonstrated that non-noble metals, such as Cu-, Ni-, and bimetallic-based catalysts [11,12,18], as well as noble metals (Ru, Pt, Pd, Au) [19–25] can attain a suitable catalytic performance. As FUR is highly reactive, both the hydrogenating character and the acid/base properties of catalysts need to be modulated. For example, even when Ni-based catalysts are highly active, they are not very selective in many cases because their active centers interact with both the carbonyl group and the aromatic furan ring [26–30]. However, Cu-based catalysts have displayed a lower hydrogenating capacity than the Ni ones, which implies that the active sites only interact with the carbonyl group [26,31–33], in such a way that the formation of furfuryl alcohol and 2-methylfuran (MF) is favored [26,31,34–38]. Fortunately, both compounds are considered valuable FUR derivatives, since FOL is employed in the resin manufacture for the foundry industry and as a starting compound for the synthesis of agrochemicals and bioproducts [12], while MF has potential to be used as additive for biofuels and for the synthesis of heterocycles [12]. On the other hand, it has been reported in the literature that the amount and the strength of acid sites also play an important role in determining the catalytic activity and selectivity. The presence of a high concentration of acid sites favors the FUR polymerization and causes the formation of carbonaceous deposits, leading to the partial blockage of active sites for FUR hydrogenation [17,37]. Generally, the use of Cu-based catalysts in the gas-phase hydrogenation of furfural results in FOL as the main product, while the formation of MF is quite low [33]. However, several papers have reported that the presence of weak acidity and small metal Cu nanoparticles bring about the hydrogenolysis of FOL to MF [35,37].As non-noble metal catalysts are prone to deactivation, most catalysts are prepared with a very high metal loading. One of the greatest challenges for the scientific community is to design catalysts in which textural properties, the proportion of acid and basic sites, and the nanoparticle size of the active phase can be easily modulated. For example, Jiao et al. [39] dispersed several metals, Cu among them, on different supports by using the strong electrostatic adsorption (SEA) method and compared them with others prepared by the conventional incipient wetness impregnation approach. They showed that the SEA method provided better metal dispersion, easier reducibility, and smaller particle size. Recently, Wong et al. [40] attained much lower metal particle sizes on silica supports (0.9–1.4 nm), in comparison with metal particles obtained by conventional impregnation (4.2–15 nm), after studying different supported metal catalysts, including Cu-based ones. Regarding the textural properties, the synthesis of porous silica-based solids with ordered structure has grown exponentially [41]. SBA-15 has been the most studied mesostructure in the last 20 years for its applications in the fields of adsorption and catalysis [42]. This mesostructured silica displays a hexagonal ordering formed by parallel cylindrical channels with a diameter between 4 and 9 nm [43], although the main advantage is ascribed to the ease of preparation, good thermal and mechanical resistance, and the tunability of the pore size. However, one of the main disadvantages of mesoporous materials is that, in some cases, active metal nanoparticles are greater than the pore diameter of the support, so these nanoparticles remain on their outer surface, leaving a significant fraction of their specific surface area without active phase. On the other hand, the long length of the cylindrical mesochannels facilitates the deactivation, and, in many cases, these pores are too narrow to allow the access of reagents to the active sites. Additionally, these narrow pores can be easily blocked by the strong interaction of the reactants or products with the active centers, or by the formation of coke [44]. To overcome those challenges, the pore diameter of the mesochannels can be enlarged by the incorporation of aromatics [45,46], alkanes [47,48], or amines [49] in the synthesis gel, although this strategy often promotes the loss of ordering in the porous silica [46]. On the other hand, incorporating fluoride species in the synthetic step limits the growth of the mesochannel length, obtaining other structures with poorer ordering denoted as mesocellular foams [45,46,50].Considering these premises, this work aimed to study the effect of the textural properties of several porous silicas on the dispersion of Cu nanoparticles by using strong electrostatic adsorption (SEA) method. Thus, different synthetic strategies were employed to prepare mesoporous silica with different textural properties, and these were compared to commercial fumed silica. Although this study is focused on establishing a correlation between the textural properties and the catalytic behavior of Cu catalysts supported on mesoporous silicas, the effect of the Cu loading on the catalytic activity was also evaluated by supporting highly dispersed Cu nanoparticles on a commercial fumed silica support using the SEA method. The study of these catalysts in the gas-phase hydrogenation of furfural allowed us to establish a correlation between the number of available metal sites and the catalytic activity. Additionally, an analysis of the deactivation processes is performed, and a strategy for the regeneration of the catalysts is put forward.The chemicals used for the synthesis of the porous silica were P123 Pluronic (PEO20PPO70PEO20, average Mn ∼ 5800, Sigma-Aldrich) as a template, tetraethylorthosilicate (TEOS) (Aldrich, 98%) as the silicon source, and hydrochloric acid (HCl) (VWR, 37%). For the modification of the textural properties of porous silica, ammonium fluoride (NH4F) (Aldrich, 99%) was employed. Commercial CAB-O-SIL® EH-5 fumed silica was also used. Before using it as a support, it was washed to remove any potential metal impurities with a 0.1 M HNO3 aqueous solution for 1 h at room temperature; then, the solution was filtered and washed with deionized water until neutral pH and dried at 60 °C for 24 h. In all cases, Cu was incorporated by using [Cu(NH3)4]SO4·H2O (Sigma-Aldrich, 98%) as the metal precursor. Furfural (FUR) (Sigma-Aldrich, 99%) and cyclopentyl methyl ether (CPME) (Sigma-Aldrich, 99.9%) were employed in the hydrogenation reactions. Likewise, the gases utilized were H2 (Air Liquide 99.999%), He (Air Liquide 99.99%), H2/Ar (10% vol. H2, Air Liquide 99.99%), N2 (Air Liquide 99.9999%), and N2O/He (35% vol. N2O, Air Liquide 99.99%).The synthesis of the SBA-15 was carried out by adjusting the temperature of the hydrothermal treatment, following the method previously described by Fulvio et al. [51]. Briefly, Pluronic P123 was dissolved in a 1.7 M HCl aqueous solution under stirring at 40 °C. Once P123 was dissolved, the silicon source (TEOS) was added dropwise to the mother solution to obtain a gel with a molar ratio of 1 P123: 55 SiO2: 350 HCl: 11,100 H2O. This gel was stirred at 40 °C for 24 h and, on the one hand, it was aged at room temperature for 48 h in order to obtain a porous SBA-15 silica with narrower pore diameter (SBA-LT) and, on the other hand, it was transferred to a Teflon-lined autoclave, where it was aged under hydrothermal conditions at 120 °C for 48 h to get a SBA-15 with a wider pore diameter (SBA-HT).In another synthesis, and in order to modify the textural properties of the SBA-15, NH4F was incorporated into the synthesis step. The synthesis of this modified porous silica was carried out following the synthesis described by Santos et al. [52] in such a way that firstly, both the template (P123) and NH4F were dissolved in a 1.7 M HCl aqueous solution under stirring at 40 °C. After the total dissolution of the template, the silicon source, TEOS, was also incorporated. The final gel had a molar ratio of 1 P123: 1.8 NH4F: 350 HCl: 55 SiO2: 11,100 H2O. Finally, the obtained gel was stirred at 40 °C for 24 h, and then it was aged for 48 h at room temperature. Santos et al. [52] reported that the obtained material displayed a structure with shorter channels, being denoted as mesocellular foam (MFC-LT).In all cases, the obtained gel was washed with distilled water, filtered, and dried overnight at 80 °C. Finally, all the solids were calcined at 550 °C for 6 h (heating rate of 1 °C min−1).The physicochemical properties of the obtained porous silicas were compared with those of a fumed silica with submicron particle size provided by Cabot Corporation.Once porous silicas were synthesized and characterized, Cu was incorporated through the strong electrostatic method (SEA). This consists of preparing a dispersion of the support in water and selecting a pH above its point of zero charge (PZC) in order to have a negatively charged surface, thus promoting the interaction with Cu(NH3)4 2+ species, which results in highly dispersed Cu nanoparticles on silica [39]. For this, 2 g of porous silica was added to 1 L of deionized water under stirring. In the next step, an aqueous solution of [Cu(NH3)4]SO4·H2O was added to obtain a final copper loading of 10 wt%. Then, the pH of the solution was adjusted with diluted NH3 and H2SO4 solutions to pH = 9. Later, the solution was stirred for 20 h at room temperature, and, after that, it was filtered and washed with deionized water, thus obtaining the pertinent catalytic precursors. However, in no case coloration was observed in the washing liquid, so all the Cu(NH3)4 2+ species were incorporated to the supports.The catalyst precursor was thermally treated to decompose the salt and reduce the metal phase in the same process. To do so, the catalytic precursors were heated with a heating rate of 5 °C min−1 from room temperature up to 300 °C under a He flow (60 mL min−1). Then, the gas carrier was changed to H2 (60 mL min−1) and the temperature was increased up to 400 °C (5 °C min−1), which was maintained for 3 h in order to reduce copper species. Later, the gas was switched back to He again, and the sample was maintained at 400 °C for 1 h before cooling down to room temperature under He flow. Finally, the catalysts were passivated using 0.5 vol% O2/N2 for 30 min at room temperature.The samples were labeled as XCu-Y, where X indicates the theoretical metal Cu loading, expressed as wt.%, and Y the support employed. Thus, the following catalysts were synthesized: XCu-SBA-LT for the Cu-based catalyst supported on SBA-15 aged at room temperature, XCu-SBA-HT for the Cu-based catalyst supported on SBA-15 aged at 120 °C, XCu-MCF-LT for the Cu-based catalyst supported on mesocellular foam, and XCu-SiO2 for the Cu-based catalysts supported on a commercial fumed silica. First, the effect of adding different Cu loadings (2.5–20 wt%) was evaluated by using the commercial silica as the support (XCu-SiO2). Then, an intermediate loading of 10 wt% Cu was chosen to evaluate the influence of each support on the furfural hydrogenation (SBA-LT, SBA-HT, and MCF-LT).Small-angle X-ray scattering (SAXS) measurements were collected on a D8 DISCOVER-Bruker instrument at 40 mA and 40 kV. Powder patterns were recorded in capillary transmission configuration by using a LYNXeye detector and a Göbel mirror (CuKα1 radiation). The powder patterns were performed between 0.2 and 10°, with a total measuring time of 120 min.X-ray diffraction (XRD) patterns of the Cu-based catalysts were obtained by using the Bragg–Brentano reflection configuration in a PANanalytical X’Pert Pro automated diffractometer equipped with a Ge (111) primary monochromator (strictly monochromatic CuKα1) and an X’Celerator detector with a step size of 0.0178 (2θ) between 2θ = 10 and 70°, with an equivalent counting time of 712 s per step. The crystalline particle size (D) was evaluated by using the Williamson–Hall equation, B cosθ = (Kλ/D) + (2ε sinθ), where B is the full width at half-maximum (FWHM) of the XRD peak, θ is the Bragg angle, K is the Scherrer constant, λ is the X-ray wavelength, and ε is the lattice strain [53].A FEI Talos F200X equipment was utilized in order to study the catalyst morphology by transmission electron microscopy (TEM). This equipment combines high-resolution STEM and TEM imaging with energy-dispersive X-ray spectroscopy (EDS) signal detection, and 3D chemical characterization with compositional mapping. Samples were suspended in isopropyl alcohol and dropped onto a perforated carbon film grid.Textural properties were obtained from the N2 adsorption–desorption isotherms at −196 °C, as determined by an automatic Micromeritics ASAP 2420 system. Before the adsorption analysis, samples were outgassed overnight at 200 °C and 10−4 mbar. Surface areas were determined using the BET equation considering a N2 cross-section of 16.2 Å2 [54], while the microporosity was determined by the t-plot method [55]. The total pore volume was evaluated from the adsorption isotherm at P/P0 = 0.95, and the average pore size was determined by applying the Barrett–Joyner–Halenda (BJH) method to the desorption branch [56]. On the other hand, a density functional theory (DFT) method was employed to determine the pore size distribution [57].The number of acid sites was determined by temperature-programmed desorption of ammonia (NH3-TPD). For each analysis, 0.08 g of catalyst was placed in a U-shaped quartz reactor and treated with H2 (60 mL min−1) at 300 °C with a heating rate of 5 °C min−1 to remove the passivation layer. Then, the sample was cooled in He (40 mL min−1) until 100 °C, and once the temperature was reached, the sample was saturated with ammonia for 5 min, and then, a He flow was employed to remove the physisorbed NH3. Ammonia desorption was performed by heating the sample from 100 to 300 °C, with a rate of 5 °C min−1, while registering the signal using a thermal conductivity detector (TCD).In order to know the reducibility of the catalysts, hydrogen temperature-programmed reduction (H2-TPR) experiments were conducted. In each test, 0.08 g of the catalyst precursor was employed, being firstly treated under He (60 mL min−1) at 350 °C (heating rate of 5 °C min−1) for 1 h to decompose the precursor salt. After that, the sample was cooled to room temperature and the H2 consumption was monitored between 50 and 800 °C with a heating rate of 10 °C min−1 by using a 10 vol% H2/Ar flow (48 mL min−1). The consumed H2 was quantified by using an on-line TCD in such a way that the water formed in the reduction process was trapped through a cold finger immersed into a liquid N2/isopropanol slurry (-80 °C).The exposed Cu surface and the dispersion of Cu nanoparticles were determined by N2O titration. This methodology is based on the superficial oxidation of Cu0 under a N2O flow, described in previous research [36,52], according to the reaction: 2Cu0 + N2O → Cu2O + N2 Before the analysis, the catalytic precursor was treated under a He flow (60 mL min−1) up to 350 °C (5 °C min−1) for 1 h, followed by a reduction step at 300 °C for 1 h, with a heating rate of 5 °C min−1, under a 10 vol% H2/Ar flow (48 mL min−1). Later, the reduced catalyst was cooled to 60 °C under a He flow. Then, the mild oxidation of Cu0 to Cu+ was performed by flowing N2O (5 vol% N2O/He) at 60 °C for 1 h. After that, the sample was cleaned under an Ar flow and cooled to room temperature. Finally, the Cu2O reduction to Cu0 was performed by heating the sample from room temperature to 300 °C, under a 10 vol% H2/Ar flow (48 mL min−1) with a heating rate of 5 °C min−1, by using a TCD to quantify the H2 consumption.The metallic surface area was estimated according to the equation proposed by Pakharukova et al. [58] (Eq. (1)): (1) S Cu N 2 O ( m 2 g C u - 1 ) = M H 2 · S F · N A 10 4 · C M · W Cu where M H 2 is the number of mol of hydrogen consumed per unit mass of catalyst (μmol H2 gcat −1), SF is the stoichiometric factor whose value is 2, NA is the Avogadro number, CM is the number of copper atoms per surface area unit (1.46·1019 atom m−2), and WCu is the Cu content (wt%).Considering the spherical morphology of Cu0 nanoparticles, the average size (nm) ( d Cu N 2 O ) was determined from Eq. (2): (2) d Cu N 2 O ( n m ) = 6 · 10 3 S Cu N 2 O · ρ Cu where ρ is the density of copper (8.92 g cm−3).The dispersion of the Cu0 nanoparticles was determined by Eq. (3) [59]: (3) D i s p e r s i o n % = 6 · v cu a Cu d · 100 where v Cu is the occupied volume per atom (1.183·10−29 m3 atom−1), aCu is the occupied surface per atom (6.85·10−20 m2 atom−1), and d is the average size of the Cu0 nanoparticles.XPS spectra were obtained in a Physical Electronics PHI 5700 spectrometer with non-monochromatic Mg Kα radiation (1253.6 eV, 300 W, 15 kV) and multichannel detector. Spectra of catalysts were recorded in the constant-pass energy mode at 29.35 eV using a diameter analysis area of 720 μm. Charge referencing was measured against adventitious carbon (C 1s) at 284.8 eV as binding energy (BE). The acquisition and data analysis was performed by using the PHI ACCESS ESCA-V6.0F software package. A Shirley-type background was subtracted from the signals. All the recorded spectra were fitted with Gaussian-Lorentzian curves to determine the binding energies of the different element core levels more accurately. As the catalysts were previously reduced, the samples were stored in sealed vials with cyclohexane as inert solvent to avoid their oxidation. Thus, the samples were prepared in a dry box under a N2 flow and analyzed directly without previous treatment, and the solvent was evaporated before the introduction of the samples into the analysis chamber.The gas-phase hydrogenation of FUR was carried out in a ¼” tubular quartz reactor. The pelletized catalyst (325–400 μm) was placed in the middle section between two layers of quartz wool, discarding diffusion problems through the Weiss-Prater criterion as shown in a previous publication [60]. Prior to the catalytic tests, catalysts were depassivated under a H2 flow (60 mL min−1) at 300 °C for 1 h. Then, the reduced catalysts were cooled down to the selected reaction temperature under a H2 flow (10 mL min−1). After reaching this temperature, a flow of 3.87 mL h−1 of a 5 vol% FUR solution in cyclopentyl methyl ether (CPME) was continuously injected using a Gilson 307SC piston pump (model 10SC). The temperature of the reaction was controlled with a thermocouple located at the same height of the catalytic bed. FUR was dissolved in CPME to avoid problems related to the use of pure furfural, such as blockage of the lines due to FUR polymerization. CPME possesses interesting physicochemical characteristics, such as that it is an environmentally friendly solvent, its low solubility in H2O in comparison to other ethereal solvents, low boiling point (106 °C), low formation of peroxides, and relatively high stability under acid or basic conditions [38]. Liquid samples were collected and kept in sealed vials, being subsequently analyzed by using gas chromatography (Shimadzu GC-14B) with a flame ionization detector (FID) and a CP Wax 52 CB capillary column.In a preliminary test, a Cu-SiO2 catalyst was chosen to evaluate the stability of CPME as solvent in the absence of FUR at 190 °C after 5 h of time-on-stream (TOS). FOL and MF were the only products obtained from the hydrogenation of FUR. These products were quantified from the calibration lines obtained with commercial reagents supplied by Aldrich. The furfural conversion [%], selectivity [%] and yield [%] were calculated as follows (Eqs. (4) to (6)): (4) Conversion ( % ) = mol o f F U R c o n v e r t e d mol o f F U R f e d · 100 (5) S e l e c t i v i t y % = mol of product mo l of FUR converted · 100 (6) Y i e l d % = mol of product mol of FUR fed · 100 XRD diffraction patterns of the XCu-SiO2 catalysts were very noisy (Fig. 1 ), even for the catalyst with the largest Cu loading. On 5Cu-SiO2, diffraction peaks located at 2θ = 43.6 and 50.6°, assigned to Cu0 crystallites, can be observed (PDF: 01-077-3038), whereas the signal located at 2θ = 36.5° was attributed to a crystalline Cu2O phase (PDF: 01-077-0199). The analysis of crystallite sizes by the Williamson-Hall method [53] confirmed that both Cu0 and Cu2O must be highly dispersed as a result of the SEA method since the average crystallite size is lower than 5 nm, although the intensity of these diffraction peaks increased with the Cu loading.On the other hand, the long-range order of porous silicas was determined by SAXS (Fig. 2 ). In the case of the SBA-15 supports, it can be observed a single peak attributed to the d 100 reflection, which is shifted towards lower 2θ values when the aging temperature is increased (from 2θ = 1.03° to 0.88° for SBA-LT and SBA-HT, respectively). This shift can be due to an increase in the pore diameter, wall thickness, or both. It is also noteworthy that the intensity of SBA-LT and SBA-HT signals was similar, which could point out that the ordering of the porous structure does not depend on the synthesis temperature. Similarly, the addition of a modifying pore agent in the synthesis step to form the MFC support also caused a shift of the d 100 reflection to a lower value (2θ = 0.36°). This fact implies a greater increase in the pore size, although the incorporation of this additive also causes a relevant decrease in the intensity of the d 100 reflection, suggesting the formation of a structure with a lower ordering, as was previously described by other authors [48,50].Once the ordering of the porous silicas was evaluated by SAXS, these materials were used as supports for the dispersion of Cu nanoparticles, which were incorporated by the SEA method. In order to compare Cu-based catalysts supported on a commercial fumed silica with those supported on mesoporous silicas, a metal loading of 10 wt% was selected. X-ray diffractograms of the reduced samples (Fig. 3 ) show, in all cases, a broad band at 2θ = 23-25°, which is assigned to the amorphous walls of porous silica. In addition, all samples exhibit intense wide and noisy diffraction peaks at 2θ = 43.6 and 50.6°, which are attributed to the formation of metallic copper crystallites. Additionally, all diffractograms display another small diffraction peak located at 2θ = 36.5°, which is ascribed to the d 111 reflection of Cu2O crystallites. The presence of Cu2O could be attributed to the partial oxidation of samples, as a result of the handling between the preparation and the XRD analysis or, most likely, to a fraction of CuO nanoparticles interacting more strongly with the support, in such a way that they are only partially reduced, forming Cu2O crystallites. Interestingly, the modification of the support does not modify the intensity of the diffraction peaks of Cu0 and Cu2O. This fact implies that all the catalysts must have a similar crystallite size, which is very interesting in order to evaluate the role of the morphology of the support on the catalytic performance.The morphology of supports and the dispersion of Cu nanoparticles were determined by TEM (Fig. 4 ). The TEM micrograph of 10Cu-SBA-LT (Fig. 4A) evidences an ordered support with a honeycomb morphology and parallel channels. In the case of 10Cu-SBA-HT (Fig. 4B), the support maintains an ordered structure, although there is a greater separation between adjacent silica walls, in agreement with XRD data (Fig. 2). This fact would suggest the formation of a porous structure with a larger pore diameter than SBA-LT. As shown in Fig. 4C1, incorporating a structure modifying agent (MCF-LT) causes a disorder of the porous structure, as the mesochannels detected in other mesoporous silicas cannot be observed here. Regarding copper supported on commercial silica (10Cu-SiO2) (Fig. 4D), small particles with spherical morphology can be seen. From these micrographs, it can be concluded that all the supports exhibit different morphology, which could influence their catalytic behavior. Furthermore, these micrographs also show a high dispersion of Cu nanoparticles on the different supports, with some nanoparticles small enough to enter the porous structure (Fig. 4C2), except for the 10Cu-SBA-HT catalyst where the agglomeration of small Cu crystallites can be observed on the surface of SBA-15 (Fig. 4B). This fact could be due to the hydrothermal synthesis leading to a more hydroxylated external surface than inside the pores, or the existence of regions with different points of zero charge (PZC) within the material, which could promote the strong electrostatic adsorption of Cu particles on the external surface of 10Cu-SBA-HT.In order to determine the textural properties of catalysts, N2 adsorption–desorption isotherms at −196 °C were obtained (Fig. S1A). According to the IUPAC classification, 10Cu-SBA-LT, 10Cu-SBA-HT, and 10Cu-MCF-LT catalysts exhibit Type IV isotherms [61], which are ascribed to mesoporous materials; however, the N2 adsorption–desorption profiles are different due to their distinct textural properties. The isotherm of 10Cu-SiO2 shows an increase in the N2 adsorbed at higher relative pressure, and, unlike the other ones, it can be classified as Type II, typical of macroporous materials [61]. In this latter case, the porosity can be ascribed to interparticle voids, and the presence of the hysteresis loop suggests the presence of cavities whose size is greater than 4 nm [61]. The hysteresis loops of 10Cu-SBA-LT, 10Cu-SBA-HT, and 10Cu-MCF-LT catalysts can be considered as Type H1, found in porous materials with a narrow range of uniform mesopores [61]. However, the hysteresis loop of the 10Cu-SiO2 catalyst resembles the Type H3, which is given by aggregates of spherical silica particles, in agreement with the results obtained by TEM (Fig. 4D).Concerning the BET surface area (SBET, Table 1 ) [54], 10Cu-SBA-LT, 10Cu-SBA-HT, and 10Cu-MCF-LT catalysts show very similar values (356–409 m2 g−1). In contrast, 10Cu-SiO2 possesses a much lower SBET (179 m2g−1). However, the pore volume analyses evidence clearer differences than the SBET values. Interestingly, while 10Cu-SBA-LT displays the lowest pore volume (VP), the use of hydrothermal conditions (10Cu-SBA-HT) leads to the largest one. On the other hand, 10Cu-MCF-LT possesses the highest microporous surface area (t-plot, 34 m2g−1) and volume (VMP, 0.0139 cm3g−1). The pore size distribution was determined by a DFT method (Fig. S1B) [57]. In all cases, the microporosity (<2 nm) was relatively low, being very similar in all of them, which is in agreement with the t-plot analysis and the micropore volume (VMP). In this sense, it has been reported in the literature that the microporosity of SBA-15 is directly related to the aging temperature. Thus, several authors have affirmed that lower aging temperatures favor the interaction of the P-123 molecules with adjacent micelles, leading to structures interconnected by microchannels [46,62]. The presence of these microchannels has enormous importance in the adsorption and catalysis fields, as they might help to minimize the possible diffusion problems associated with the long mesochannels of SBA-15. However, by increasing the aging temperature, isolated P-123 micelles without connection between them are formed, which results in decreased microporosity after calcination [62]. Despite this, it has been hardly seen any differences between the microporosity of 10Cu-SBA-LT and 10Cu-SBA-HT. In this sense, it has been reported that these micropores are easily blocked when porous silicas are functionalized by grafting or subjected to impregnation [46]. However, regarding the mesoporosity, there are clear differences between supports (Fig. S1B). Even when 10Cu-SBA-LT, 10Cu-SBA-HT, and 10Cu-MCF-LT catalysts present a narrow distribution of pore sizes, these are centered at different pore diameters. It has been previously reported in the literature [62] that an increase in the aging temperature causes a rise of the average pore diameter (dP); therefore, as the pore distribution for 10Cu-SBA-LT is centered at 4.3 nm and for 10Cu-SBA-HT at 8.6 nm, our results are consistent with previous studies. In the case of 10Cu-MCF-LT, the pore diameter increases compared to 10Cu-SBA-LT, which was synthesized at the same aging temperature, as a consequence of the addition of fluoride in the synthetic step that modified the growth and ordering of the porous silica, giving rise to a mesocellular structure. Finally, the 10Cu-SiO2 catalyst does not show a homogeneous pore size distribution because its specific surface can be mainly attributed to interparticle voids between silica microspheres.The quantification of the available Cu0 sites was performed by N2O titration at 60 °C (Table 2 ), following the methodology described in previous studies [37,38,58]. The incorporation of higher loadings of Cu species over commercial silica led to a higher metallic surface area, reaching the maximum value (25.0 m2 Cu gcat −1) on 15Cu-SiO2 (Table 2). However, a higher Cu loading (20Cu-SiO2) reduces the available metallic surface area, likely due to the formation of larger Cu crystallites. Regarding the dispersion of the Cu0 sites, a decrease is observed as the Cu content increases.Regarding the influence of the support (10Cu-Y catalysts) (Table 3 ), 10Cu-MCF-LT displays the highest metallic surface area, 25.9 m2 Cu gcat −1, revealing the highest Cu dispersion and the lowest particle size (2.6 nm). On the contrary, the 10Cu-SiO2 catalyst shows the lowest metallic surface area and the largest Cu particle size for the same theoretical Cu loading. However, in all cases, the average Cu particle size, estimated by N2O titration, is lower than 5 nm, which is in agreement with previous studies using the SEA method [39,63,64], thus corroborating that this methodology provides highly dispersed metal particles in comparison with other conventional methods, such as incipient wetness impregnation or precipitation [64].The reducibility of the Cu(NH3)4 2+ species was elucidated from their H2-TPR profiles (Fig. 5 ). Considering that all catalysts possess the same Cu loading (10 wt%) and a silica support, the reducibility of the Cu nanoparticles should only be attributed to their size because the interaction should be similar. In this sense, it has been reported in the literature that sometimes it is not feasible to distinguish the different reduction steps for copper species (Cu2+ → Cu+ → Cu0) [37]. As shown in Fig. 5, the maximum of the H2-TPR profiles shifts from 205 to 230 °C due to the slight increase in the particle size, as was inferred from the N2O titration (Table 3). Thus, copper nanoparticles are more easily reduced in the case of 10Cu-MCF-LT than in 10Cu-SiO2. With respect to the samples supported on SBA-15, the maximum slightly shifted toward a higher temperature for 10Cu-SBA-HT compared to 10Cu-SBA-LT, which could be attributed to those agglomerated particles detected by TEM (Fig. 4) that would be more difficult to reduce. Additionally, in most H2-TPR profiles, and mainly in the case of 10Cu-SiO2, a shoulder is observed at higher reduction temperatures, which could imply the existence of a small fraction of bigger Cu2+ nanoparticles that require even higher temperatures to be reduced.In order to analyze the chemical composition on the surface, including the oxidation state of the active phase, XPS analyses were also carried out (Fig. 6 and Table 4 ). The Cu 2p core level spectra analysis for the Cu-based catalysts with a theoretical Cu loading of 10 wt% (Fig. 6A) shows a band located between 932.6 and 932.9 eV, which is ascribed to reduced Cu species [65]. However, as it is not possible to differentiate between Cu+ and Cu0 species, it is also necessary to use the CuLMM Auger line (Fig. 6B). The coexistence of Cu0 and Cu+ is observed in all cases, being Cu+ the most abundant (56.6–60.2%) (Table 4). The presence of Cu0 and Cu2O agrees with XRD data (Fig. 3), although those showed more intense peaks for Cu0 species. The higher proportion of Cu+ species could be ascribed to partial oxidation of Cu0 species during handling, or to the presence of small Cu2O particles well dispersed on the surface, which would be hardly observed by XRD. On the other hand, in the Si 2p core level spectra, there is a contribution located at about 103.3–103.5 eV, which is typical of SiO2 [65], while the O 1 s core level spectra also show the characteristic band of oxygen in SiO2 at about 532.7–532.8 eV (Table 4) [65].The surface Cu composition obtained by XPS analysis of the Cu-based catalysts with a 10 wt% theoretical Cu content is between 1.0 and 2.2% (Table 4). On those catalysts with greater dispersion and consequently smaller particle size, one should expect higher surface copper contents. However, this is not the case when comparing the dispersion obtained by N2O titration (Table 3) and the XPS results (Table 4). In this sense, TEM micrographs have revealed that, in the case of 10Cu-SBA-LT (Fig. 4A) and especially on 10Cu-SBA-HT (Fig. 4B), there is a large proportion of Cu nanoparticles deposited on the external surface of the porous structures, whereas smaller particles are able to penetrate into the pores of 10Cu-MCF-LT (Fig. 4C2). Taking this premise into account and considering that XPS is a surface technique, it is possible to justify the higher surface Cu content on 10Cu-SBA-LT and 10Cu-SBA-HT when compared to 10Cu-MCF-LT.The quantification of the acid sites, performed by NH3-TPD (Table 1), reveals a relatively low concentration between 32 and 57 μmol g−1 for all the 10Cu-Y catalysts. This low acidity is mainly ascribed to the Brönsted acid sites associated to silanol groups that have not been involved in the SEA process, whereas the existence of Lewis acid sites can be assigned to Cu+ species, which are more easily formed when small Cu0 particles are present. The acidity is necessary to obtain MF, although the amount and strength of those acid sites need to be modulated to minimize the deactivation in the gas-phase hydrogenation of FUR. It should also be noted that 10Cu-MCF-LT showed the lowest acidity. Regarding the density of acid sites, 10Cu-SiO2 exhibited the highest value, while 10Cu-MCF-LT showed the lowest one, despite having a surface area similar to that of 10Cu-SBA-LT and 10Cu-SBA-HT.These Cu-based catalysts with different textural properties were studied in the gas-phase FUR hydrogenation. In this work, the catalytic conditions were selected to observe the differences between the catalysts prepared with different silica supports, instead of choosing the best experimental conditions to achieve the complete conversion of furfural. In addition, the physico-chemical characterization of fresh and used catalysts also allows to elucidate the causes of the observed deactivation of the catalysts.The first study focused on the influence of the Cu loading on the catalytic behavior. For this purpose, the commercial SiO2 with Cu contents between 2.5 and 20 wt% was selected (Fig. 7 ). In most cases, it was observed a progressive decrease in the FUR conversion with time-on-stream (TOS) (Fig. 7A), being more pronounced with those catalysts with Cu loadings lower than 15 wt%. Thus, FUR conversion increases from 47% for 2.5Cu-SiO2 to almost full conversion for 15Cu-SiO2 after 5 h of TOS at 190 °C. It is striking that the catalytic activity slightly decreased for the catalyst with the highest Cu content (20Cu-SiO2) compared to 15Cu-SiO2 under similar experimental conditions. In this sense, previous research studies have shown similar trends with other Cu/SiO2 catalysts with small particle sizes [37]. In that study, it was observed that the increase in the Cu content caused the sintering of metal particles, thus decreasing the metallic surface area (i.e., amount of available active sites).Regarding the product distribution (Fig. 7B-C), the catalysts with lower Cu contents tend to form a larger amount of FOL, even at short reaction times, with the 5Cu-SiO2 catalyst reaching the highest FOL yield after 1 h of TOS at 190 °C (54%), with this slightly decreasing to 49% after 5 h of TOS. An increase in the Cu loading causes a progressive decrease in the FOL yield, which is accompanied by a rise in the MF yield. This trend is clearly observed in the case of 10Cu-SiO2, although the MF yield decreases from 70% after 1 h to 12% after 5 h of TOS at 190 °C. On the contrary, the FOL yield increases from 19% after 1 h to 51% after 5 h of TOS at 190 °C.Considering that all catalysts display similar acidity and metal particle size, these should show similar catalytic behavior. Thus, it is expected that all catalysts are highly selective towards MF at t0; however, the hydrogenolysis sites (FOL → MF) in the catalysts with lower Cu content (lower metallic surface areas) are poisoned very fast, in such a way that FOL is the main product after the first hour of reaction. In the case of the catalysts with higher Cu contents, their higher metallic surface areas maintain the hydrogenolysis process along the TOS, reaching a MF yield of 82% with the 15Cu-SiO2 catalyst after 5 h of TOS at 190 °C. In any case, the hydrogenolysis sites involved in the FOL → MF reaction are more susceptible to deactivate than the hydrogenation sites involved in FUR → FOL, as the increase in the FOL yield with TOS suggests. Interestingly, the 20Cu-SiO2 catalyst provides a lower proportion of MF and larger FOL after 5 h of TOS compared with the product distribution at the beginning of the reaction. These data are in agreement with a previous study in which it was demonstrated that the incorporation of higher Cu loadings led to the formation of bigger Cu nanoparticles, which favor the hydrogenation reaction (FUR → FOL) with respect to the hydrogenolysis reaction (FOL → MF) [37]. In this sense, Shan et al. pointed out that the formation of Cu–O-Si-O- bonds favored the formation of Lewis acid sites, even under H2 flow, which can be involved in the hydrogenolysis step [66]. In this regard, the presence of Cu+ species, identified by XRD (Figs. 1 and 3), may also imply the existence of Lewis acid sites, which may be involved in the FOL → MF hydrogenolysis process [66,67].The catalytic activity reported in Fig. 7A follows a similar trend to the metallic surface area (m2 gcat −1) shown in Table 2, in such a way that a higher content of available Cu sites provides a greater catalytic activity and resistance against deactivation. Consequently, the catalyst with the highest metallic surface area (15Cu-SiO2) is the most active. However, the catalyst with the highest Cu content (20Cu-SiO2) displays a selectivity pattern that resembles that of 7.5Cu-SiO2 and 10Cu-SiO2 catalysts, even though the catalytic activity almost stays unaltered, which is likely due to the decreased metallic surface area and the stronger deactivation of the hydrogenolysis sites (FOL → MF).These results have been compared with those obtained with copper chromite, the commercial catalyst used in the industrial process, under similar experimental conditions [68]. It should be noted that copper chromite reached a FUR conversion of about 75% after 30 min of reaction, but it was very prone to deactivation, with negligible activity after 5 h of reaction. Regarding the selectivity, the main product was FOL with a selectivity above 90%. Those data differ from the reported in the present work because deactivation is more limited and the main product, especially at short reaction times, is MF, which indicates that the hydrogenolysis reaction FOL → MF is promoted due to the presence of a small fraction of weak acid sites.Taking into account that 10Cu-SiO2 is the most prone catalyst to modify its selectivity pattern with TOS (Fig. 7A) and that this catalyst does not maintain a total conversion, a 10 wt% loading was selected to study the influence of the textural properties and morphology of the porous silica on the catalytic behavior (Fig. 8 ). Considering that all catalysts display similar Cu crystallite size (Table 3) and active site-support interaction (Cu-SiO2), the difference in the catalytic behavior must be ascribed to the textural properties of the catalysts.The data reveal that all the catalysts, except 10Cu-SBA-HT, reach high conversion values at short TOS, with initial values close to 100% (Fig. 8A). While 10Cu-SBA-LT and 10Cu-MCF-LT render FUR conversions higher than 90% after 5 h of TOS at 190 °C, 10Cu-SiO2 deactivates (FUR conversion of 66% after 5 h of TOS). However, in the case of 10Cu-SBA-HT, the conversion is low from the beginning of the reaction, and the conversion decreases further after 5 h of TOS, obtaining the poorest FUR conversion (11%).The analysis of the product distribution (Fig. 8B-C) reveals that the most active catalysts (10Cu-MCF-LT, 10Cu-SBA-LT, and even 10Cu-SiO2) show higher MF yields (greater than 70%) at shorter reactions times (1 h), whereas 10Cu-SBA-HT is more selective towards FOL. As the reaction time progresses, the selectivity towards MF decreases, although the most active catalyst (10Cu-MCF-LT) still maintains a MF yield of 76% after 5 h of TOS at 190 °C. The deactivation process is accompanied by a concomitant increase in the formation of FOL, obtaining the highest FOL yield of 49% with the 10Cu-SiO2 catalyst after 5 h of TOS. Interestingly, the formation of FOL is more pronounced on the catalysts that deactivate. Therefore, it can be concluded that MF formation occurs through two consecutive reactions, FUR → FOL on hydrogenation sites, and FOL → MF on hydrogenolysis sites, which are more likely to be deactivated, as had previously been suggested by other authors [31,35,37]. In this respect, DFT studies have suggested that the first step, the hydrogenation reaction (FUR → FOL), occurs by the interaction of the Cu nanoparticles with the oxygen atom of the carbonyl group in a top η1(O)-aldehyde binding mode [26]. In the second step, the hydrogenolysis reaction (FOL → MF), FOL adopts a similar disposition on the copper sites [31] through an interaction with the oxygen atom of the hydroxyl group, while the furan ring is nearly parallel to the surface. Several authors have pointed out that the existence of weak acidity, together with the presence of small Cu particles, favors the hydrogenolysis reaction [35,37]. However, the acidity cannot be high, as this would favor the polymerization and the strong adsorption of FUR on the catalyst surface, blocking the active sites in such a way that the catalysts suffer a drastic deactivation at short reaction times. NH3-TPD data (Table 1) confirmed the low acidity of these catalysts; therefore, using the SEA method for the incorporation of small Cu nanoparticles onto amorphous and porous silica seems to be appropriate to minimize, or slow down, the generation of carbonaceous deposits. Considering that most of the catalysts here synthesized initially produce MF, it could be proposed that the acidity, even if low, would be associated with the silanol groups and partially reduced Cu species (Cu+), which provide Lewis-type acidity. In fact, other authors concluded that small Cu+ particles exhibited Lewis acidity, which promoted hydrogenolysis processes [66,67]. On the other hand, it should be noted that the hydrogenolysis reaction (FOL → MF) generates H2O as a by-product, which might have an adverse effect on the catalytic performance as it could favor the oxidation of Cu0 to Cu+, thus increasing the Lewis acidity. In this sense, it has been reported in the literature that Cu+ sites can interact more strongly with CO than the Cu0 ones [69]. This fact can be extrapolated to the carbonyl group of the FUR molecule, being able to infer that the presence of a higher proportion of Cu+ sites could favor a stronger interaction of the FUR molecules with the active centers, favoring a faster formation of carbonaceous deposits that block the active sites involved in the FUR hydrogenation. Moreover, previous studies have reported that the adsorption heats of FUR and H2O are very similar (12.3 and 12.4 kcal mol−1, respectively) [31,70]; therefore, they could compete for the same active centers, negatively affecting the catalytic behavior. In addition, the interaction of H2O with FUR can also favor its polymerization. Thus, the presence of H2O could promote the partial oxidation of copper species; however, the Cu+/Cu0 ratio detected by XPS was very similar in all cases, discarding its influence on the selectivity pattern. Taking into account all these premises, it seems clear that the sites involved in the second step (FOL → MF) are more prone to deactivate than those involved in the first step (FUR → FOL), as shown in Fig. 8B-C, since the decay of the formation of MF is accompanied by larger amounts of FOL. This fact could be explained by the formation of carbonaceous deposits that would mainly block those active sites involved in the hydrogenolysis reaction.Considering that all the 10Cu-Y catalysts possess the same Cu loading (10 wt%) and all the supports have the same chemical composition (SiO2), a similar catalytic behavior should be expected, although this is not the case (Fig. 8A). As previously stated, 10Cu-Y catalysts displayed different textural properties (Supplementary Information, Fig. S1 and Table 1), and those with the narrowest pore diameter between 3 and 7 nm (10Cu-MCF-LT and 10Cu-SBA-LT) provided the highest FUR conversions, likely due to the closer proximity of furfural to the active sites within the pores. Moreover, both catalysts exhibited the highest metallic surface area and, consequently, the lowest particle size (Table 3). Thus, a suitable pore size together with the small particle size could favor the reaction of FUR on the Cu sites. Conversely, the lower activity of the 10Cu-SBA-HT catalyst could be attributed to its pore diameter being too large (7–12 nm), which makes difficult the adsorption of FUR on the active sites, in such a way that FUR molecules can go across the SBA-15 mesochannels without interacting with the active sites. That catalyst also presents some agglomeration of the Cu nanoparticles on the outer surface, as observed by TEM (Fig. 4B). This was also observed by N2O titration (Table 3), with 10Cu-SBA-HT and Cu-SiO2 presenting the lowest dispersion values. On the other hand, when comparing the catalysts with narrower pore size distributions (10Cu-MCF-LT and 10Cu-SBA-LT), the 10Cu-SBA-LT catalyst is more prone to modify its selectivity pattern than 10Cu-MCF-LT. In this sense, even when the presence of narrow pores could favor the interaction between FUR and Cu sites, the longer channels could promote the subsequent interaction of FUR, or the reaction products, with the active centers, in such a way that these pores would be blocked. However, with the addition of fluoride in the synthesis step, the length of the silica channels decreases, which facilitates the furfural access and exit of the reaction products from the channels, thus showing a more gradual deactivation. On the other hand, the N2O titration data also show that the 10Cu-MCF-LT catalyst displays the highest metallic surface area, so this catalyst possesses the highest number of available Cu sites for the FUR hydrogenation. It should also be noted that the density of acid sites on 10Cu-SBA-LT is twice that observed for 10Cu-MCF-LT, which could also contribute to a faster deactivation through the formation of carbonaceous deposits. Likewise, it was previously mentioned that the 10Cu-SiO2 catalyst was more prone to deactivation, which agrees with its highest number of acid sites and the lowest metallic surface area. Therefore, it can be concluded that the 10Cu-MCF-LT catalyst possesses suitable textural and acidic properties to provide a high FUR conversion with a low deactivation rate.The following study aims to evaluate the effect of the reaction temperature on the catalytic performance (Fig. 9 ). Considering that the boiling point of FUR is 161.7 °C, the catalytic studies were carried out between 170 and 230 °C. A volcano-type conversion as a function of temperature has been previously reported in the gas-phase hydrogenation of FUR [17,36,71]. Thus, the FUR conversion increases at lower reaction temperatures because the FUR hydrogenation is thermodynamically favored under those conditions [17]. However, higher temperatures worsen the catalytic performance due to the polymerization processes that take place, which cause a fast deactivation [17]. Furthermore, the hydrogenolysis reaction is favored at high temperatures [17,36]. As expected, the reaction at 170 °C provides the lowest conversion values (70% after 5 h of TOS), although the catalyst is highly selective to FOL (59% yield). Increasing the reaction temperature improves the FUR conversion, but it also modifies the selectivity pattern. Thus, at 190 °C, the FUR conversion is higher than 93% and the MF yield is above 75% in all cases, after 5 h of TOS, confirming that the hydrogenolysis reaction is favored at higher temperatures, as previously reported [36,71]. Surprisingly, a further increase of the reaction temperature did not decrease the conversion, as previously pointed out in the literature [17,36]. This could be due to the small size of the Cu crystallites (<4 nm), which implies that the metallic surface area is very high, so that the catalysts are more resistant to deactivation than others with bigger metal nanoparticles [36–38]. It has also been demonstrated that the 10Cu-MCF-LT catalyst possessed a lower density of acid sites and suitable textural properties due to the use of fluoride for the synthesis of this support, which could contribute to slowing down its deactivation.Finally, in order to further evaluate the stability of the 10Cu-MCF-LT catalyst, another experiment was carried out at a longer reaction time (Fig. 10 ). This stability test was performed at 230 °C because high conversion and MF selectivity were observed under these experimental conditions, in order to determine if that high MF production could be maintained along with TOS. Even though 10Cu-MCF-LT showed almost complete conversion initially, it suffered gradual deactivation with TOS, obtaining a FUR conversion of 44% after 48 h of TOS. Regarding product distribution, the main product during the first hours of TOS is MF, resulting from the consecutive hydrogenation-hydrogenolysis reactions taking place. However, after 6 h of TOS, the MF selectivity decreases likely due to some metal sites involved in the hydrogenolysis reaction being blocked by carbonaceous deposits, as was previously reported [37], so that the FUR → FOL → MF process mostly ends in FOL, maintaining a FOL yield of 39% after 48 h. This fact confirms that the hydrogenation sites are less susceptible to deactivation than the hydrogenolysis ones. As the gas-phase FUR hydrogenation is a reaction that undergoes strong deactivation processes, it is essential to evaluate the regeneration capacity of catalysts. According to previous studies, the carbonaceous deposits involved in the deactivation could be removed by calcination at 500 °C [36,37]. After calcination at this temperature, 10Cu-MCF-LT displayed a lower FUR conversion value (80%) than the first cycle (Fig. 10). However, the deactivation follows the same trend to that observed along the first run, obtaining a conversion of 25% after 48 h of TOS. Nevertheless, the selectivity profile is different to that shown after the first run, with FOL as the main product (62% FOL yield after 1 h) instead of MF. The change in the selectivity pattern is ascribed to an increase in the Cu particle size during the regeneration process, which is reflected on the loss of the hydrogenolysis sites due to the reduction of the metallic surface area. In this sense, it has been reported in the literature that Cu-based catalyst favor the formation of MF in gas-phase when Cu nanoparticles sizes are lower than 5 nm [35,37,64], due to the generation of Lewis acid sites [72] besides those associated to the slightly acidic silica support [35,37].In order to elucidate the changes of the active phase throughout the catalytic tests, the 10Cu-MFC-RT catalyst was collected after two cycles of 48 h of TOS at 230 °C. XRD analysis shows that the 10Cu-MFC-RT catalyst hardly suffered any changes (Fig. S2), discarding the transformation of Cu2O, or Cu0, in other crystalline phases. However, the XPS analysis revealed that the proportion of Cu+ remained almost unchanged after two cycles of 48 h (Fig. S3). While the presence of Cu+ species could favor the hydrogenolysis reaction to produce MF, the modification of the selectivity pattern would indicate that the presence of Cu+ does not have a predominant role in determining the selectivity pattern, as inferred from a higher production of FOL.From the XPS data (Fig. S3 and Table 4), it is observed a decrease in the intensity of the Cu 2p core level spectrum and the Cu AugerLMM line after the two catalytic cycles, which suggests a decrease in the dispersion of Cu species on the catalyst surface. The analysis of the surface chemical composition, before and after the catalytic cycles, also shows an increase in the carbon content after the reaction (Table 4), confirming that a fraction of the Cu species could be partially blocked by the formation of carbonaceous deposits (polymerized FUR and FOL), which seems to agree with the progressive decline of the FUR conversion.A series of Cu-based catalysts was prepared by the strong electrostatic adsorption (SEA) method using several silica supports with different morphologies and textural properties (commercial fumed silica, SBA-15 synthesized at room temperature and under hydrothermal conditions, and mesocellular foam), and these were studied in the gas-phase hydrogenation of furfural, obtaining Cu0 crystallites with a size below 5 nm. Regarding Cu supported on mesoporous silica catalysts, it can be concluded that the presence of large pores (greater than 7 nm), like those found on the SBA-15 synthesized under hydrothermal conditions, complicated the intimate interaction between the reagents and the active sites along the channels, giving rise to very low catalytic activity. In addition, this catalyst also showed that the agglomeration of the Cu nanoparticles could worsen the activity even more. In contrast, mesoporous silica supports synthesized at room temperature, 10Cu-SBA-LT and 10Cu-MCF-LT catalysts, offered a more suitable pore size for reaction. Moreover, it was demonstrated that the addition of fluoride on the synthesis shortened the length of the silica channels (mesocellular foam), which facilitates the furfural access and exit of the reaction products on 10Cu-MCF-LT, thus providing the highest values of furfural conversion and 2-methylfuran yield. Besides, the high production of 2-methylfuran at the beginning of the reaction could be attributed to the low acidity of this catalyst. However, its selectivity pattern changed along 48 h of TOS, even after intermediate catalyst regeneration, due to the fact that hydrogenolysis sites are more prone to deactivation processes.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 for financial support from the Spanish Ministry of Science, Innovation and Universities (RTI2018‐94918‐B‐C44 project), FEDER (European Union) funds (UMA18-FEDERJA-171 and UMA20-FEDERJA-88), Junta de Andalucía (P20-00375), and the University of Malaga. C.P.J.G. and C.G.S. acknowledge Junta de Andalucía and FEDER funds, respectively, for their postdoctoral contracts. A.C.A.R. is thankful to the University of Toledo for her start-up package. Funding for open access charge was provided by the University of Malaga/CBUA.Supplementary data to this article can be found online at https://doi.org/10.1016/j.fuel.2022.123827.The following are the Supplementary data to this article: Supplementary data 1

Cu nanoparticles were incorporated on different porous silica supports (commercial silica, SBA-15 synthesized both at room temperature and under hydrothermal conditions, and mesocellular foam) by the strong electrostatic adsorption method and tested in the gas-phase furfural hydrogenation for the production of furfuryl alcohol and 2-methylfuran, being the latter a promising biofuel. The incorporated copper species provided metal particle sizes lower than 5 nm in all cases. However, different catalytic behaviors, both in terms of conversion and selectivity, were detected due to the morphology and textural properties of each support. Cu over commercial silica was more prone to suffer from deactivation, and it provided higher furfuryl alcohol yields, probably due to its higher acidity and lower metallic surface area. On the other hand, the agglomeration of Cu nanoparticles together with larger pore sizes complicated the access of furfural to the active sites using the hydrothermal SBA-15 support, thus decreasing the activity. In contrast, the addition of fluoride in the synthesis of mesoporous silica, which shortened the length of the silica channels (mesocellular foam), facilitated the furfural access and provided both higher metallic surface area and lower acidity. This fact led to a more gradual deactivation, still attaining high values of furfural conversion and 2-methylfuran yield (95 and 76%, respectively) after 5 h at 190 °C. However, Cu over mesocellular foam changed its selectivity pattern along 48 h and after regeneration, increasing the furfuryl alcohol selectivity due to the decreased number of available active sites.

Data will be made available on request.Photocatalysts, projected as one of the most promising catalytic media of producing H2, are key for the development of sustainable energy sources to achieve the goal of averting imminent climate change while sustaining economic growth. The typically used industrial photocatalysts, such as TiO2, or carbon nitrides are active at ultraviolet wavelengths constituting < 5 % of the solar spectrum. Unfortunately, practical strategies often fall short for binary/ternary photocatalysts due to under-explored various interfacial interactions. Two-dimensional (2D) transition metal dichalcogenides, such as MoS2, have recently attracted attention for environmental-friendly applications [1]. However, bulk MoS2 is known to be less efficient for hydrogen evolution reaction (HER) due to photocatalytically inactive basal plane of MoS2, as only the extremely active sulfur terminated edges display robust nature, and slightly more positive conduction band position relative to the redox potential of H+/H2 [2]. Nevertheless, the physical properties of MoS2 can substantially be modulated by molecular/nanoparticles (NPs) adsorption [3], applied electric field [4], and metallic contacts [5]. In the case of metal-MoS2 contact, the photoactivity not only depends on layered materials’ structure and properties, but also on the interface between the photocatalyst and cocatalyst [2,6]. The well-defined interface between metal and MoS2 can hamper electron-hole recombination and optimize the light absorption of photocatalysts. Moreover, the physical properties MoS2 can effectively be tailored through these interfaces via carrier injection and strain engineering [7,8]. Herein, taking the HER as an example, a combination of synchrotron and lab source-based spectro-microscopic investigations of various interfacial interactions in Ni-Ag-MoS2 (MAN) ternary heterostructure is carried out.Despite the importance, study of the interfacial contact between the layered semiconductor and hetero-junctional sites is not an easy task. As per the previous report, metallic contact between MoS2 and Ni NP happens via Au [9] / Ag [10] nano-buffers as inferred from X-ray absorption (XA)-photoelectron emission microscopy (PEEM), referred to as X-PEEM, a spectro-microscopic full-field, synchrotron-based surface sensitive technique. The high sensitivity of the XA spectrum (XAS) to oxidization states, local structure, and electronic structure makes X-PEEM highly attractive technique. However, X-PEEM fails to probe the effect of interaction between Ag and MoS2 on MoS2 due to dilute concentration of Ag and technical difficulties to record XAS of Ag at the soft X-ray regions.Raman spectroscopy, on the other hand, is an excellent local and non-destructive optical tool to probe structure and local environment in 2D materials. Furthermore, Raman mapping can monitor changes in Raman spectra at spatially different positions [11,12]. In case of metal-MoS2 contact, Raman spectroscopy has been used to probe the effect of metal on the vibrational properties of MoS2. Different observations, such as i) red shift in both the phonon modes due to mechanical strain induced because of metallic contact [13], ii) splitting of both E2g 1 and A1g modes due to biaxial strain [14], iii) splitting of E2g 1 mode only due to uniaxial strain [15], and iv) shift in A1g due to charge doping, whereas shift in E2g 1 due to stress have been made [7,16]. In hybrid structures, dark localized surface plasmon resonances (LSPR) modes can relax non-radiatively by transferring energy to electrons via Landau damping leading to charge carrier (electron) injection in the lattice of the adjacent 2D semiconductor [17,18]. Till date, mainly Raman spectroscopy has been used to probe these interactions. However, this doping and strain distribution can be highly inhomogeneous, depending on the contact between the two, which mandates simultaneous microscopic and spectroscopic determinations.In this work, we have investigated various interfacial interactions in Ni-Ag-MoS2 (MAN) via X-PEEM and Raman mapping. Due to the interaction at Ag-MoS2 interfaces, Raman spectrum of the heterosystem is dominated by compensated charge doping and compressive strain depending upon the laser power used along with intimate contact between MoS2 and Ag. We believe that this study pushes forward the frontier of binary/ternary photocatalyst design towards efficient water splitting.MAN heterostructure was synthesized by utilizing a sonication based wet chemical synthesis following the method reported earlier [9]. Materials for composites include Ni nanopowder (Ningxia Orient Tantalum Industry, Co. Ltd.), MoS2 (99.5 % assay, were purchased from Nanjing Emperor Nano Material Co. Ltd.) and AgNO3 (0.01 mol/L, Sigma-Aldrich). MoS2 powder (43.4 mg) and Ni powder (4 mg) were mixed in 100 mL de-ionized (DI) water and shaken vigorously to get evenly dispersed suspension followed by addition of AgNO3 aqueous solution (1.3 mL, 0.01 mol/L) to the flask. The material synthesis was carried out by sonication using a Skymen JTS-1018 water bath ultrasonic cleaner. The cleaner was pre-heated to 70 °C, and the materials were then sonicated @35 kHz using maximum power (∼ 3 A current) for a duration of 4 h. Post sonication, the samples were kept at room temperature overnight for the sedimentation of particles. Post sedimentation, water was carefully rinsed without losing synthesized particles, followed by decanting the synthesized particles. Drying the remaining water off from the as-prepared samples was carried out by evaporation of the water in an open beaker by utilizing a hot plate to heat samples to approximately 100 °C in the ambient air. Dried samples were then scraped off the beaker and stored in sample bottles as a dry powder. The synthesized product was then dispersed in ethanol (Sigma-Aldrich, 95.0 %) followed by drop cast on pure Si substrate. The same sample was employed for scanning electron microscopy, X-PEEM, and Raman mapping measurements. In a control experiment, Ag-MoS2 binary system was also prepared to compare the HER of MAN with Ag-MoS2.It is worth noting that to reach fundamental aspects down to illustrations at electronic structural levels, the composite was synthesized and treated with the above fractions and conditions so that the compositions are well distinguished in the following spectro-microscopic studies.Zeiss Ultra plus Field emission scanning electron microscope (FESEM) was used to study the morphology of heterostructures. X-PEEM measurements on selected MAN (based on FESEM) was carried out at the AC-SPLEEM end-station of MAXPEEM Beamline at MAX IV laboratory (Lund, Sweden), using a modified SX-700 monochromator equipped with 1220.9 lines.mm−1 (high-density) Au/Si grating. The beamline energy resolution was estimated to be 2 × 10-4 with a photon flux of 1–5 × 1012 ph⋅s−1 (200–900 eV range). The photon energy was scanned across Ni L2,3 edge with a 0.2 eV step. More details regarding the PEEM measurements can be found in a previous study [19].Micro-Raman mapping measurements were performed in backscattering geometry at room temperature using an InVia Raman spectrophotometer from Renishaw equipped with an air cooled charged coupled device detector. During the measurements, the confocal mode and a × 100 long working distance Leica objective with a numerical aperture (NA) of 0.75 were used. With this experimental setup, a spectral resolution of ∼1 cm−1 and a spatial resolution of about 0.8 μm (approximating to the relation 1.22·λ/NA, where λ is the wavelength of the incoming laser) could be achieved. The mapping measurements were performed with different values of the laser power varying it from 0.5 to 10 % of the source power (100 mW). The Raman data was collected with 532 nm laser excitation. Before the measurement, the spectrophotometer was calibrated to the first-order vibrational mode of a Si wafer centered at 520.3 cm−1.The catalytic activity of the pristine and synthesized heterostructure was carried out in a photoreactor to measure any activity differences. Pristine MoS2, Ag decorated MoS2 and MAN heterostructure were used to run experiments in the same conditions, such as identical cells, same volume, an equal amount of DI-water, a total weight of 5 mg of each sample, and fixed illumination time of 2 h. A magnetic stirring bar for stirring the suspension was always present during the experiment. The 5 mg samples were added into a quartz bottle with a total volume of 68 mL. For each experiment, 3 s sonication was carried out to provide a consistent starting point in terms of the dispersion of the suspension. The gaseous content was flushed for a fixed repeatable starting point by using argon. In the first round of preparation, we applied careful Ar-flushing, including heating of the water, to minimize the water-dissolved air. The argon flush's primary function was to establish a reliable control method to monitor gas leaks that arose through effusion and, to a lesser extent, diffusion mechanisms, thus detecting potential abnormally high leak rates. By utilizing the available monochromatic wavelengths each at the time, eight experiments per sample were carried out. The repetitive nature of the experiments also provided initial results of the reusability of the samples.For HER measurements, samples were cycled in Perfect Light PCX50B photo reactor through eight monochromatic LED light sources covering near UV regime to visible light (in a range of 365 to 630 nm) equipped with magnetic stirring. Sampling and analysis of gaseous species were carried out using Agilent Micro 490 GC gas chromatograph equipped with an H2 sensitive column (10MS5A) after the end of each illumination round.Pristine MoS2, Ag decorated MoS2 and MAN heterostructure were used to run HER experiments and it was found that MAN system is superior to the pristine MoS2 and Ag decorated MoS2. Thus, MAN is further investigated by employing spectro-microscopic techniques. In MAN, the MoS2 sheets and Ni NP are bonded via Ag NPs. The diameter of the Ag NPs is found to be ∼20 nm, whereas the diameter of Ni NPs is ∼100 nm.The SEM image of MAN is shown in Fig. 1 a. Fig. 1b shows the X-PEEM image while the corresponding XAS spectrum is shown in Fig. 1d. By tuning the synchrotron beam energy from 845 to 875 eV, the XAS covered the main features of Ni L2,3 edges. The main peaks in the Ni 2p XAS are associated with the Ni 2p3/2,1/2 → 3d dipole transitions, separated by spin–orbit splitting of 17.3 eV. The general spectral line shape in the MAN displays similar features to Ni metal foil, suggesting the stability of Ni NPs during the wet chemical synthesis. For the XAS spectra collected from MAN, a small peak turns out at the photon energy of 861 eV. However, its rather weak intensity leaves questions to deduce its origin, e.g., satellite feature of the Ni0, or Ni atoms bonded with S atoms of MoS2, as observed in nickel chalcogenide nanofilms [20]. Even in the latter case, the rather low intensity compared with the main peak show that the bonded Ni atoms are rather few in comparison to that of the Ni-Au-MoS2 system [9]. Nevertheless, a rather low charge transfer between the Ni NPs and the flat MoS2 basal plane beneath and the protruding layer is possible. The metal particle is thus connected to the semiconductor through basal and side contacts. Although X-PEEM uncovers the interaction between MoS2 and Ni, the interaction between Ag and MoS2 and its effect on MoS2 cannot be explored using the same. For this reason, we employed lab source-based Raman mapping. Fig. 1c shows the Raman image of MAN at the same location (as that for SEM and X-PEEM) and corresponding Raman spectra are plotted at various spatial position within MAN (Fig. 1e). Raman spectra of MAN exhibits mainly-two phonon modes, E2g 1 and A1g. Along with these two modes, E1g mode at ∼286 cm−1 and higher intensity of out-of-plane A1g are noted. According to the Raman selection rules, the E1g mode is forbidden in backscattering experiment on the basal plane of bulk MoS2 [21]. However, when the incident light scatters on the surface of edge terminated MoS2, the corresponding scattering Raman tensor undergoes a rotation transformation, leading to a nonzero differential scattering cross-section and hence the E1g mode can be observed indicating film formation of MoS2. Moreover, the peak intensity of the out-of-plane A1g mode is like that of the in-plane E2g 1 mode in the bulk MoS2 and 3 times that of the E2g 1 mode in the MoS2 film. Such preferred excitation of an out-of-plane mode is also consistent with the vertical-aligned crystal texture of the film sample considering the polarization dependence of the Raman scattering cross-section. When measured at spatially different positions, Raman measurements of MAN show variation in the Raman spectra, such as red shift in the phonon frequency of E2g 1 and A1g modes compared to pristine MoS2 (Fig. 1c & e). The observed variation of optical phonons is further investigated using power dependent Raman mapping to probe the effect of Ag NPs on MoS2.Many sets of power dependent (0.5, 1, 5, and 10 %) Raman mapping are collected on various sites to make appropriate observations. At a low laser power (0.5 %), no variation could be seen as E2g 1 and A1g modes are observed consistently at ∼384 and 409 cm−1, which is similar to pristine MoS2. However, when the same region is excited with higher laser power (1, 5 and 10 %), variation in phonon frequency is noted. Representative Raman images for different regions are shown in Figs. 2 & 3 and Fig. S1. All the peak positions are obtained by fitting Raman spectra with Lorentzian line shape (Fig. S2 and Table S1/S2).Increase in the laser power from 0.5 to 1 % led to either no change or slight red shift in both E2g 1 and A1g modes (Fig. 4 ). Enhanced red shift in both E2g 1 and A1g modes is further observed with the increase in laser power to 5 %. However, further increment in laser power led to both red and blue shift in E2g 1 and A1g modes at different locations (Fig. 4 & Fig. S1f). Comparison of different Raman mapping data collected at four different laser power shows that increase in the laser power led to red shift as well as blue shift in E2g 1 and A1g modes at spatially different locations. Moreover, one can see that both E2g 1 and A1g modes follows the same pattern with the change in the laser power.As a further analysis, Fig. S3 shows the average phonon frequency for both the A1g and the E1 2g mode, including all the 13 points investigated in Fig. 3. Fig. S3 reveals that, on average, the laser power leads to red shift in the frequencies (note that the vibrational mode of Si maintains a constant value and is hence used as our internal reference). The slightly larger errors visible for higher power of 5 and 10 % are a direct consequence of the more spread values found while investigating the different spots. The observed spatial variation could be ascribed to i) laser heating, ii) mechanical strain induced due to presence of Ag NPs, and iii) interplay between charge injection and stress.Vasa et al. [16] have reported Raman spectra of MoS2 as a function of excitation power in the range from 0.69 and 21.84 mW. No appreciable peak shift as a function of laser power was noted for either of the two modes. This suggests that MoS2 is quite stable with respect to laser power and observed changes here cannot possibly arise due to increase laser power. Moreover, laser heating give rise to uniform red shift in the phonon frequency and increment in red shift due to increased laser power. However, here we have observed both red and blue shift in E2g 1 and A1g modes at different laser powers. Moreover, different phonon frequencies are observed at spatially different location. Thus, the effect of laser heating can be negated.Observed changes in the Raman spectra could be due to local mechanical strain due to metal-MoS2 contact [14,15,22]. However, this does not seem plausible here as this strain should be laser power independent, whereas we have not observed any shift in phonon frequency at very low laser power. Moreover, as mentioned in the literature, biaxial strain leads to splits in both the E2g 1 and A1g modes, whereas uniaxial strain leads to splitting of mainly E2g 1 mode. However, no splitting is noted here (Fig. 2 and Fig. 3). Moreover, the effects of mechanical strain in thicker MoS2 layers are significantly weaker and dominantly found in monolayer or bilayer. Thus, this possibility can be negated as well.The interplay between charge transfer and compressive stress [13,16] seems to be the most probable cause of the observed power dependent Raman shift in MAN. As per literature, A1g mode corresponding to the structural distortions arising due to the out-of-plane motion of sulphur atoms on either side of Mo atoms preserves the lattice symmetry, whereas E2g 1 mode corresponding to the collective in-plane motion of two layers of the sulphur atoms in the opposite direction of Mo atoms, does not preserve the lattice symmetry, hence only A1g is sensitive to doping [16]. Whereas, in another work, it has been shown that charge-transfer via external perturbation leads to change in both the E2g 1 and A1g modes [13]. Herein, Fig. 4 shows that both E2g 1 and A1g mode follow the same trend indicating that both modes are sensitive to charge transfer and or stress. Since n-doping (p-doping) softens (stiffens) the modes, the observed red shift in the Raman spectra from 0.5 to 5 % suggests n-doping in the present case. In the current case, the electron transfer to the conduction band of MoS2 can occur via LSPR of Ag NPs excited by the laser excitation. Since ELSPR > EMoS2 − EAg, the LSPRs can relax non-radiatively via Landau damping transferring the energy to free electrons in Ag. These electrons subsequently get injected into the adjacent MoS2 leading to red shift in phonon modes. Moreover, laser annealing can cause compression in the MoS2 leading to blue shift in the phonons at higher laser power [13].Thus, it is inferred that power dependent laser excitation of Ag NPs in hybrid structure results in optically tunable electron transfer and compressive deformation of MoS2 by electron injection via non-radiative relaxation of LSPRs excited in Ag NPs. The observed shifts of two modes can be related to the extent of electron transfer concentration and strain and by the relations given as follows [13], (1) ω i = k n i n (2) Δ ω = - 2 γ i ω i 0 γ ε i Equations (1) and (2) enable the calculation of Raman peak frequencies for constant carrier concentration and strain. Here, ε is the lattice strain, n is the charge carrier concentration, γ i is the Grüneisen parameter corresponding to phonon frequency ω i , ω i 0 is the phonon frequency of unstrained and undoped MoS2, and k n is −0.33 cm−1 per 1013 cm−2 e- and −2.22 cm−1 per 1013 cm−2 e- for E2g 1 and A1g mode [13], respectively. Average room temperature γ value is 0.86 and 0.15 for E2g 1 and A1g mode, respectively. Since the observed phonon frequencies at 0.5 % laser excitation is same as that of pristine MoS2, we have chosen E2g 1 ∼ 384 cm−1 ( ω 1 0 ) and A1g ∼ 409 cm−1 ( ω 2 0 ) as the undoped and unstrained value. The inferred laser power dependence of electron transfer concentration and the associated strain averaged over the laser spot size is noted in Table 1 & Table 2 . The electron density as well as strain are different at different spatial locations, even though Raman mapping was collected at same laser power.Observed difference in Raman spectrum at different spatial locations can be ascribed to various possible reasons, such as i) non-uniform growth of Ag NPs, ii) larger size of laser spot (0.8 μm) compared to Ag NPs (10–20 nm) and iii) intercalation of Ag in MoS2 layers. As per electron microscopy imaging, the deposition of Ag is not continuous, rather they form NPs on the surface of MoS2, which can be interpreted as Volmer-Weber island growth mode [23]. Since the growth of Ag is inhomogeneous, and this inhomogeneity can give rise to two regions, one is the intimate contact between Ag clusters and MoS2 surface while the other region is spaced from MoS2 with a notable separation form Ag. Intimate contact between the two will give rise to larger effect of Ag NPs, whereas the effect decreases with increase in the distance between the two. Moreover, observed no changes in the Raman spectra at certain locations (Fig. 4) can be explained with reported density functional theory calculations, which predict that an interface separation of larger than 6 Å between metal and MoS2 is enough to decouple MoS2 from the electronic perturbation of atop metal layers [22]. This further shows that effect of Ni on the Raman spectra of MoS2 can be neglected as Ni is in contact with MoS2 via Ag NPs, thus, spaced far enough to have any effect. Since the size of laser spot is much bigger than the size of Ag NPs, laser exposure covering an area might include different number of Ag NPs, when collecting Raman data at different locations, which can give rise to varied doping concentration and strain value. The third possibility can be the intercalation of Ag between MoS2 layers during the growth of Ag from AgNO3. It is known that intercalation of Ag between layered structure MoS2 is possible [24]. Moreover, intercalation can generate stress in the layered structures along with the possibility of charge transfer [25,26]. In the present case, as mentioned above, AgNO3 is added to MoS2 solution and sonicated for 4 h. Since the growth of Ag on MoS2 follows the Volmer-Weber island growth mode [23], wherein nucleation followed by growth of clusters leads to isolated metal islands on MoS2. The islands keep growing until a continuous and polycrystalline film forms. However, for highly mobile materials, such as Ag, the metal islands change dynamically even near room temperature, where the large islands grow at the expense of the shrinking of small islands. During this process, it can happen that smaller Ag clusters can intercalate within the MoS2 layers. Growth of these clusters will depend on the separation between two adjacent layers, which is around 6 Å in the case of MoS2. Thus, this intercalation can lead to varied charge doping and stress as the laser used for excitation has a depth of focus ∼0.66 μm, thus covering many layers of MoS2. However, the intercalation of Ag within MoS2 layers and its effect on electronic and vibrational properties of MoS2 needs further investigation, which will be explored separately.In order to establish the effect of laser-induced electron transfer and strain, we carried out further power dependent measurements in reverse. In this process, Raman mapping was collected at lowest laser power of 0.5 % after carrying out Raman mapping at 10 % laser power (Fig. 5 & S4). The observed reversible effect further establishes the above interpretation of electron transfer and compressive strain in MAN. Fig. 6 a shows the measured water splitting experiment results for all the samples (bare MoS2 and Ni-Ag-MoS2) under white light illumination. Bare MoS2 show negligible H2 yield, whereas MAN shows significant H2 production. Based on the white light water splitting experiment, we carried out detailed experiment in the visible to UV region for MAN. Fig. 6b shows the H2 production rate of MAN in the light range of 365 nm to 630 nm. As can be seen from Fig. 6b, at the excitation wavelength of 485 and 535 nm, the observed HER performance for MAN is ∼55 μmol g−1h−1. The observed lesser HER at 595 nm could be due to experimental error.A cyclic test on the MAN system was carried out for 3 cycles with each cycle of 2 h under white light irradiation. An Agilent 8860 GC was employed to quantify the evolved H2. The catalyst was stably functional in HER as shown in Fig. S5 of the Supplementary material. The SEM was performed on MAN sample after the cycling tests. Depicted in Fig. S6a, the Ni NPs were firmly attached on to the MoS2 through side and basal contacts. The metals of Ni and Ag coexist with these of S and Mo from the semiconductor matrix according to the element mapping of Fig. S6b. A high-resolution transmission electron microscopic evaluation (HRTEM) was further performed on the sample after the cyclic tests. Both the Ni NPs and flakes are well retained in the composite in Fig. S7(a). The selected area electron diffraction (SAED) pattern further proved that the main crystal structures of MoS2 and Ni were kept in Fig. S7(b) and (c). From the above microscopic and microstructural studies, samples are stable after 3 cycles of photocatalysis. The control experiment by using the Ag-MoS2 was also performed and < 2 µmol/g/h HER yield was found when using the white light as the incident source. Thus, the MAN system is superior to the bare MoS2 and binary Ag-MoS2 systems within the photocatalytic HER abilities. It’s worth noting HER is mainly investigated here and the full water splitting requires dedicated instrumental setups and calibrations along with spectro-microscopic investigations of OER reactive sites to deduce mechanisms. However, a recent work of the similar system [10] showed the oxygen evolution is accompanied with the HER, denoting a full water splitting capability of the MAN system. Beside the same composites, the fraction of Ni used here falls into the range of the MAN system studied previously. Thus, full water splitting capability is also expected on the present sample.The possibility to not only probe locally, but visualize a chemical structure, composition, conformational state, and effect of various components on heterostructures and its catalytic activity has stimulated the development of imaging techniques. Both the spectro-microscopic techniques used here are fingerprint, rich, non-destructive and give information of different interactions (Fig. 7 ). MoS2 being a cheap and abundant mineral has promising applications in transistors, optoelectronics, and UV–vis light convertors. Furthermore, it has photocatalytic abilities in degrading organic pollutants. Despite these achievements, the MoS2 is a poor photocatalysts. However, the chemical robustness could be enhanced by heterostructure engineering. Herein, Ni-Ag-MoS2 contact is established using facile wet chemical synthesis method. Lab-based Raman mapping and synchrotron-based X-PEEM verified the successful bonding of Ni to the layered MoS2 at the nanoscale interface regions via the Ag buffer (Fig. 7). Moreover, power dependent Raman mapping showed the same pattern of downshift or upshift in the phonon frequency of E2g 1 and A1g modes with the increase in laser power due to various possibilities, such as non-uniform growth of Ag NPs, the contact between MoS2 and Ag, larger laser spot size covering different concentration of Ag NPs, and intercalation of Ag in layered MoS2 (Fig. 7).A side view of ternary structure scheme is illustrated in Fig. 8 a denoting metal/semiconductor contact via the MoS2 basal and side modes. In the first mode, no biaxial strain from the dichalcogenide is needed when joining the Ni NPs onto the MoS2 basal with the Ag as the buffer. This is supported by the absence of splits in both the E2g 1 and A1g of the Raman spectra in Figs. 2 and 3. In fact, the lattice of Ag (111) buffer well matches the one of the MoS2, providing a metallic contact between the metal and semiconductor [10]. In the second case, the silver is involved in bonding the Ni NPs and the atoms at the MoS2 edge. As a result, the M/S contact was also metallic [9] when the noble metal buffers the Ni and the MoS2′s defective side [27,28]. A band alignment scheme is depicted in Fig. 8b following the spectro-microscopic determinations. The figure adopts the work functions of 5.35 eV, 4.74 eV, and 5.20 eV for Ni, Ag and MoS2, respectively, [10,29] and the metal–semiconductor (M/S) contact model of metal-induced gap state [30]. Therein, the interface dipole was formed by the charge transfer across the bonds at the M/S interfaces. The reference level is set to the charge neutrality level (CNL), similar to Fermi level in semiconductor itself and referring to the highest occupied surface state for the common surface [31]. Here the CNL is very close to the Ag Fermi level due to Ag’s buffer status and a lower work function of the Ag than the MoS2. As per the band alignment scheme when Ni NPs and MoS2 are joined by Ag NPs, free electrons will transfer from Ni to MoS2 via Ag due to the work function difference. This leads to the accumulation of electrons next to the valence band of MoS2 adjacent to the interface region and the decrease of contact resistance which is contributing to the observed higher H2 yield by MAN [32]. The Fermi levels of the Ni, MoS2, and the interface region will be aligned after thermodynamic equilibrium [33], resulting in the band bending of MoS2. During the photocatalysis, electron-hole (e--h+) pairs are first created on the semiconductor MoS2 matrix. Following the aligned bands, the photoexcited electrons from the MoS2 can thus move easily from the valence band of MoS2 to the metal side, as shown in the Fig. 8(b). As a result, the e--h+ recombination is substantially inhibited [34], and a longer time window is left for the water redox than in the pure semiconductor case. Holes in the MoS2 oxide led to formations of the OH· and H+ radicals. The protons get reduced by the electrons to form H·, and later H2. Remaining OH· possibly combines to form the H2O2 or partially stack on the metal sides. In the former case, the product gets easily dissolved and emits oxygen under vigorous stirring and light irradiations. The latter one results the increase of the nickel hydroxides along with the native oxides [10] which indeed benefits the oxygen evolution reactions [35].”Spectro-microscopic X-PEEM and Raman mappings have efficiently been used to probe Ni-Ag-MoS2 heterostructure. The interaction of Ni with MoS2 is evidenced through synchrotron X-PEEM, whereas the interaction mechanism at the Ag-MoS2 interface is probed via Raman mapping. The large variation in E2g 1 and A1g phonon modes in Ni-Ag-MoS2 with the increase in laser power and pristine like behavior during reverse power dependent measurement is observed. These variations are assigned to non-uniform growth of Ag NPs and their intimate contact with MoS2, larger laser spot size covering different concentration of Ag NPs, and Ag intercalated between layered MoS2. These observations further reveal compensation between downshift in E2g 1 and A1g modes due to charge carrier injection and upshift in E2g 1 and A1g modes due to laser annealing. The improved photocatalytic activity of Ni-Ag-MoS2 heterostructure (H2 yield ∼ 55 μmol h−1 g−1) compared to pristine MoS2 (H2 yield < 5 μmol h−1 g−1) is attributed to successful bonding of Ni, Ag and MoS2. Thus, an interesting possibility of achieving a tunability between electron injection and strain is attained by employing laser induced charge doping and stress. This LSPR-induced electron injection offers unique possibility of spatially localized dynamical electron doping and provides an active manipulation and tuning of 2D semiconductors.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 acknowledge Academy of Finland grant #311934, and The University of Oulu and The Academy of Finland Profi5 - project #326291. W. C. acknowledges funding provided by European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement 101002219). Authors gratefully acknowledge the Center of Materials Analysis (CMA), University of Oulu for characterizations and Dr. J. Fernández-Catalá and Dr. R. Greco for helping with GC operations. Funding from the Knut and Alice Wallenberg Foundation (Wallenberg Academy Fellowship award, 2016-0220) is kindly acknowledged. We acknowledge MAX IV Laboratory for time on Beamline (MAXPEEM) under Proposal (20200401). Research conducted at MAX IV, a Swedish national user facility, is supported by the Swedish Research council under contract 2018-07152, the Swedish Governmental Agency for Innovation Systems under contract 2018-04969, and Formas under contract 2019-02496.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcat.2022.09.006.The following are the Supplementary data to this article: Supplementary data 1

Despite the boom in catalytic response via constructing interfaces, understanding interfaces’ interaction in heterostructures is still a paradox. In this work, the interaction of Ni with MoS2 in Ni-Ag-MoS2 heterostructure are unveiled through synchrotron X-PEEM and what’s more, the missing interaction mechanism at the Ag-MoS2 interface is probed via Raman mapping. The observed competition between the downshift of the E2g 1 and A1g modes due to charge carrier injection and the upshift of the E2g 1 and A1g modes due to compressive strain during reverse laser power experiment is assigned to the non-uniform growth of Ag nanoparticles, their intimate contact with MoS2, and Ag intercalated layered MoS2. The substantial improvement of the H2 yield of the Ni-Ag-MoS2 (∼55 μmol h−1 g−1) over the pristine MoS2 and the binary Ag-MoS2 evidence successful bonding of Ni, Ag and MoS2. This study highlights the importance of considering both electronic coupling and strain to optically tune electromechanical properties of MoS2.

Heterogeneous catalysis plays a significant role in synthesizing organic compounds for agrochemicals, pharmaceuticals, and fine chemicals for a sustainable future. These catalysts have contributed to developing technology to produce green chemicals from biomass-derived platform chemicals under environmentally benign processes and avoid using toxic or hazardous materials [1]. However, the design and development of novel eco-friendly solid catalysts require appropriate techniques and strategies for improved recoverability, recyclability, and eco-friendliness [2]. In addition, there is an urgent need to acquire new organic compounds which might find possible applications in diverse areas of study. Hence more environmentally and eco-friendly heterogeneous catalysts are being sought to expedite the synthesis of nitrogen-based heterocycles because of their enhanced biological activities [3].Porous hexagonal boron nitride (h-BN) is a material that demonstrates unique physical and chemical properties, including low density, high specific surface area, high thermal conductivity, oxidation resistance, and chemical durability [4, 5]. These features make h-BN a promising catalyst for their application in various research areas, especially those related to adsorption, like gaseous uptake and pollutant adsorption. Furthermore, h-BN is excellent catalyst support [6] since they possess a hexagonally shaped crystal structure composed of continuous boron-nitrogen bonds, wherein lone pair of electrons on the nitrogen atoms can coordinate with certain metals. The hBN/metal substrates functions as supports for metal clusters with Cr(110), Molybdenum(110), Rhenium(0001), Iron(110), Iridium(111), Copper(111), Gold(111), Silver(111), Nickel(111), and Platinum(111), metal-organic complexes, and organic molecules [7]. Water-soluble and porous boron nitride is usually biocompatible and can effectively employ as a nanocarrier for loading anti-cancer drug doxorubicin [8]. Moreover, several studies supported the use of interactions of h-BN through calcium coordination bonding towards reduced graphene oxide yielded nanocomposites to enhance the mechanical, electrical and thermal properties substantially [9, 10, 11]. Therefore, we aimed to investigate a metal and concoct a novel h-BN metal-based material that could be used as a catalyst to synthesize new nitrogen heterocycles. Hence, we investigated the suitability of a novel zinc boron nitride (Zn-BNT) material to synthesize new benzimidazoles through a simple condensation reaction.Although benzimidazoles are N-based heterocycles are important scaffold for medicinal or pharmaceutical compounds [12, 13], they are also display a wide variety of biological activities such as anti-microbial [14], anti-viral [15], anti-cancer [16], anti-protozoal [17], anti-inflammatory [18], and analgesics [19]. They also exhibit significant antiviral activity against different viruses, including HIV (AIDS), HSV-1, influenza, and HCMV [20, 21, 22, 23]. Some selected medicinal compounds containing benzimidazole moiety are presented in Figure 1 .Benzimidazole scaffold is widely used in the organic synthesis of drugs and drug intermediates [24]. Principally, two protocols are adopted for synthesizing 2-substituted benzimidazoles and their analogs. The first protocol involves a coupling of 1,2-phenylene-diamines with carboxylic acids, whereas the second protocol involves condensation of 1,2-phenylenediamine with aldehydes followed by oxidative cyclodehydrogenation [25, 26]. Benzimidazole moiety can be constructed through reactions involving the use of several types of catalysts such as ammonium chloride [27], alumina [28], sodium metabisulphite [29], lanthanum chloride [30], indium triflate [31], sodium hexafluroaluminate [32], nickel acetate [33], iodine [34], and sodium dodecyl sulfate [35]. At the same time, some biologically active quinoline molecules are synthesized using boron modified sulphonic acid catalyst [36, 37, 38]. However, the methods have several drawbacks, such as the use of hazardous organic solvents, strongly acidic conditions, high moisture-sensitivity or tedious workup conditions, low yields, and difficulty isolating the products from the reaction mixture, and strong oxidizing nature of the reagents employed. Hence our research thrust was to prepare a new catalyst and evaluate its potential to synthesize benzimidazole derivatives. Furthermore, the theoretical calculations are studied for FT-IR and NMR spectra, HOMO-LUMO gap, Mulliken charge analysis, molecular electrostatic potential map, and Fukui function analysis.All the reagents were purchased from commercial sources and used as received. The reactions' progress and the compounds' purity were monitored by thin-layer chromatography (TLC) on pre-coated silica gel plates procured from E. Merck and Co. (Darmstadt, Germany). TLC spots were visualized by UV light and using an iodine-vapor chamber. The melting points of the synthesized compounds were determined using a Stuart SMP 10 melting point apparatus and are uncorrected. The IR spectra were recorded on Varian Scimitar 1000 FT-IR using KBr pellets, and the absorption frequencies are expressed in reciprocal centimeters (cm−1). In addition, 1H and 13C NMR were recorded on either a Bruker 400 or 600 MHz spectrometer: DMSO-d 6 was the solvent while TMS was an internal reference. The chemical shift values are recorded on δ scale, and the coupling constants (J) are in hertz. The elemental analyses (C, H, N) were obtained from a PerkinElmer precisely 2400 analyzer. Differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA) were conducted using TA Instruments. A Carl Zeiss Ultra Plus scanning electron microscope with EDX detector was used. The X-ray diffraction analysis was conducted with a Philips PW 1050 diffractometer set at 1°/min with a scanning step size of 0.02° from 40° to 100° 2Ѳ using monochromated Cukα radiation. Data were captured with a sietonics 122D automated microprocessor linked to the diffractometer [39].To a solution of Zn(OAc)2 (24.9 mg), in acetonitrile (50 ml), was added boron nitride (2.50 mg, 0.1 mmol), and the suspension was stirred at room temperature for seven days under an inert atmosphere. The resulting suspension was filtered, and the solid was washed with aqueous methanol and dried under reduced pressure to give Zn-BNT as a white powder in 95% yield.Equimolar quantities (1 mmol) of o-phenylenediamine (1) and aromatic aldehydes (2a–h) in acetonitrile (6 ml) were transferred to a 50 ml round bottom flask, and Zn-BNT (10 mol%) was added. The mixture was heated under reflux for12 h, and TLC was used to monitor the progress of the reaction mixture. Upon completion of the reaction, the mixture was filtered to recover the catalyst. The solid was washed with chloroform followed by methanol and dried at 120 °C while the filtrate was purified by silica-gel column chromatography using ethyl acetate and petroleum ether 10:90 (v, v 10%) as an eluent to yield the 2-arylbenzimidazoles 3a–h, quantitatively.The spectra data is presented below:White solid; Yield: 97%; mp: 132–134 °C; FT-IR (ATR, ν max , cm−1): 3321 cm−1 (N–H stretching), 3177 cm−1 (=CH), 1625 (C=N), 1551, 1467, 1422 (Ar–C=C); 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.91 (1H, s, N–H), 8.20–8.22 (2H, m, Ar–H), 7.45–7.62 (5H, m, Ar–H), 7.20–7.26 (2H, m, Ar–H) ppm. 13C-NMR (100 MHz, DMSO-d 6 ): δ = 151.18, 130.07, 129.82, 128.92, 128.78, 126.38, 122.09 ppm. Anal. Calc. for C13H10N2: C, 80.39; H, 5.19; N, 14.42%. Found: C, 80.40; H, 5.21; N, 14.44%.Brown solid; Yield: 95%; mp: 296–298 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.98 (1H, s, N–H), 7.93 (1H, m, Ar–H), 7.60–7.62 (1H, d, J = 8 Hz, 1H, Ar–H), 7.47–7.49 (1H, d, J = 8 Hz, Ar–H),7.22–7.17 (3H, m, Ar–H), 6.71 (1H, m, Ar–H), ppm. Anal. Calc. for C11H8N2O: C, 71.73; H, 4.38; N, 15.2%. Found: C, 71.75; H, 4.40; N, 15.23%.Yellow solid; Yield: 90%; mp: 294–296 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.82 (1H, s, N–H), 7.96–8.03 (2H, m, Ar–H), 7.86–7.88 (1H, m, Ar–H), 7.73–7.77 (1H, m, Ar–H), 7.59–7.62 (2H, m, Ar–H), 7.23–7.25 (2H, m, Ar–H), ppm. Anal. Calc. for C13H9N3O2: C, 65.27; H, 3.79; N, 17.56%. Found: C, 65.29; H, 3.81; N, 17.58%.yellow solid; Yield: 88%; mp: 306–308 °C; FT-IR (ATR, ν max , cm−1): 3321 cm−1 (N–H stretching), 3177 cm−1 (=CH), 1625 (C=N), 1551, 1467, 1422 (Ar–C=C); 1H NMR (400 MHz, DMSO-d 6 ): δ 13.60 (1H, s, N–H), 8.39 (1H, d, Ar–H), 8.15 (1H, d, 1H, Ar–H), 8.00 (1H, d, 1H, Ar–H), 7.64–7.66 (1H, dd, Ar–H), 7.25–7.27 (1H, dd, Ar–H) ppm. Anal. Calc. for C13H9N3O2: C, 65.27; H, 3.79; N, 17.56%. Found: C, 65.29; H, 3.81; N, 17.57%.Brown solid; Yield: 89%; mp: 280–282 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.76 (1H, s, N–H), 8.17 (2H, m, Ar–H), 7.58–7.63 (4H, m, Ar–H), 7.20–7.22 (2H, m, Ar–H), ppm. Anal. Calc. for C13H9ClN2: C, 68.28; H, 3.97; N, 12.25%. Found: C, 68.30; H, 3.98; N, 12.27%.White solid; Yield: 93%; mp: 186–188 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.91 (1H, s, N–H), 8.20–8.24 (1H, q, Ar–H), 7.74–7.78 (1H, q, Ar–H), 7.72 (1H, t, Ar–H), 7.49 (1H, t, Ar–H), 7.34–7.42 (1H, m, Ar–H), 7.25 (1H, t, Ar–H), 7.10 (1H, t, Ar–H), 7.00–7.03 (1H, q, Ar–H) ppm. Anal. Calc. for C13H9FN2: C, 73.57; H, 4.27; N, 13.20%. Found: C, 73.59; H, 4.29; N, 13.21%.Greyish solid; Yield: 82%; mp: 178–180 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.82 (1H, s, N–H), 8.04 (2H, m, Ar–H), 7.54–7.56 (2H, m, Ar–H), 7.34–7.36 (2H, m, Ar–H), 7.16–7.20 (2H, m, Ar–H), 2.34 (3H, s, CH3), ppm. Anal. Calc. for C14H12N2: C, 80.74; H, 5.81; N, 13.45%. Found: C, 80.76; H, 5.83; N, 13.47%.White solid; Yield: 94%; mp: 222–224 °C; 1H-NMR (400 MHz, DMSO-d 6 ): δ = 12.72 (1H, s, N–H), 9.61 (1H, s, Ar-OH), 8.03–8.06 (1H, m, Ar–H), 7.70–7.72 (2H, m, Ar–H), 7.27–7.36 (2H, m, Ar–H), 6.88 (2H, m, Ar–H), 6.60 (1H, m, Ar–H), ppm. Anal. Calc. for C13H10N2O: C, 74.27; H, 4.79; N, 13.33%. Found: C, 74.29; H, 4.81; N, 13.35%.The new catalyst Zn-BNT (Figure 2 ) was synthesized through a simple reaction between boron nitride, zinc acetate in acetonitrile under an inert atmosphere by stirring at room temperature for seven days. The novel catalyst was characterized using FT-IR, XRD, SEM, SEM-EDX, SEM-mapping, BET, and DSC-TGA.The FT-IR spectrum of Zn-BNT is presented in Figure S1 (Supporting Information). The FT-IR revealed the presence of functional bond stretching frequencies N–B–N at 1738 cm−1, B–N at 1365 cm−1, and B–N–B at 920 cm−1, with other additional bands for Zn–N at 1365 cm−1 and 1217 cm−1 assigned as asymmetric and symmetric stretching frequencies, respectively.The powder XRD diffraction analysis [5] showed the crystal-like Zn-BNT. The characteristic Bragg's XRD peaks at 2θ boron nitride peaks are observed at 26.67°, 41.62°, 55.02° and 75.91° are indexed to the (002), (100), (004), (220) while the Zn peaks 43.81° and 50.15° were indexed to (111), (012), respectively (Figure 3a and 3b).Zn-BNT was analyzed by scanning electron microscopy (SEM) for morphological characteristics. The particles appeared to be randomly distributed, spherical and elongated needle-shaped for 1% Zn-BNT (Figure 4 a). An agglomeration of particles was observed, which probably led to the formation of the larger particles for 2% Zn-BNT (Figure 4b). Consistent morphology and clusters of metal particles on the support surface of 4% Zn-BNT was observed in Figure 4c. The size of the rod-like particles of Zn was observed as 50–150 nm (length) and 16 nm (width) on the BN surface for 4% Zn-BNT (Figure 4d). The particles observed in 1% Zn-BNT were small and have a high surface area with a smaller number of active sites compared to 2% Zn-BNT. Similarly, the particles appeared larger and therefore had a smaller surface area in 4% Zn-BNT compared with 2% Zn-BNT.The particles were observed as small and had a high surface area with a smaller number of active sites. Furthermore, the SEM mapping indicated the elements of Zn-BNT (Figure 5 a, b, c, and d) Zn-BNT, boron, nitrogen, and zinc. The morphology of Zn-BNT material was further confirmed by SEM-EDX analysis.The SEM-EDX spectrum of Zn-BNT (Figure 6 ) displayed elements B, N, O, and Zn of weight (%) 20.11, 9.13, 70.72, and 0.04, respectively. Au peaks were due to the sample being coated with Au in the sampling process.Brunauer-Emmett-Teller (BET) specific surface was determined from the nitrogen adsorption data at the relative pressure using a multipoint method. Figure 7 (a) and (b) shows the isotherms of a 2 and 4% Zn-BNT in N2 adsorption-desorption. The sample exhibited type-III isotherms. The type-III isotherm is closely related to the mesoporous structure of the catalyst and therefore no identifiable monolayer formation; the adsorbent-adsorbate interactions are now relatively weak and the adsorbed molecules are clustered around the most favorable sites on the surface of the solid. 2% Zn-BNT material showed a surface area of 25.42 m2 g−1 with a pore volume of 0.119 cm3 g-1 and 4% Zn-BNT material showed a surface area of 23.58 m2 g−1 with a pore volume of 0.114 cm3 g-1 and a pore size 188.74 Aº thereby indicating a mesoporous structure and effective as a catalyst for organic reactions.Zn-BNT was analysed for its thermal properties from room temperature to 800 °C. Figure 8 presents the DSC-TGA profile of the Zn-BNT material. The DSC curve of the sample indicated a broad exothermic peak. The weight loss occurred in stages; about 1.2% loss up to 200 °C due to moisture and other volatile materials, whereas a loss of about 1.8% occurred until 800 °C. These results demonstrated that Zn-BNT has good thermal stability; hence, it can be used effectively as suitable for the catalyst for several organic reactions.Briefly, equimolar quantities of o-phenylenediamine (1) and differently substituted aromatic aldehydes (2a–h) were added in acetonitrile. Zn-BNT was added, and the mixture was subjected to microwave irradiation for 15 min. The crude product was purified by silica-gel column chromatography using ethyl acetate and petroleum ether (10:90) to yield the pure 2-arylbenzimidazoles (3a–h) quantitatively. The recovered catalyst was reused for other reactions. The facile, one-step synthetic route is shown in Scheme 1 . Notably, various aromatic aldehydes containing both electron-withdrawing and electron-donating groups underwent condensation to afford the target compounds. The results show that aromatic aldehydes bearing electron-donating groups such as hydroxyl and methyl afforded corresponding 2-arylbenzimidazole derivatives in better yields than the aldehydes having electron-withdrawing groups.The effect of different solvents, such as protic and aprotic, with different polarities were investigated (Table 1 ) on the model reaction to synthesize 3a. An initial reaction of ortho-phenylenediamine with benzaldehyde, in acetonitrile under reflux for 12 h produced a 75% yield of 3a. Protic solvents such as methanol and ethanol and aprotic solvents such as toluene, dichloromethane, tetrahydrofuran, and dioxane were investigated. The yield of 3a varied with the nature of the solvents used in the reaction. Since dioxane and acetonitrile gave the highest yield of 3a, we selected these two solvents and used microwave irradiation to decrease the reaction time and increase the yield of 3a. For a reaction time of 15 min, the yield of 3a was 55% and 97% for dioxane and acetonitrile, respectively (Table 1, entry 8–9). Published data on benzimidazole synthesis was analyzed and presented in Table 2 . It was obvious that Zn-BNT outclassed published data; however, further work should be undertaken with named catalysts and under similar reaction conditions. Although the synthesis of the Zn-BNT material was recorded as seven days, the catalyst showed excellent catalytic activity, is environmentally friendly, non-toxic, and produces a high product yield (Table 2).After optimizing the reaction, 3a–h were synthesized using acetonitrile and Zn-BNT under MW conditions at 140 °C for 15 min. Appropriate ortho-phenylenediamine 1 and different aldehyde (2a–h) were used. The yield of the products (Table 3 ) ranged from 82 to 97%. The products 3a–h were characterized by spectroscopic techniques FT-IR, 1H-NMR, 13C-NMR, and elemental analysis. The spectral confirmation of 3a is discussed as a template analysis. The FT-IR spectrum displayed the characteristic vibrational band as a single sharp peak at 3321 cm−1, which indicated the formation of new aromatic secondary amino -N-H stretch, while there was no shoulder peak in the region of 3500–3400 cm−1, which would strongly indicate the NH2. A strong peak at 1625 cm−1 demonstrated the C=N stretch. Also, there was no signal for the C=O stretch in the region of 1680–1715 cm−1, indicating the absence of the aromatic aldehyde, one of the starting materials. Thus, the new C=N bond and N–H bond confirmed the main functional groups of 3a. Moreover, the characteristic peaks at 1551 cm−1, 1467 cm−1 and 1422 cm−1 indicated the Ar–C=C stretch and a sharp peak at 3177 cm−1 represented the = CH stretch.The 1H-NMR spectrum of 3a showed a distinctive N–H proton broad singlet at δ 12.91, and no broad singlets were observed in the region of δ 5–7 ppm corresponding to the starting compound. The 13C-NMR displayed the N=C–N signal at δ 151.18 while the remaining 10 aromatic carbons appeared in the range δ 130.07–122.09 ppm. Elemental analysis revealed Anal. Calc. for C13H10N2: C, 80.39; H, 5.19; N, 14.42%. Found: C, 80.40; H, 5.21; N, 14.44%.After optimizing the reaction, 3b–h was synthesized using acetonitrile and Zn-BNT under MW conditions at 140 °C for 15 min. Appropriate ortho-phenylenediamine 1 and different aldehyde (2a–h) were used. The yield of the products (Table 3) ranged from 82 to 97%. The products 3a–h were characterized by spectroscopic techniques FT-IR, 1H-NMR, 13C-NMR, and elemental analysis.The condensation reaction was conducted using acetonitrile as the solvent at 140 °C and different amounts of 2% Zn-BNT catalyst in the range of 0.05–1 g. The best yield of product 3a was obtained with 0.07 g of the catalyst. Higher amounts of the catalyst did not improve the yield. The catalyst could be quickly recovered by simple filtration of the reaction mixture, followed by washing with chloroform and then methanol: the solid was dried at 105 °C and then used for subsequent reactions. The reusability potential of Zn-BNT was also investigated in the model reaction to synthesize 3a. Recovery of the catalytic activity of Zn-BNT was better than other catalysts. Zn-BNT could be reused up to 8 times with only 5% loss of catalytic activity (Table 4 ), indicating good potential if undertaken on an industrial scale for synthesizing any benzimidazole derivatives.The ground state (DFT) optimization energy calculations were performed using the hybrid exchange-correlation functional B3LYP [40] with the basis set for the 6–311++G(d,p) [41]. The various possible conformers for compound 3f were optimized in the gas phase. The conformer with the lowest energy with the fundamental frequencies in the ground state was considered and shown in Table S1 and Figure S13 (Supplementary Information). The total energy and Cartesian coordinates for the optimized minimum energy structure are given in Table S2. The infra-red (IR) spectrum and frequency values of 3f from DFT calculations are shown in Figure S15 and Table S3. Further, the NMR shielding tensors computed 3f with the gauge-independent atomic orbital (GIAO) method is shown in Figure S16 and Table S4. All the calculations were carried out with the Gaussian09 program package [42].The energy level of the highest-occupied molecular orbital (EHOMO) and lowest-unoccupied molecular orbital (ELUMO) energies, along with the EHOMO-LUMO gap is an important parameter [43] to evaluate the reactivity of 3f. The HOMO-LUMO plot is given in Figure 9 . The energy gap value of 3f was EH-L = 4.48 eV. The large EH-L energy gap denotes the compound's high stability and lower reactivity.The atomic charge distribution of 3f is obtained using Mulliken charge analysis [44] in Figure S14. The nitrogen and fluorine atoms show negative charge distribution due to electronegative properties. Consistently, the hydrogen zones represent a positive charge distribution in compound 3f.The electrostatic potential surface (ESP) [45] ranges an isosurface from −0.02 a.u. to 0.02 a.u., as shown in Figure 10 . The ESP map showed the high electron density for the blue region, which is due to higher electronegativity (nitrogen zones). In contrast, the red region denoted low electron density (hydrogen zones). The results mentioned above indicated that the C–N bond is an active site.Fukui function predicts the most probable sites of the nucleophilic, electrophilic, and radical attack selectivity and chemical reactivity using DFT calculation shown in Figure S17 [46]. The Fukui function indices on the i th atomic site, for nucleophilic (f i +), electrophilic (f i −), and free radical (f i 0) are following equations, (1) f i + = q i ( N + 1 ) − q i ( N ) (2) f i − = q i ( N ) − q i ( N − 1 ) (3) f i 0 = 1 2 [ q i ( N + 1 ) − q i ( N − 1 ) ] The f i + , f i − and f i 0 is represent the nucleophilic, electrophilic, and free radical attack on the reference compound 3f. From Eqs. (1), (2), and (3), qi denotes the atomic charge at the ith atomic sites of the chemical species are anionic (N + 1), cationic (N − 1), and neutral (N), respectively. Here, the Fukui function for f i + , f i − and f i 0 attack is calculated using Mulliken charge analysis.The dual descriptor ΔF(r), the difference between the nucleophilic ( f i + ) and the electrophilic ( f i − ) Fukui function as the given equation is [47], (4) Δ F ( r ) = [ f i + − f i − ] Eq. (4) represents the nucleophilic and electrophilic (electron added and removed from the LUMO and HOMO) sites given in Figure 9. The nucleophilic and electrophilic attack sites are favored, whereas dual descriptor ΔF(r) > 0 and ΔF(r) < 0. Here, the dual descriptor ΔF(r), is the clear distinction between nucleophilic and electrophilic attacks at a particular site with positive/negative values. The dual descriptor values are presented in Table 5 which indicated that the nucleophilic reactivity of 3f was in the following order N7 > H12 > C1> H13 > C5 > F24 > H10 > H11 > C23 > H22 > H21 > C4 > C19 > H18 > H25 > C8 > H17 > C14 > N9 > C20 > C3 > C16 > C6 > C2 > C15 and electrophilic reactivity in the order C16 > C15 > C1 > C23 > N7 > C4 > H22 > H21 > C8 > H12 > H13 > F24 > H10 > H11 > H18 > H17 > C5 > H25 > C19 > N9 > C20 > C6 > C3 > C2 > C14 while the free radical attack reactivity order was C1 > N7 > C16 > H12 > C23 > H13 > F24 > H10 > H22 > H11 > H21 > C4 > C5 > C8 > C15 > H18 > C17 > C19 > C25 > N9 > C20 > C6 > C3 > C2 > C14. Furthermore, the condition of dual descriptor, the positive values denoted nucleophilic site (F(r) > 0) for C2, C3, C5, C6, N7, N9, H10, H11, H12, H13, C14, C19, C20, and F24. Similarly, the electrophilic site was negative values (F(r) < 0) are C1, C4, C8, C15, C16, H17, H18, H21, H22, C23, and H25. Finally, behavior for the nucleophilic, electrophilic, and free radical attack indicated the highly reactive sites were N26 (0.096), C8 (1.611), and C8 (0.801).In this work, we developed a facile, one-pot synthesis of 2-substituted benzimidazole derivatives through a reaction between o-phenylenediamine and different aromatic aldehydes in the presence of a novel zinc-boron nitride catalyst. The advantages of the developed method include the environmental-friendly reaction conditions, simple operation, broad substrate scope, satisfying yields, easy isolation of the product, and reusability of the catalyst. Furthermore, 2-(4-fluorophenyl)-1H-benzo[d]imidazole (3f) was selected for a computational study of the IR and NMR spectrum, which matched the experimentally generated spectra. The HOMO-LUMO gap was calculated as 4.48 eV. Moreover, the novel catalyst could be employed to synthesize other heterocyclic compounds, and their biological activity can be evaluated.Sureshkumar Mahalingam, Arul Murugesan: Performed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.Thangaraj Thiruppathiraja, Senthilkumar Lakshmipathi: Performed the experiments; Wrote the paper.Talent Raymond Makhanya: Performed the experiments; Analyzed and interpreted the data.Robert Moonsamy Gengan: Conceived and designed the experiments; Analyzed and interpreted data.Sureshkumar Mahalingam was supported by National Research Foundation (NRF) of South Africa (Grant No. 98725).The data that has been used is confidential.The authors declare no conflict of interest.Supplementary content related to this article has been published online at [URL].Supplementary content related to this article has been published online at https://doi.org/10.1016/j.heliyon.2022.e11480.The following is the supplementary data related to this article: Supplementary information Supplementary information

A new zinc-based boron nitride (Zn-BNT) material was synthesized from boron nitride and zinc acetate in 95% yield. The morphological and spectroscopic properties of Zn-BNT were elucidated by SEM, XRD, BET, DSC-TGA, and FT-IR. Zn-BNT catalyzed the synthesis of benzimidazoles (3a–3h) through a reaction between o-phenylenediamine and different aromatic aldehydes under microwave conditions for 15 min. The compounds were purified by silica-gel chromatography. The synthesized compounds were characterized by FT-IR, 1H-NMR, 13C-NMR, and elemental analysis. Zn-BNT was reused eight times with only a 5% loss of catalytic activity. Furthermore, 2-(4-fluorophenyl)-1H-benzo[d]imidazole (3f) was selected for a computational study of the IR and NMR spectrum, which matched the experimentally generated spectra. The HOMO-LUMO gap was 4.48, and the Fukui function analysis showed high activity in the reactive sites.

Data will be made available on request.An important class of ligands known as Schiff bases has found wide applications in the coordination chemistry of inner transition, transition and main group elements [1–5]. Generally, aromatic or aliphatic amines and carbonyl compounds are condensed by nucleophilic addition to produce Imine bases [6–12]. All the metal complexes bind to DNA in a non-covalent manner on the groove through electrostatic bonding and intercalation [13–18]. Designing less harmful, more affordable and non-covalently bound novel chemotherapeutic medicines is urgently needed to solve the challenge in anticancer treatment [18–25]. The synthesis of a novel series of benzothiazole Schiff base metal complexes is what we have concentrated on due to their diverse range of pharmacological actions, including antibacterial, anticancer[26], anti-inflammatory[27], anti-tuberculosis[28], antioxidant[29], anticonvulsant[30] and anthelmintic[31]. Schiff bases are among the compounds most thoroughly explored in medicinal chemistry. Schiff bases are employed as stabilisers for polymers, catalysts, dyes and organic synthesis intermediates [32–34]. The benzothiazole-derived Schiff base is used in a wide range of analytical, biological, inorganic, medical and pharmaceutical applications [35–38].Synthesis of 2-EBTMCP ligand and their metal complexes is presented in this research article. All the Metal complexes of Co(II), Ni(II), Cu(II) and Zn(II) prepared from 2-EBTMCP ligand. The Physical-chemical spectrum techniques and their analytical data have been used to characterise the spectral properties of the 2-EBTMCP ligand and associated 3d series transition metal complexes. The biological activities of DNA binding studies and MMT test was used to screen the cytotoxicity investigations against the cervical carcinoma (HeLa) and breast adenocarcinoma (MCF-7) cell lines.TLC (Thin Layer Chromatography) method was used to check the purity of the newly prepared 2-EBTMCP ligand and their metal complexes. On a Polmon MP-96 model instrument, measurements of the melting points of the ligands and their metal complexes were obtained.The SHIMADZU Perkin-Elmer Infrared model was used to record FT-IR spectral analysis of ligand and metal complexesin a range of 550 to 4000 cm−1.On a Bruker 400 MHz NMR instrument, the 1H NMR/13C NMR of the ligand and metal complex was recorded using DMSO‑d 6 as the solvent and TMS as the internal reference.SEM was used to examine the surface morphology of the CT complex (scanning electron microscope on Zeiss evo18). EDX spectra were used to analyze the elemental composition of the ligand and metal complexes (SEM, Quanta FEG 250). The device was powered by a 20 kV acceleration voltage.A Rigaku MiniFlex 600 X-ray diffractometer was used to perform powdered XRD analysis. Cu-Kα radiation (λ = 1.5406 A°) in the range 2θ from 5 to 80° with a step size of 0.02° and a scan step time of 0.15 s.6-ethoxy-2-amino benzothiazole and 5-Chloro-2-hydroxy benzaldehyde was purchased by Sigma-Aldrich. The HPLC solvents, which included acetone, chloroform, n-hexane, ethyl acetate methanol, Dichloromethane, acetic acid, Hydrochloric acid (32%), Triethylamine and aqueous ammonia (25%) solutions, were procured from SD fine AVRA company. The metal chloride salts CoCl2·4H2O, NiCl2·4H2O, CuCl2·4H2O and ZnCl2·4H2O were purchased from AVRA Company. MMT dye purchased from Bangalore Genei in Bengaluru India and kept at −20° C.To an equimolar quantity of 5-Chloro-2-hydroxybenzaldehyde (0.279 g, 3 mmol) dissolved in hot methyl alcoholcarbinol (10 ml) solution, 6-ethoxy-2-aminobenzothiazole (0.194 g, 3 mmol) was added. The final mixture was refluxed for five-six hours at 75–80 °C. After the reaction was finished, a solid precipitate with a yellow tint emerged. This was filtered to separate it (using a suction pump), then washed with cold methyl alcoholcarbinol before being dried in a vacuum (Scheme 1 ).Colour: Yellow. Yield: 83%. Melting point: 243–245. ESI-Mass spectra (m/z): Calculated: 332 Obtained: 333[M + H]+. Infrared spectra (cm−1, KBr): ν(HC=N) 1649, ν(OH/H2O) 3032, ν(C-O) 1267. Elemental analysis: C16H13ClN2O2S for Calculated % (Found %): C, 57.63; H, 3.854; N, 8.35 (C, 57.74; H, 3.94; N, 8.42). 1H NMR in Fig. 1 (400 MHz, chloroform-d) δ 9.19 (s, 1H, OH), 7.85 (d, J = 8.8 Hz, 1H), 7.60 (d, 1H), 7.52 (s, J = 8.4 Hz, 1H), 7.10 (s, 1H), 7.85 (d, J = 8.2 Hz, 1H), 7.87 (s, J = 8.7 Hz, 1H), 4.12 (q, 2H), 1.42 (T, 3H). 13C NMR in Fig. 2 (400 MHz, CDCl3): δ 14.40(C1), 64.05(C2), 77.01(t, CDCl3), 104.80(C4), 109.12(C13), 116.21(C11), 119.03(C8), 119.84(C7), 123(C15), 124.27(C16), 127.68(C14), 128.73(C5), 129.40 (C6), 137.34 (C3), 156.96 (C12), 160.90 (C10) and 165.86 (C9).The hot methyl alcohol Carbinol (10 ml) solution containing the 2-EBTMCP (0.332 g and 3 mmol)) ligand was slowly added to the methyl alcohol Carbinol (10 ml) solutions of the corresponding metal chloride ions such as CoCl2·4H2O (0.201 g), NiCl2·4H2O (0.201 g), CuCl2·4H2O (0.206 g) and ZnCl2·4H2O (0.208 g). The reaction mass was heated for 3–4 hrs at the same temperature while being refluxed at 70–75 °C. Following the completion of the initial material, solid precipitates of various colours, including Green, Maroon, Blue and Yellow were obtained. These precipitates were then washed in cold methyl alcohol Carbinol (CH3OH), n-hexane and dried in vacuum ( Scheme 2 ). Analytical data: Colour: Light orange. Yield: 66%. Melting point: 266–268. ESI-Mass spectra (m/z): Calculated: 720 Obtained: 722 [M + 2H]+. Infrared spectra (cm−1, KBr): ν(HC=N), ν(OH/H2O), ν(C-O), ν(M−O) and ν(M−N) are shows at 1602, 3107, 1249, 707 and 410 respectively. Elemental analysis: C32H24Cl2CoN4O4S2 for Calculated % (Found %): C, 53.27; H, 3.24; N, 7.66 (C, 53.19; H, 3.35; N, 7.75). Analytical data: Colour: Light orange. Yield: 68%. Melting point: 259–261. ESI-Mass spectra (m/z): Calculated: 721 Obtained: 722 [M + H]+. Infrared spectra (cm−1, KBr): ν(HC=N) 1602, ν(OH/H2O) 3109, ν(C-O) 1251, ν(M−O) 765, ν(M−N) 414. Elemental analysis: C32H24Cl2N4NiO4S2 for Calculated % (Found %): C, 53.14; H, 3.38; N, 7.64 (C, 53.21; H, 3.35; N, 7.76). Analytical data: Colour: Dark green. Yield: 71%. Melting point: 273–275. ESI-Mass spectra (m/z): Calculated: 724 Obtained: 725 [M + H]+. Infrared spectra (cm−1, KBr): ν(HC=N) 1606, ν(OH/H2O) 3197, ν(C-O) 1238, ν(M−O) 767, ν(M−N) 455. Elemental analysis: C32H24Cl2CuN4O4S2 for Calculated % (Found %): C, 52.75; H, 3.38; N, 7.63 (C, 52.86; H, 3.33; N, 7.71). Analytical data: Colour: Dark orange. Yield: 66%. Melting point: 271–272. ESI-Mass spectra (m/z): Calculated: 725 Obtained: 726 [M + H]+. Infrared spectra (cm−1, KBr): ν(HC=N) 1595, ν(OH/H2O) 3182, ν(C-O) 1247, ν(M−O) 748, ν(M−N) 410. Elemental analysis: C32H24Cl2N4O4S2Zn for Calculated % (Found %): C, 52.63; H, 3.24; N, 7.62 (C, 52.72; H, 3.32; N, 7.69).At a temperature of 25 °C, the DNA stock solution had a composition of 5 mM Tris-HCl and 50 mM NaCl in double-distilled water. The solution became clear after being stirred continuously all night. HCl was utilised in order to get the pH of the homogeneous transparent buffer solution up to 7.2. DNA solution quality was examined using the electronic absorbance ratio of 1.8–1.9 at 260 and 280 nm. Ct-DNA has no impurity protein [38]. Using absorption spectroscopy with a molar absorptivity of 6600 M−1 cm−1 and absorbance wavelength of 260 nm, the proportion of Ct-DNA was determined. The generated solutions were stored at a lower temperature (3–4 °C) and utilised for three to four days. The compound was dissolved in a buffer solution consisting of 50% Tris-HCl and 50% acetonitrile throughout the whole experiment. DNA concentrations from 0 to 10 μM were used in the titration experiments, with the complex concentration held constant.The synthesised ligands and their metal complexes were evaluated for cytotoxicity in vitro using the MTT assay. At DMSO, the compounds were dissolved in concentrations ranging from 1 to 100 μM. The cells were seeded into a 96-well plate, and then they were cultured for 48 h at 5% CO2. The cell was then incubated for 24 h after being treated with various concentrations DMSO solutions of the metal complexes. The culture media was removed, and then 15 μL of MTT dye solution was added to each well, before being re-incubated for another 4 h in the dark. The use of MTT was discarded, and dimethyl sulfoxide (DMSO) was used for product solubilisation. With an Elisa reader, check the absorbance at 620 nm in each sample well. The IC50 values were determined by plotting the absorbance of the dosage response curves[39–42]. Table 1 and Fig. 3 displays the chemical and physical characteristics of the ligand (LH) and its metal complexes. The mass spectra match up with what would be predicted for each system. According to the analytical results, the molar ratio of metal to the ligand in all of the complexes is 1:2.The useful information about functional groups is provided by the IR spectral data in Fig. 4 . Table 2 provides a summary of 2-EBTMCP ligand and its metal complexes primary IR characteristic stretching frequencies. The shift in stretching frequency values can be understood by comparing the IR spectra of the free ligand and its metal complexes. The strong and distinct band seen at 1649 cm−1 in the free ligand is due to the azomethine's HC = N stretching vibration. This band is shifted to a lower wave number by 47 cm−1 to 54 cm−1 in metal complexes. These changes provide evidence that the ligand imine nitrogen group participates in the binding of metal ions. Additionally, a prominent peak at 1267 cm−1 in the ligand IR spectra was identified as being associated with the phenolic C-O stretching frequency, which is slightly shifted to a lower wave number in metal complexes. These changes confirm that the phenolic –OH of the ligand participates in the binding of the Co(II), Cu(II), Ni(II), and Zn(II) ions, as well as the elimination of the –OH stretching frequency. Peak in complexes provides further proof that phenolic oxygen participated in metal binding via proton dissociation. Additional novel frequencies are seen in metal complexes in the ranges of 707 cm−1, 767 cm−1 and 410 cm−1, 455 cm−1 respectively, confirming the development of M−O and M−N.To ascertain the crystalline size and structure of the produced Schiff base ligand and metal complexes, powder X-ray diffraction analysis was performed. Fig. 5 shows the X-ray powder patterns for ligand and metal complexes. The semi crystalline nature of the newly synthesised metal complexes is attested to by the sharp and well-defined Bragg Peaks at certain 2θ angles, as determined by this study. The particle size of the complexes was calculated using the Debye-Scherer formula based on the intensity of the highest intensity line in comparison to the other lines. D = 0.94 λ/β cos θ. where, λ is the wavelength of the X-ray used, D is the crystalline size in nm, 0.94 is the sheerer constant, β is the full width at half maximum (fwhm) and θ is the position of the particular diffraction peak. Calculations for the size distribution of ligand and metal complex particles using these diffraction peaks provide values of 19.26 nm for ligand, 24.13 nm for Cu, 29.76 nm for Ni, 31.26 nm for Co, and 41.58 nm for Zn.SEM examination was used to examine the surface morphology of the Schiff base ligand and their metal complexes and energy dispersive X-ray analysis was used to examine the elements of the compounds (EDX). The Schiff base ligand showed sheet like structures observed in Fig. 6 . The Co metal complex shows gross like structure, the Ni metal complex has spherical shape, the Cu metal complex has long flakes, and the Zn metal complex rod shaped particles are observed. The surface morphology of metal complexes differs from their ligands and from one another, as shown by the SEM micrographs, as a result of complexation and metal ion displacement.One of the most well-known techniques for examining the interaction of DNA with metal complexes is UV–vis electronic absorption spectroscopy. In the absence and presence of Ct-DNA, the electronic spectra of all metal complexes Co(II), Ni(II), Cu(II) and Zn(II) were recorded (Fig. 7 ). Due to the strong contact between the aromatic chromophore and the DNA base pairs in the intercalative mode, compound binding through intercalation typically results in hypochromism with or without a slight red or blue shift. The absorbance of the metal complexes and the shift in wavelength due to the increase in DNA concentration has been used to assess their binding capacity[43,44]. A considerable hypochromism is detected when increasing amounts of CT-DNA are added. This is due to a strong association between DNA and complexes, and it is also suggested that these complexes bind to the DNA helix by intercalation. The absorption data were analysed to calculate the intrinsic binding constant (Kb) via the Wolfe-Shimmer equation. [ D N A ] / ( ε a - ε f ) = [ D N A ] / ( ε b - ε f ) + 1 / K b ( ε b - ε f ) Kb is the intrinsic binding constant; [DNA] is the concentration of CT-DNA; εa is the apparent coefficient; εf and εb represent the extinction coefficients for unbound and bound DNA, respectively; Kb can be calculated from a plot of DNA/(εa-εf) vs [DNA] by dividing the slope by the intercept. The binding constants Kb is calculated from spectral data found to be 8.78 × 106 M −1 (Zn), 7.62 × 106 M −1 (Ni), 6.99 × 105 M −1 (Co), and 4.39 × 104 M −1 (Cu). From the above values, it is clear that Zn(II) and Ni(II) complexes bound strongly to DNA than the other metal complexes.Using the MTT assay, the Schiff base ligands and metal complexes were tested for in vitro cytotoxicity against the cell lines HeLa and MCF-7. The Percentage inhibition of cancer cell development was evaluated after treating two cell lines with CT complex in a range of concentrations (12.5–100 μM) for 24 h. The cell viabilities (%) vs concentrations obtained with continuous exposure for 24 h were depicted in Fig. 8 . Cisplatin was used as control. The complexes cytotoxicity was determined to be concentration dependant. The screening results indicated that all of the metal complexes had significant anticancer activity (Table 3 ). The order of IC50 values of the metal complexes against both cell lines as Zn(II) > Ni(II) > Co(II) > Cu(II) > Ligand (Fig. 8).In this study, the coordination capabilities of a new ligand were investigated utilising computational and equilibrium techniques. The complexes of solid metals of the substances involving Co(II), Ni(II), Cu(II) and Zn(II)were developed, then characterised by different spectrum analytical methods FT-IR, ESI-Mass spectra, XRD and SEM-EDX. Co(II), Ni(II), Cu(II) and Zn(II) complexes analytical spectrophotometric investigation revealed that metal and ligand formed stoichiometric in a 1:2 ratio. The DNA binding testing for electronic absorption showed that the intercalation mode was used by all complexes to engage with CT-DNA. Following cytotoxic screening, all metal complexes displayed more potency than that of the corresponding ligand. Zn complex was the most active of all the complexes when compared to other complexes. Kamble Gopichand: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Writing – original draft. Varukolu Mahipal: Writing – review & editing, Resources, Software, Validation, Visualization. N. Nageswara Rao: Methodology. Abdul Majeed Ganai: . P. Venkateswar Rao: Supervision, Conceptualization, 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.GK thanked the Head of the Chemistry Department at Osmania University in Hyderabad for his help with my research.Supplementary data to this article can be found online at https://doi.org/10.1016/j.rechem.2023.100868.The following are the Supplementary data to this article: Supplementary data 1

2-((E)-(6-ethoxybenzo[d]thiazol-2-ylimino)methyl)-4-chlorophenol (2-EBTMCP) of Schiff base (HL) which would be produced from 6-ethoxy-2-amino benzothiazole and 5-Chloro-2-hydroxybenzaldehyde was prepared and characterised using 1H NMR, Infrared, ESI-mass spectra, elemental analysis, SEM, EDX and powder XRD spectroscopic methods. Its complexes with Ni(II), Co(II), Zn(II) and Cu(III) were made, isolated as solid compounds, and identified by various spectroscopic methods. The research of metal complexes here includes the intercalative form of DNA binding studies with CT (Calf thymus) DNA. The UV–Vis electron absorption spectroscopy approach has been used to explore the DNA-binding capabilities of transition metal complexes. Additionally, a MTT assay was performed to investigate their in vitro cytotoxic potential.

Depletion of natural energy resources (crude oil, natural gas, solid fuels) has prompted a great interest in the use of renewable energy sources, mainly biomass [1]. An increasing interest in the biomass conversion is not only because of the energy related reasons but also ecological reasons, mainly the need to restrict the emission of greenhouse gases [2,3]. Biomass is composed of cellulose, hemicellulose and lignin units [4,5]. Because of its chemical decomposition and processing abilities it is a valuable raw material that can be processed into a number of useful products, e.g., substitutes of crude oil fuels [6,7]. This aspect is very important because at present transport uses up to 1/5 of energy on the global scale [8,9].One of the technologies used for transformation of biomass is fast pyrolysis, which is thermal decomposition of high-molecular chemical compounds to simpler components under anaerobic conditions [10,11]. At these conditions, the building blocks of biomass undergo degradation leading to the formation of a liquid fraction known as biooil or pyrolytic oil [12]. It is a mixture of over 300 different chemical compounds containing a significant number of oxygen atoms [13], mainly in the form of oxygen lignin derivatives, aldehydes (including heterocyclic furfurals), alcohols, phenols, ketones, carboxylic acids or carbohydrates [6,14]. Biooil has a high energy potential as it maintains up to 70 % of the initial biomass energy [5]. However, it cannot be applied directly as a fuel because of a too high content of water (up to 25 wt.%) and oxygen (up to 40 wt.%), which finally results in a low calorific value [5,9]. In order to make its properties more resembling those of crude oil, biooil must be subjected to physicochemical processing [15].A promising pathway to get high quality hydrocarbon fuel from biooil is the catalytic hydrodeoxygenation (HDO) [16–18]. This process is based on the reaction of chemical compounds contained in biooil with gaseous hydrogen, under elevated pressure, elevated temperature and in the presence of a catalyst. Under these conditions the reagents are transformed to simpler compounds, mainly hydrocarbons [18,19]. This process permits a removal of oxygen atoms from chemical compounds of biooil but also leads to the saturation of multiple bonds of the reagents, which improves the system stability. Finally, the process causes a decrease in the O/C ratio at the simultaneous increase in the H/C ratio, which means that the improved biooil, called a biofuel, can be used as a motor vehicle fuel [9]. The HDO process is usually carried out at a temperature between 200 and 500 °C under hydrogen at a pressure reaching 200 bar [20]. Under such conditions, the majority of chemical compounds are able to convert into their deoxygenated analogues [6]. In the laboratory scale the mechanism of HDO is studied using the so-called model chemical compounds that are identical to those contained in biooil [21]. Very often anisole is selected for investigation because it contains an isolated methoxyl group (OCH3) which is the most often present in the chemical structure of biooil [22–24].High yield of HDO process depends on the catalyst which should by characterized by high activity even at lower pressure and temperature [21,25]. The most effective catalysts for this process are those with transition metal ions (Pt, Pd, Ru, Rh, Ir, Co, Ni) as the active components because they show high activity in the processes involving hydrogen [26,27]. Because of their high cost [28], they are usually deposited on supports in order to enlarge the surface area of the active phase and improve the mechanical strength and thermal stability of the system [29]. A particularly interesting group of supports are ordered mesoporous silicas [30–32], characterized by well-developed system of pores with diameters between 2 and 50 nm, high pore volume (∼0.7 cm3/g) and large surface area reaching even 1000 m2/g [29,33]. Very attractive silicas are those of SBA (Santa Barbara Amorphous) type materials because they are non-toxic, possess ordered mesoporous structure with interconnected micropores in the mesopore walls and can be easily synthesized by block copolymer soft templating [30,34].The best-known representative of the SBA class of materials is SBA-15 silica with two-dimensionally ordered mesopores. Below are some examples from the literature where this material was used as catalyst in the HDO reaction.A very interesting silica from the SBA group is SBA-12, which has a hexagonal distribution of channels with the p63/mmc space group. The specific surface area of this silica often exceeds 1000 m2/g [35]. The pores of the SBA-12 material have a three-dimensional architecture that essentially facilitates the transport of larger reagent molecules, which in turn provides easier access to active sites. The material in question has not yet attracted much attention, although there is a good chance that it will because, like the SBA-15 silica, it has thick pore walls, thanks to which it is thermally and hydrothermally stable.Catalytic systems based on mesoporous silicas have been explored by many authors in hydrodeoxygenation of anisole. Yang et al. [36] have studied the mechanism of HDO of anisole catalyzed by a catalyst having nickel atoms on SBA-15 type mesoporous silica. The process was performed at 310 °C under hydrogen pressure of 30 bar for 6 h. The products formed in large amounts at 100 % anisole conversion were hexane (selectivity 26 %), cyclohexane (30 %) and benzene (26 %). Results of this experiment proved high activity of the catalyst used because it was possible to obtain a high degree of deoxygenation and anisole hydrogenation.Similar studies have been carried out by Sankaranarayanan et al. [5] who investigated the catalytic activity of ordered mesoporous silica, specifically SBA-15 with deposited cobalt atoms as the active phase. The process was performed at 220 °C under hydrogen pressure of 50 bar for 2 h. These authors achieved 99 % anisole conversion to the following products: methoxycyclohexane (selectivity 72 %), cyclohexane (12 %), cyclohexanone (1 %) and cyclohexene (0.5 %). The catalytic system was also very active, although the highest selectivity was achieved for methoxycyclohexane, the compound with single bonds between carbon atoms but still having oxygen in the structure. Although the results of hydrodeoxygenation of chemical compounds of biooil are good, the search for catalytic systems capable of providing high yields but at possibly lower temperatures and pressures is continued [26]. According to literature, the HDO process is the most effective at elevated temperatures and pressures, but at these conditions it is an energy consuming process, uneconomical from industrial viewpoint [37].Previously only SBA-15 among the SBA-family materials was used as a support of the HDO catalyst. Here we report data for hydrodeoxygenation of anisole over the ruthenium catalyst deposited on SBA-12 silica at relatively low temperature (90−130 °C) and under low pressure of gas hydrogen (25−60 bar). We studied the effect of the reaction time, the amount of catalyst and the active phase loading on the conversion process. The impact of these parameters on the degree of anisole conversion, types and amounts of the reaction products is discussed. It is the first attempt of using SBA-12 silica with transition metal catalysts in the HDO conversion. Such ordered mesoporous silica with symmetry group p63/mmc can be of consideration when outstanding stability and three-dimensional mesopore structure is desired for the proper adsorption/diffusion necessary in heterogeneous catalysis. A comparison with other catalysts is assessed.SBA-12 silica was synthesized by adding structure-directing agent BRIJ 76 surfactant (Sigma Aldrich; 8 g) to a mixture of water (40 g) and 0.1 M hydrochloric acid (160 g) and stirring for 2 h. Then, 17.6 g of tetraethyl orthosilicate (TEOS, Fluka, 99 % purity) was added and the mixture was stirred for 20 h. After this time, the mixture was poured to PP bottles and placed in a drying apparatus at 80 °C for 24 h. The solid product was filtered, dried and calcined in order to remove the template at 550 °C for 6 h. The resulting silica was subjected to wet impregnation in order to introduce different amounts of ruthenium to obtain catalysts with different amount (wt.%) of Ru with respect to the mass of the support (the incremental dosage was 0.5 wt.% up to maximum 3 wt.%, i.e., catalysts with 0.5; 1; 1.5; 2; 2.5 and 3 wt.% were prepared). Ethanol solution of ruthenium(III) chloride was introduced dropwise to the dry SBA-12 support in the amount sufficient to fill the pores and wet the external surface of particles. The content of the beaker was stirred. The beaker was tightly closed with a polyolefin-paraffin foil Parafilm® and left to rest for 24 h to allow penetration of the solution into the pores of SBA-12. Afterwards, the foil was removed and the beaker was placed under fume hood at room temperature until the entire solvent was evaporated under stirring the content from time to time. Next, the catalysts were dried for 1 h at 30 °C, 1 h at 40 °C, and 18 h at 60 °C, followed by reduction in a tube furnace in hydrogen atmosphere. To remove air from the system argon of N5.0 purity (Linde Gas Poland) (50 cm3/min) was blown to the furnace, then hydrogen (50 cm3/min) was introduced and after 0.5 h the furnace was switched into the heating mode. The catalysts were reduced at 250 °C for 3 h (the temperature was established based on the TPR results). The catalysts studied were labelled as y RuSBA-12, where y is the wt.% content of ruthenium in relation to the support.The process of anisole hydrodeoxygenation was carried out in a high-pressure reactor (CAT 24 HEL) placed on a magnetic stirrer and equipped with a thermocouple and manometer to measure the initial pressure in the reactor. The catalysts dispersed on SBA-12 silica support and containing 1 or 3 wt.% ruthenium (labelled as 1% RuSBA-12, 3% RuSBA-12) with respect to the support, were placed in a furnace at 250 °C for 3.5 h under dry argon flow prior to the hydrodeoxygenation in order to remove traces of water. An appropriate amount of the catalyst (0.025, 0.05 or 0.1 g) together with anisole (1 g) was placed in a glass vessel equipped with a magnetic stirrer. No solvent was added as anisole was in the liquid state. The glass vessel with the reaction mixture was placed in a high-pressure reactor, which was tightly closed and flushed with argon of N5.0 purity in order to remove air from the reactor. Then hydrogen of N5.0 purity was introduced to the reactor and then the reactor was filled with hydrogen up to the pressure of 25, 40 or 60 bar. The reaction was conducted at three temperatures 90, 110 or 130 °C, for the time of 1, 2.5 or 4 h. The reaction mixture was stirred at 700 rotations per minute. After the reaction time, the reactor was cooled to room temperature. The reaction mixture was centrifuged (VWR) in order to separate the liquid reaction products from the solid catalyst.The reaction substrates and products were analyzed using a gas chromatograph made by VARIAN 3900 GC equipped with a capillary column CPWAX57CB (length 25 m, diameter 0.32 mm, film thickness 1.2 μm) and a flame-ionization detector. The products were identified on the basis of comparison of retention times of the compounds obtained with those of the standards. The results were confirmed by the results of gas chromatography analysis using a chromatograph coupled with a mass detector made by VARIAN 4000 GC and a capillary column VF-5MS (length 30 m, diameter 0.25 mm, film thickness 0.25 μm).The Ru content in all synthesized catalysts was evaluated by inductively coupled plasma (ICP) analysis. In addition, these catalysts were characterized by X-ray diffraction (XRD), transmission electron microscopy (TEM), low-temperature nitrogen adsorption/desorption (N2 ads/des). The XRD diffractograms were recorded on a diffractometer Bruker AXS D8 Advance, using CuK α (0.154 nm) radiation, in the range 2Ɵ = 0.6–10° with 0.02° step (small-angle range) and 2Ɵ = 4–60° with 0.05° step (wide-angle range). TEM images were recorded on a JEOL-2000 TEM microscope at 80 kV.Low-temperature nitrogen adsorption/desorption isotherms were recorded on Nova Quantachrome instrument at 176 °C, after prior degassing of the sample in vacuum at 350 °C for 24 h. The surface area of the catalysts was determined by the BET (Brunauer – Emmet – Teller) method. The pore volume was determined using the KJS-BJH method based on the BJH (Barret – Joyner – Halenda) algorithm [38,39].All the materials studied were characterized by using some of the techniques described in chapter 3. This allowed selection two materials for detailed characterization. Additionally, due to the lack of space, only data for 1 and 3 wt.% are discussed here. Fig. 1 presents the diffractograms recorded for pristine SBA-12 and SBA-12 support with deposited ruthenium, in the small-angle range (Fig. 1a) and wide-angle range (Fig. 1b). The small-angle diffractograms show one main reflection at 2Ɵ = 1.8° corresponding to the plane [002] and a lower-intensity reflection at 2Ɵ = 1.27°, corresponding to the plane [100]. The material of SBA-12 type structure belongs to the space group p63/mmc and has hexagonal symmetry [40,41] so the elementary cell parameters were calculated for hexagonal symmetry.The a and c lengths of the elementary cell for hexagonal RuSBA-12 sample vary in the range 8.0–9.6 nm, as shown in Table 1 . The unit cell parameters of the SBA-12 (a = 8.0–8.4 nm, c = 8.9–9.1 nm, c/a = 1.63) are in good agreement with previously reported results [42,43]. After introduction of ruthenium in the amounts smaller than 1 wt.% the values of a and c lengths decrease and then increase when the content of ruthenium increases from 1 to 3 wt.%.The nitrogen adsorption/desorption isotherms recorded for samples 1% RuSBA-12 and 3% RuSBA-12 (Fig. 2 ) are of type IVa, characteristic of mesoporous materials. Also, the type of hysteresis loop is typical for mesoporous materials [40,44,45]. As was mentioned above, the isotherms obtained for the SBA-12 samples studied match the patterns reported for other mesoporous materials with p63/mmc symmetry [40]. The additional hysteresis loop recorded for samples with 1 wt.% of Ru can be interpreted as a result of the presence of voids between particles [42,46]. Surface area of the samples is close to 1000 m2/g, while the average pore size is 5 nm, as shown in Table 1.The TEM images presented in Fig. 3 , confirm the ordered mesoporous structure of the materials studied. They have hexagonal 3D structure with densely packed mesopores (hcp) [47]. No crystalline ruthenium oxide species were observed (wide-angle XRD, TEM), which can suggest that agglomerates at the outer surface were not formed. Thus, Ru probably exists in the catalyst in the form of isolated Ru atoms The results of SEM/EDX analysis (not shown here) confirm homogeneous distribution of ruthenium species.Hydrodeoxygenation of organic compounds is the process of obtaining hydrocarbons in the presence of gaseous hydrogen supplied under elevated pressure that takes place at elevated temperature and in the presence of a catalyst, as shown in Fig. 4 .In this study a series of hydrodeoxygenation reactions of anisole were performed using SBA type silica loaded with different amounts of ruthenium catalyst. The reaction does not proceed without a catalyst and proceeds with low conversion with unmodified SBA-12 catalyst (Table 2 , entry 19). The effects of the reaction time, the amount of catalyst, hydrogen pressure and temperature were evaluated. According to literature data, the sequence of transformations of anisole molecules under elevated pressure and temperature includes a few reactions such as hydrogenation, dehydration, demethoxylation and demethylation [12,48]. The type and amounts of products of HDO of anisole depend on many parameters, including the type of catalyst and conditions of the process.The main products of HDO of anisole identified by GC or GC–MS, irrespective of the conditions, are methoxycyclohexane and 1-methoxycyclohexene, as shown in Figs. 6–9. Besides the above main products, the formation of cyclohexane, cyclohexene, benzene, 1,1’-dimethoxycyclohexane, cyclohexanone, cyclohexanol, was confirmed. The selectivities to these products are listed in Table 2. However, these products are not presented on the plots as their total content did not exceed 20 %. Based on the identified products of the reaction, three pathways of HDO of anisole to a fully hydrodeoxygenated compound - cyclohexane - are proposed (Fig. 5 ). The first pathway involves hydrogenation of the aromatic ring, leading to 1-methoxycyclohexene and subsequently to methoxycyclohexane, then the CO bond between the carbon from the aromatic ring and the oxygen from methoxyl group is broken, which leads to cyclohexane formation. The formation of undesirable product of 1,1’-dimethoxycyclohexane (that contains more oxygen atoms than the initial compound) as a result of methanol attachment to 1-methoxycyclohexene is also possible.In the second pathway, cyclohexane is obtained via an intermediate product benzene. This pathway starts with direct demethoxylation of anisole to the aromatic structure and then the aromatic ring is hydrogenated.In the third pathway, the first reaction is demethylation of the substrate, leading to the formation of phenol, which is converted to cyclohexanol via total hydrogenation of the aromatic ring. Cyclohexanol can undergo isomerization to cyclohexanone, dehydration of which gives cyclohexene. Cyclohexene having a double bond in its structure can easily undergo hydrogenation to cyclohexane. Because of possible cyclohexanol dehydration, the reaction can also give dicyclohexyl ether.Analysis of Figs. 6–9 and Table 2 data suggests that in this study the main pathway seems to be the first one, involving direct hydrogenation of aromatic ring: anisole → 1-methoxycyclohexene → methoxycyclohexane → cyclohexane. The domination of this pathway may be related to the character of transition metal used as a catalyst that prefers hydrogenation of aromatic ring [37] and to the fact that the reactions were run at low temperatures and low pressures, at which the energy needed for hydrogenation of aromatic ring is lower than that needed to break the bond between the carbon atom from the aromatic ring and the methoxyl group or the oxygen atom and carbon from the methyl group [5]. A general conclusion is that in the HDO reaction of anisole catalyzed by SBA-12 silica with deposited ruthenium atoms the hydrogenation of aromatic ring is preferred over demethylation and demethoxylation. The product forming in the largest amount is methoxycyclohexane with higher H/C ratio to be achieved in the process. The teams of Shi [49], Khromov [50] and Sankaranarayanan [5] studied also HDO of anisole and reported the highest selectivity to methoxycyclohexane.The plots illustrating the influence of the reaction time, the amount of catalyst, catalyst loading, pressure and temperature of the reaction are shown below.Preliminary studies of hydrodeoxygenation of anisole were carried out to establish the optimum time of the process. The reaction was carried out for 1, 2.5 or 4 h at 110 °C, under hydrogen pressure of 60 bar and in the presence of 0.05 g SBA-12 containing 1 or 3 wt.% of ruthenium. Decomposition of the main products of the reaction and the degree of anisole conversion as a function of time are presented in Fig. 6, while the selectivities to the side products are given in Table 2 (entries no. 1, 3, 5 for the catalyst 1 wt.% RuSBA-12, and 2, 4, 6 for the catalyst 3 wt.% RuSBA-12). The results unambiguously show that the quantitative composition of the products and anisole conversion depend significantly on the time of the HDO process. In the experiment with 1% RuSBA-12 catalyst the time of reaction had a great effect on the anisole conversion, which increased from 24 % after 1 h to 93 % after 2.5 h, to 100 % after 4 h (Fig. 6a). Moreover, the catalyst became increasingly selective towards methoxycyclohexane with time of the reaction because the amount of this compound increased. A similar correlation was observed for the process with 3% RuSBA-12, for which after 4 h of the process the selectivity towards methoxycyclohexane increased to 96 % (Fig. 6b). Using the 3% RuSBA-12 catalyst, the degree of anisole conversion was smaller than in the presence of 1% RuSBA-12. The conversion of anisole was already high - 93 % after 1 h of the process, then after 2.5 h it increased to 95 % and after 4 h it was 100 %. Greater activity of 3% RuSBA-15 catalyst after 1 h of the process was a direct consequence of a greater content of metal with respect to the substrate. Based on the results of these experiments, we decided to carry out further experiments for 4 h.At the next stage, our aim was to establish the optimum amount of the catalyst used in the process. The HDO of anisole was performed at 110 °C under a hydrogen pressure of 60 bar for 4 h, in the presence of 0.025, 0.05 or 0.1 g of samples 1% RuSBA-12 or 3% RuSBA-12. Analysis of HDO of anisole in the presence of sample 1% RuSBA-12 (Fig. 7a) (Table 2, items 5, 7, 9) reveals that the greatest catalytic activity was obtained for the catalyst amount of 0.05 g. The use of catalyst in this amount gave the maximum conversion of anisole to 94 % of methoxycyclohexane. The reaction in the presence of 0.025 or 0.1 g of the catalyst led to the formation of noticeable amounts of 1-methoxycyclohexene – a compound of a lower content of hydrogen than methoxycyclohexane – and the degree of anisole conversion was lower than when an intermediate amount of catalyst was used, i.e., 0.05 g. When 0.1 g catalyst was used in the reaction, a decrease in the anisole conversion and a decrease in the selectivity to methoxycyclohexane were observed compared to the reaction catalyzed by 0.05 g of silica. A decrease in the catalyst activity can be attributed to the physico-chemical properties of the catalyst, and of course the reaction conditions according the Sabatier principle stating that the bond strength between reactants and catalyst should be intermediate (i.e., not too weak, so that the reactants adsorb on the catalyst surface but not too strong to avoid, its poisoning).For the processes in the presence of 3% RuSBA-12 (Fig. 7b) (Table 2, entries no. 6, 8, 10) no significant effect of the amount of the catalyst was observed as in all reactions the selectivity to methoxycyclohexane was greater than 90 % and the anisole conversion was maximum. Based on these results, we decided to use in the further study the catalyst in the amount of 0.05 g.The studies aimed at establishment of the optimum reaction time and optimum amount of the catalyst also provided information on the activity of the catalysts used, 1% RuSBA-12 or 3% RuSBA-12, differing in ruthenium loading. The more selective catalyst with higher anisole conversion was 3% RuSBA-12. Its higher activity can be directly related to the greater content of metal on the support surface. Therefore, further studies on the effect of pressure and temperature on HDO of anisole were performed with the catalyst 3% RuSBA-12.The influence of hydrogen pressure on the distribution of the reaction products and the degree of anisole conversion was studied for three pressure values, 25, 40 or 60 bar at three temperatures 90 °C (Fig. 8a), 110 °C (Fig. 8b) or 130 °C (Fig. 8c). Table 2 gives the amounts of products of HDO of anisole besides the main ones (entries no. 6, 11–18). The reaction was performed for 4 h using 0.05 g of 3% RuSBA-12 catalyst. The results imply a significant effect of hydrogen pressure on the character of HDO of anisole. For each temperature, the conversion of anisole increased with increasing hydrogen pressure. Under the hydrogen pressure of 60 bar the aromatic ring hydration occurred in the highest degree. It resulted in the formation of a large amount of methoxycyclohexane and small amount of 1-methoxycyclohexene, which is particularly evident for the processes carried out at 110 and 130 °C. For the reactions performed at 90 °C, the change in hydrogen pressure had practically no effect on the distribution of the main reaction products.The influence of temperature on HDO of anisole on its conversion and selectivity of the reaction towards individual reaction products is illustrated in Fig. 9a–c. The experiments were performed at 90, 110 or 130 °C under hydrogen pressures of 25 bar (Fig. 9a), 40 bar (Fig. 9b) or 60 bar (Fig. 9c). Table 2 presents the amounts of the reaction products other than the main ones (entries no. 6, 11–18). These data show that the reaction temperature, similarly as pressure, has a significant influence on the character of the process. At each pressure an increase in temperature caused greater conversion of anisole. However, temperature changes had no effect on the amounts of the particular reaction products. The catalyst became more selective with increasing temperature, which is well visible for the processes at 130 °C.As can be seen from Fig. 9a–c, anisole conversion at 90 °C is very low and the main products are 1-methoxycyclohexene and methoxycyclohexane. At higher temperature, 110−130 °C, the amount of generated cyclohexane increases (Table 2), while the amount of 1-methoxycyclohexene decreases (Fig. 9a–c). In other words, the Caryl O bonds cleave and anisole is converted preferentially to cyclohexane at higher temperature, which indicates the occurrence of the reaction according to the pathways I and II in Fig. 5.The main reaction product in our experiment was methoxycyclohexane - chemical compound with a higher H/C atom ratio than the starting substrate. According to the literature, methoxycyclohexane is very often obtained in this reaction regardless of the type of catalyst used. In accordance with the data collected in the Table 3 (examples 2 and 6), scientists also reported methoxycyclohexane as the main product, however the reaction temperature used was much higher. Moreover, in our experiments, methoxycyclohexane was obtained with higher selectivity and higher conversion of anisole. Some research groups have reported deoxygenated chemicals (Table 3, examples 1, 3, 4, 7, 8), however, several factors contributed to this, such as harsher process conditions and the presence of acid centers. The table below collects more examples from the literature, where anisole was subjected to the hydrodeoxygenation reaction.The reaction of hydrodeoxygenation of chemical compounds obtained from biomass degradation is generally considered as a new and attractive method for production of biofuels. Further research needs to be conducted to optimize the process and make it energetically efficient. Studies of hydrodeoxygenation of anisole, chosen as a representative of organic compounds present in biooil, catalyzed by RuSBA-12 at varying pressures (25−60 bar) and temperatures (90−130 °C), permitted the following conclusions: (I) ordered mesoporous silica such as SBA-12 with ruthenium deposited on their surface are highly active in HDO of anisole, (II) larger content of ruthenium on the support surface is directly related to a higher conversion of anisole, (III) the product obtained in the highest amount, irrespectively of the reaction conditions, was methoxycyclohexane, (IV) the pressure and temperature of the process have significant effect on the degree of anisole conversion and the selectivities to the reaction products. The yield of the process increased with increasing pressure and temperature.Briefly, Ru catalyst catalyzed aromatic CO bonds to produce aromatic hydrocarbons due to its oxophilicity. The strong metal oxophilicity of Ru favors the direct Caryl O bond cleavage to benzene, while the weak oxophilic catalyst (e.g. Pt/SiO2) favors the aliphatic Calkyl O bond breaking to phenol. Both demethylation and demethoxylation occur over the moderately oxophilic Ru catalysts.To Maria, a very outstanding scientist, sincere thanks for all she has done and will continue to do in the future in the field of heterogeneous catalysis. Thank you, for being always ready to share your knowledge.The authors wish to thank the National Centre for Science for financial support of the studies reported within the research project HARMONIA-5 (no. DEC-2013/10/M/ST5/00652).

Hexagonally ordered mesoporous silica SBA-12 catalysts containing various amounts of Ru (1 or 3 wt.%) were obtained by wet impregnation. These catalysts were thoroughly characterized by X-ray diffraction (XRD), transmission electron microscopy (TEM), and low-temperature nitrogen adsorption/desorption (N2 ads/des). Anisole conversion was measured over catalysts at different process conditions, where the process temperature was 90−130 °C and the hydrogen pressure – 25−60 bar. Prior the experiments the process was optimized, i.e., the amount of catalyst used in the reaction and the time of reaction were adjusted. This study shows a significant effect of hydrogen pressure and process temperature on the hydrodeoxygenation of anisole, the conversion of which increased (16–100 %) with both increasing parameters. For all the catalysts studied, the highest selectivity was obtained for two main reaction products, methoxycyclohexane and 1-methoxycyclohexene. Along with increasing conversion of anisole, the selectivity to the main reaction product - methoxycyclohexane generally increased (65–96 %) reducing the amount of 1-methoxycyclohexene (0–27 %).

Diesel engines commonly have higher thermal efficiency than gasoline engines due to their higher compression ratio, without throttle loss and low pump gas loss characteristics. Additionally, since diesel engines have low-speed torque characteristic advantages, they are widely used in non-road mobile machinery such as engineering machinery, agricultural machinery, generating set, ship and train moving mechanical equipment (Pirjola et al., 2017; Hu et al., 2021). It is well known that increasing the compression ratio raises the combustion temperature, which eventually results in serious emissions (Mohiuddin et al., 2021; Saxena and Maurya, 2017). It's much worth noting that non-road mobile machinery has a large amount of ownership (Feng et al., 2022), covers a wide range of industries, the working conditions are complex and harsh environment, and there are mobile operation and trans-regional operation characteristics, all of which will considerably exacerbate the emissions. The main pollutants discharged include CO, NOX, solid PM, and HC, which will cause serious air pollution (Kumar et al., 2021). The emissions will pose a major threat to human health and the climatic environment. Moreover, the emissions will induce extremely significant ailments in individuals, such as immune system destruction, influence the blood supply and impair the respiratory tract (Margaryan, 2021; Marco et al., 2019; Dong et al., 2021). World Health Organization (WHO) shows that air pollution kills an estimated seven million people worldwide every year and 9 out of 10 people breathe air that exceeds WHO guideline limits containing high levels of pollutants (World Health Organization, 2020). Carbon dioxide (CO2) was not considered as a pollutant in the past, but the excessive use of carbon-containing fossil fuels has unbalanced the earth's carbon cycle and intensified the “greenhouse effect”. So CO2 emissions have also drawn much global attention. For example, the United States (US) Energy Information Administration (EIA) estimates that diesel fuel consumption resulted in 456 million metric tons of CO2 emissions in 2019 (Energy Information Administration, 2021). This amount was equal to about 24% of total US transportation sector CO2 emissions and nearly 9% of total US energy-related CO2 emissions in 2019, and diesel-fueled machinery is a major source of harmful pollutants. Similarly, according to 2014 statistics from the US Environmental Protection Agency (EPA) (United States Environmental Protection Agency, 2021), non-road diesel machinery in the US contributes over 35% of NOX and 44% of PM emissions from mobile sources. According to the European Environment Agency (EEA) (European Environment Agency, 2021), non-road PM emissions account for 17% of total emissions. In China, the number of non-road diesel engines is far less than on-road engines. The emission standards of road diesel and gasoline engines are much higher than those of non-road diesel engines. Non-road pollution sources have gradually become an important source of air pollution and have become more and more serious (China Mobile Source Environmental Management Annual Report (2020), 2021). According to the Ministry of Ecology and Environment released “China Mobile Source Environmental Management Annual Report (2020)” (Ministry of Ecology and Environment, 2020), 2019 in China, the HC, NOX, and PM emissions of non-road diesel engines were 43.50 million tons, 4.93 million tons and 240 thousand tons, respectively. The NOX emissions of non-road diesel engines were the same as those of on-road motor vehicles. As a result, non-road pollutant emissions have become a major global problem.More and more countries in the world attach importance to pollution and formulate more stringent emission regulations for non-road diesel engines (Olabi et al., 2020; Ni et al., 2020). The European Commission Stage V and the US EPA Tier 4 Final standards have the same emissions limits (European Commission, 2016). China has formulated the national standard GB 20891-2014 “Limits and Measurement Methods for Exhaust Pollutants from Diesel Engines of non-road Mobile Machinery (CHINA III, IV)” for non-road diesel engines (GB 20891-2014, 2014). And the China IV emission standards will be implemented after December 1, 2022.In order to cope with the environmental pollution pressure and increasingly strict emission regulations, various energy conservation and emission reduction technologies are inevitably applied to non-road diesel engines (Venu et al., 2019; Hamedi et al., 2021; Datta and Mandal, 2016). In recent years, non-road diesel engine emissions have been alleviated to a certain extent through the research of experts and scholars (Boccardo et al., 2019; Duraisamy et al., 2019; Frosina et al., 2015; Ganesh et al., 2019). Like the road diesel engine, the exhaust emission reduction technologies of non-road diesel engines include internal purification and external purification technology. The internal purification technology mainly improves the engine combustion efficiency to reduce the harmful gas composition in the tail gas (Wu et al., 2021; Agrawal et al., 2019; You et al., 2020). At present, the main internal purification technologies include intake management technology, combustion technology (Wang et al., 2020), electronic controlled combustion injection technology, multi-valve technology and exhaust gas recirculation (EGR) (Wang et al., 2019). In recent years, dual-fuel diesel engines have made great progress. Ning et al. (2020) showed that the addition of primary alcohol fuels in dual-fuel mode reduces CO and soot emissions, but the total hydrocarbon (THC) and NOX emissions increase (Lee et al., 2020). Nag et al. (2022) investigated the co-combustion characteristics of hydrogen (H2)-diesel dual-fuel with EGR, and found that the synergistic action of H2 and EGR has a higher emission reduction potential, which can reduce NOX by more than 38% (Nag et al., 2019). However, adding H2 to a diesel engine will also face overwhelming challenges. For example, Kumar et al. (2021) found that adding H2 to the diesel engine was conducive to reducing carbon-based emissions, while Karthic et al. (2020) reported that the engine is prone to knocking at a higher H2 level. Kim et al. (2021) found that natural gas substitution significantly reduces PM and NOX, but at the same time increases CO and unburned hydrocarbon (UHC) emissions. As we all know, EGR can effectively reduce NOX emissions (Ayhan et al., 2020), but engine combustion is easy to deteriorate, which will increase HC, CO, and soot emissions (Pradelle et al., 2019). Zhang et al. (2021) found that EGR can reduce oxygen (O2) and hydroxide radical (OH) concentration and flame temperature, thus reducing O2 and OH oxidation rate on the soot surface, so it will increase soot emissions, especially when the EGR rate is high. Recently, homogeneous charge compression ignition (HCCI) and reactivity controlled compression ignition (RCCI) had been introduced into non-road diesel engines due to the advantages of higher thermal efficiency, lower NOX and lower PM (Duan et al., 2021; Reitz and Duraisamy, 2015).Although internal purification technologies can effectively reduce the emissions of non-road diesel engines to a certain extent, the exhaust aftertreatment technology is widely used in non-road diesel engines because the internal purification technologies cannot meet the requirements of relevant emission regulations. The exhaust aftertreatments mainly include SCR (Jiang and Li, 2016), DOC (Tan et al., 2019), DPF (Kang et al., 2018; Zhang et al., 2020), POC and ASC. NOX emissions can be significantly removed by SCR (Lauren et al., 2020), while DOC is commonly used to decrease CO and HC emissions as well as small amounts of particles. PM is a very harmful exhaust emission and most of it can be purified by DPF or POC. The DPF is a wall-flow filter that forces airflow through the porous wall to capture particles by alternately blocking the inlet and outlet of the current-carrying hole so that a large number of particles can accumulate during the operation of the trap. Note that the PM purification efficiency of DPF can reach 90% or higher. However, a large amount of PM accumulation will block the passage and cause high back pressure in the exhaust pipe (Fang et al., 2019), finally resulting in lower engine efficiency and increased emissions. Therefore, the DPF needs continuous regeneration (Jang et al., 2017). DPF regeneration requires more fuel or electricity consumption, so diesel engines equipped with DPF are less economical. In addition, the fuel also has high requirements, because DPF is very sensitive to the sulfur in the fuel and prone to form sulfuric acid at high temperatures, high sulfur content is also easy to block DPF (Rounce et al., 2019). Unlike DPF, POC has no risk of clogging because it mainly uses a cellular channel structure with multiple folds, so there is no blockage in and out of the carrier hole, and most PM can be captured (Yao et al., 2011; Liu et al., 2012). What is more, POC has the advantages of lower back pressure and lower cost compared to DPF (Lehtoranta et al., 2007). However, POC capture efficiency of PM in emissions is only 40–70% and it is mainly influenced by soot loading density, the structure of POC, PM to nitrogen dioxide (NO2) ratio and exhaust temperature (Guan et al., 2016; Zhan et al., 2012).In recent years, DPF and POC have been widely used in diesel engines, usually combined with other aftertreatments. The general aftertreatment (DOC+DPF+SCR) technical line can effectively reduce the emissions of PM and NOX (Jung et al., 2019 ; Ko et al., 2019). Guan et al. (2016) demonstrated that diesel engine emissions can be significantly reduced by using POC under different fuel injection strategies. The optimized injection strategy can further improve the POC removal rate. Rounce et al. (2019) found that using a catalytic partial flow filter as a POC, combining oxidation and filtration functions in a single unit, helped achieve greater catalytic pollutant removal capacity and significant PM filtration efficiency, reducing the pressure increase across the DPF. Feng et al. (2014) showed that POC significantly increased the NO2/NOX ratio. It can be increased by 4.5 times on average in all modes. The average reduction rate of particle number (PN), which is another display method of PM, was 61%. When the engine speed was set at 1400 rpm, the PN decreased with the increasing particle size. Liu et al. (2012) reported that POC can remove organic components from the total PM, but can only partially remove carbonaceous particles with the particle size less than 30 nm due to the honeycomb structure of POC and insufficient time to oxidize or capture solid particles. Geng et al. used a double diesel oxidation catalyst (DDOC) and a DOC closely coupled with a POC in series (DPOC) to research the PM. They found that after DDOC and DPOC treatment, when the exhaust gas temperature is high enough, the quantity and mass concentration of particles, especially nuclear particles, can be significantly reduced. However, the filtering effect of POC on PM is much lower than that of DPF (Geng et al., 2015). Rossomando et al. have shown that the removal efficiency of DPF particles below 23 nm was higher than 90%, and the highest efficiency was 99% in the range of 20–40 nm. Due to the high removal efficiency of DPF within the study range, the engine operating conditions had little effect on particle emissions (Rossomando et al., 2021). Ko et al. found that above 300 °C, nitric oxide (NO) was converted to NO2 by Pt catalytic oxidation reaction. The highest conversion rates appeared at 450 °C for DOC and 350 °C for DPF (Ko et al., 2019). Chen et al. (2020) demonstrated that at high DOC inlet temperatures (> 190 °C), the diesel methanol dual fuel (DMDF) model combined with DOC achieved higher DPF inlet temperature and higher NO2/PM ratio, which were conducive to passive regeneration. The evolutionary mechanism of the soot layer has an important effect on the regeneration process of DPF (Ou et al., 2021). Temperature also has an important effect on DPF regeneration (Meng et al., 2020) For instance, the soot load of 8 g/L and the regeneration temperature of 575 °C resulted in a large total emission of particulate matter downstream of DPF. When the regeneration temperature is 550 °C, the regeneration efficiency of 4 g/L and 8 g/L DPFs is moderate, and the total emission of particles is relatively low.From the above discussion, although lots of work has been done in both in-cylinder purification and aftertreatment technology, most of the studies on the emission characteristics of non-road diesel engines mainly focus on steady state conditions and separately study the emission reduction effects of DPF or POC, and most of the researches have been done on the microstructure and regeneration methods of the aftertreatment processors. However, there are few studies on coupling several aftertreatment processors together, and the coupled aftertreatment processors mainly focus on the emission of the vehicle or engine under steady-state working conditions, compared with transient working conditions. It is necessary to study the emissions of aftertreatment processors under transient conditions. What's more, most of the research on this problem focuses on on-road diesel engines. Although the proportion of non-road diesel engines is small, the emissions from these engines are very serious. Accordingly, the effects of the DOC + DPF/POC + SCR + ASC integrated after-treatment system on non-road diesel emissions are contrastively studied under NRSC and NRTC simultaneously in this paper. And the difference between POC and DPF is fully compared and analyzed. The main goal of the current study is to explore the possibility and implementation of POC to replace DPF and provide valuable guidance for the emission reduction of non-road diesel engines.The experiments were performed on an in-line six-cylinder, four-stroke, turbocharging, water-cooled, non-road diesel engine, and the main parameters are listed in Table 1. The Chinese #0 diesel was used throughout the experiments. The fuel's density (20 °C) and kinematic viscosity (20 °C) were 829.20 kg/m3 and 4.68 mm2/s, respectively. The fuel sulfur content was < 1 mg/kg, the heat value and cetane number were 43.44 MJ/kg and 52.30, respectively.In this investigation, two kinds of experiments were conducted, which were named eight-condition NRSC and NRTC tests, with two kinds of aftertreatment combinations of POC or PDF coupled with DOC, SCR and ASC. The aftertreatment combinations with POC or DPF were numbered as POC and DPF, respectively. Specifications of aftertreatments are shown in Table S1 of the Supplemental materials, and the diagram of the experimental setup is shown in Fig. 1. The specific process was strictly regularized by China's National Standard GB 20891-2014, and the detailed test process is described as follows (GB 20891-2014, 2014): 1) The ambient conditions of the test were kept stable at normal temperature and pressure, namely the ambient temperature was 25 ± 1 °C, the pressure was 101.30 ± 0.10 kPa, and the ambient humidity was between 50% and 70%. Laboratory atmospheric factor fa met the condition of 0.96 ≤ fa ≤ 1.06. The intercooler temperature should be within the manufacturer’s specified range ± 5 °C under the rated net power point, and it should not be lower than 20 °C. The intercooler’s outlet temperature should be between 45 °C and 55 °C at the rated power. The fuel temperature at the inlet of the fuel injection pump should be 33–43 °C. 2) All laboratory instruments and sensors were properly connected. The main test instruments/sensors and their parameters are reported in Table S2 of the Supplemental materials. The exhaust sampling system and gas analyzer needed to preheat to normal operating conditions. During the test, the speed and load of the diesel engine, inlet air temperature, fuel flow rate, intake and exhaust flow rates, POC or DPF inlet temperature and outlet temperature and pressure were measured when the diesel engine was working steadily. All the volume and flow rates were converted to the standard atmospheric state of 273 K (0 °C) and 101.30 kPa. And various pollutants (PM, NOX, HC, CO, CO2) were collected by sampling and corresponding gas analyzers at the frequency of 0.10 s. Gaseous pollutants were carried out by AVL AMA I60 gas analyzer, and the smoke was measured with AVL 439 opacimeter smoke meter and AVL415S filter-type smoke meter. The opacimeter smoke degree was used as PM indicator and can be estimated by calculating the light absorption coefficient in Eq. (1) (1) k = − 1 L A ln 1 − N 100 where, k is the absorption coefficient (m-1), N is the opacity of the smoke meter (%), L A is the effective optical path length (m). Specific emissions of PM can be evaluated indirectly by k . 3) Under the NRSC test, every steady operating condition ran for 10 min, so that the diesel engine could run fully and stably. The exhaust gas should pass through the analyzer during the last 3 min of each working condition and the analyzer output should be recorded by a tape recorder or the equivalent data acquisition system. The eight-condition points and weighting coefficient under the NRSC test are shown in Fig. 2b and Table S3 of the Supplemental materials. According to GB 20891-2014, in Table S3, four points were selected for the rated speed. The corresponding loads were 100%, 75%, 50% and 10%. Three points were selected for the intermediate speed, and the corresponding loads were 100%, 75% and 50% respectively. The idle load is 0. And the eight-condition points correspond to the serial numbers of the NRSC in Fig. 2b. 4) NRTC test baseline cycle conditions are shown in Fig. 2a. The transient test cycle lasted 1238 s, and each 1 s was a working condition. In order to fully study the emission characteristics of various species in transient working conditions, the computer recorded data every 0.10 s, where "H" represents the hot start state in this study. The cold start cycle test was initiated when the temperature of the engine coolant, oil, engine treatment system and auxiliary equipment was maintained between 20 °C and 30 °C. After the end of the cold start, the hot immersion period of 20 min was carried out immediately, and then the hot start cycle test was carried out. The final specific emission results were 90% weight of the hot-start cycle and 10% weight of the cold-start cycle. The ambient conditions of the test were kept stable at normal temperature and pressure, namely the ambient temperature was 25 ± 1 °C, the pressure was 101.30 ± 0.10 kPa, and the ambient humidity was between 50% and 70%. Laboratory atmospheric factor fa met the condition of 0.96 ≤ fa ≤ 1.06. The intercooler temperature should be within the manufacturer’s specified range ± 5 °C under the rated net power point, and it should not be lower than 20 °C. The intercooler’s outlet temperature should be between 45 °C and 55 °C at the rated power. The fuel temperature at the inlet of the fuel injection pump should be 33–43 °C.All laboratory instruments and sensors were properly connected. The main test instruments/sensors and their parameters are reported in Table S2 of the Supplemental materials. The exhaust sampling system and gas analyzer needed to preheat to normal operating conditions. During the test, the speed and load of the diesel engine, inlet air temperature, fuel flow rate, intake and exhaust flow rates, POC or DPF inlet temperature and outlet temperature and pressure were measured when the diesel engine was working steadily. All the volume and flow rates were converted to the standard atmospheric state of 273 K (0 °C) and 101.30 kPa. And various pollutants (PM, NOX, HC, CO, CO2) were collected by sampling and corresponding gas analyzers at the frequency of 0.10 s. Gaseous pollutants were carried out by AVL AMA I60 gas analyzer, and the smoke was measured with AVL 439 opacimeter smoke meter and AVL415S filter-type smoke meter. The opacimeter smoke degree was used as PM indicator and can be estimated by calculating the light absorption coefficient in Eq. (1) (1) k = − 1 L A ln 1 − N 100 where, k is the absorption coefficient (m-1), N is the opacity of the smoke meter (%), L A is the effective optical path length (m). Specific emissions of PM can be evaluated indirectly by k .Under the NRSC test, every steady operating condition ran for 10 min, so that the diesel engine could run fully and stably. The exhaust gas should pass through the analyzer during the last 3 min of each working condition and the analyzer output should be recorded by a tape recorder or the equivalent data acquisition system. The eight-condition points and weighting coefficient under the NRSC test are shown in Fig. 2b and Table S3 of the Supplemental materials. According to GB 20891-2014, in Table S3, four points were selected for the rated speed. The corresponding loads were 100%, 75%, 50% and 10%. Three points were selected for the intermediate speed, and the corresponding loads were 100%, 75% and 50% respectively. The idle load is 0. And the eight-condition points correspond to the serial numbers of the NRSC in Fig. 2b.NRTC test baseline cycle conditions are shown in Fig. 2a. The transient test cycle lasted 1238 s, and each 1 s was a working condition. In order to fully study the emission characteristics of various species in transient working conditions, the computer recorded data every 0.10 s, where "H" represents the hot start state in this study. The cold start cycle test was initiated when the temperature of the engine coolant, oil, engine treatment system and auxiliary equipment was maintained between 20 °C and 30 °C. After the end of the cold start, the hot immersion period of 20 min was carried out immediately, and then the hot start cycle test was carried out. The final specific emission results were 90% weight of the hot-start cycle and 10% weight of the cold-start cycle.It is necessary to convert the gaseous emission concentration (ppm) into specific emission [g/(kW·h)] at each operating condition. The exhaust gas flow was measured with air flow meters. Then, a tail gas analyzer was used to detect the concentration of exhaust emissions under various working conditions. After that, the mass flow rate of each exhaust emission was calculated by formulas (2)–(4). The emission calculation method (formula) in this paper came from China's National Standard GB 20891-2014 (GB 20891-2014, 2014). (2) NO x mass , i = 0.001587 × K H × NO x conc , i × G EXHW , i (3) HC mass , i = 0.000479 × HC conc , i × G EXHW , i (4) CO mass , i = 0.000966 × CO conc , i × G EXHW , i where, G EXHW , i is the exhaust flow rate (kg/h) at the i working condition; K H is the humidity correction coefficient of NO x . NO x conc , i , HC conc , i and CO conc , i are the concentrations (ppm) of NO x , HC and CO at the i working condition respectively; NO x mass , i , HC mass , i and CO mass , i are the mass flow (g/h) of NO x , HC and CO at the i working condition, respectively.Then, the mass flow rates of exhaust pollutants which calculated by Eqs. (2)–(4) were substituted into Eq. (5) to calculate the weighted specific emissions of each component under various working conditions. (5) A i = A mass , i × WF i ∑ i = 1 n P ( n ) i × WF i where, A mass , i is the mass flow rate (g/h) of A species under the i working condition; WF i is the weighting coefficient of the i working condition; P ( n ) i is the net power of i working condition; A i is the specific emission [g/(kW h)] of some gaseous pollutant under the i working condition.The specific emission of PM was calculated by the filter paper weighing method. For each operating point of the engine, the exhaust gas was passed through the filter paper for a period at a specific engine power. Then, the mass of the filter paper was weighed and the value was recorded. The unit of mass was μg, which should be converted to g/(kW h).The experimental error and validity were accessed. The test error was examined through the original emission measurement (without an aftertreatment system). The error came from the devices and random during the experiment. The error estimation method was adopted from Wu et al. (2020). The detailed accuracy of devices can be found in Table S2 of the Supplemental materials in the appendix. The random error was estimated by repeated tests, three times for the original emission measurement. The relative errors of CO, HC, NOX, and PM are 4.3%, 2.4%, 2.4% and 3.7%, respectively. And the error bar was added into the figures. Note that error bars have not been added to the transient result curves for clearer readability.The conversion rate and specific emissions between POC and DPF under the NRSC test are shown in Fig. 3. It can be seen from the figure that being equipped with an aftertreatment system can effectively reduce emissions. The overall emissions of PM, NOX, HC and CO are all lower than the China IV, US, and European Union (EU) standards emission limits. So, it is very necessary to add aftertreatment technology to non-road diesel engines. From the perspective of conversion rate, the DPF conversion efficiency of PM, NOX, HC, and CO emissions are 87%, 98.50%, 98%, and 75.50%, respectively. While the POC conversion efficiency of PM, NOX, HC, and CO emissions are 60%, 96.74%, 95.27% and 77.78%, respectively. This is mainly related to the internal structure and operating conditions of POC and DPF. In the test, we just did the same test after replacing POC with DPF (operating for 10 min in each operating condition and collecting data in the last three minutes). The catalyst and the catalyst carrier have not been changed, but the plugging of the POC outlet end has been added. The emissions of NOX and HC are very close and their conversion rate exceeds 95%, almost achieving zero emissions. However, the total CO emissions of POC are lower than DPF, and the conversion efficiencies of POC are higher than DPF. Noble metal catalysts (such as Pt and Pd) have priority to oxidize CO. The affinity of DOC for HC and CO is different. The surface catalyst of DOC can adsorb CO and HC, but CO reacts before HC, and the reaction rate of HC is lower than that of CO (Wang et al., 2008), and POC can further oxidize CO and HC to a certain extent. At the same time, the working conditions of the system also have an important impact on the emission of pollutants, as shown in Fig. 4. The detailed emission process will be discussed in detail in the next section.Non-road diesel vehicles are an important source of PM emissions (Dhital et al., 2021; Ganesh et al., 2019). The main components of PM include dry soot (DS), soluble organic fraction (SOF) and sulfate. The main reason is that oxygen-deficient fuel will undergo cracking and dehydrogenation under high temperature and pressure environment, and finally form soot particles. These soot particles will absorb various unburned or incompletely burned heavy HC and other condensed-phase substances during the cooling process and constitute the PM. Therefore, reducing PM emissions has always been a focus of research. Fig. 4a shows that the filtration effect of POC and DPF is affected by engine speed and load. The PM concentration at 2200 rpm is higher than that at 1400 rpm, and PM increases with the decrease of load, and the maximum occurs at 50% load. At the same time, in the third working condition (2200 rpm and 50% load, see Fig. 2b), both POC and DPF have the largest contribution rate (the proportion of each working condition), and they are 26.07% and 27.80%, respectively. At the rated speed, the PM contribution rate decreases with the increase of the load. At medium speed, the PM contribution rate increases with the increase of the load. This is mainly due to the low combustion temperature of low load and the richer mixture of high load, which results in fuel incomplete combustion and reduces soot oxidation rate. Meanwhile, as can be seen from Fig. 4a, the line of POC is always above DPF. That is to say, the purification conversion efficiency of DPF is higher than that of POC in each working condition. So DPF has more advantages in reducing PM than POC. For example, at 1400 rpm, the purification effect of DPF is greatly improved compared to POC, and the maximum increase is 76.51% at 100% load. That means that DPF is more conducive to reducing PM at medium speed and high load. The filtration efficiency is mainly related to pressure drop. The highly porous wall will show low pressure drop and thus low filtration efficiency (Konstandopoulos et al., 2012). The exhaust temperature has a similar trend relative to the pressure drop (Zheng and Banerjee, 2009), as shown in Fig. 4e and f, the pressure drop increases with the increase of exhaust temperature in DPF and POC channels, so the PM purification efficiency of fifth working condition (1400 rpm and 100% load) is greater than that of other working conditions.NOX is an important environmental pollutant. NOX undergoes a series of chain photochemical reactions under strong sunlight to generate ozone (O3) (Li and Cocker, 2018) and peroxyacetyl nitrate (PAN) (Sun et al., 2021), that is, photochemical smog of secondary pollutants (photochemical smog) (Sun et al., 2020). Therefore, it is also crucial to reduce NOX emissions. SCR is currently recognized as the most effective and promising technology for reducing NOX (Kang et al., 2018), and the purification efficiency of SCR on NOX is as high as 91.6% (Hu et al., 2021). At present, SCR is based on V2O5/WO3/TiO2 as a catalyst, and urea aqueous solution as a reducing agent is widely used to reduce NOX (Jung et al., 2017). Fig. 4b shows that NOX emissions decrease with load decreasing, and the NOX emissions of DPF are always lower than those of POC at the same speed and load. For example, NOX emissions of working condition 1 (2200 rpm and 100% load) and working condition 5 (1400 rpm and 100% load) are much higher than that of other working conditions. This is because the average air-fuel ratio (A/F) of the combustible mixture reduces as the engine load increases, causing an increase of the maximum combustion temperature (Nabi et al., 2021). At low engine load, although the excess air coefficient is larger, the lower combustion temperature is not conducive to the production of NOX (Deng et al., 2019). So, high load and high temperature are more conducive to the production of NOX, while the influence of speed on NOX is smaller than that of load.CO is an intermediate product produced in the combustion reaction process. It is mainly produced by partial lack of oxygen due to uneven mixing of mixed gases (Zhu et al., 2018). And too low temperature will also lead to incomplete combustion and increase CO emission. According to the latest study by Sun et al., they found that the HC and soot emissions of natural gas (NG) SI engine were much lower with hydrogen addition. In addition, the energy efficiency and fuel economy of the NG SI engine were improved with hydrogen addition (Sun et al., 2022). This provides a new idea for reducing CO in non-road diesel engines. Fig. 4c shows the CO emission characteristics under NRSC. It can be seen that CO emissions increase with load increases. The smallest CO emissions occur at idle speed, while the largest CO emissions occur at 2200 rpm and full load. The difference in CO purification effects between POC and DPF is minor, and the change trend is consistent. It is important to note that the original CO emissions are very low. Although the CO conversion efficiency of POC and DPF is between 75% and 80%, the influence of load is not obvious due to the very low original CO emissions. Combustion temperature and local hypoxia are still the key factors.Diesel engines produce HC mainly due to two reasons: uneven fuel mixing and poor fuel atomization quality. As shown in Fig. 4d, HC decreases with the increase of load at 2200 rpm (50–100% load). The load increase makes the engine burn more thoroughly, leading to an increase in exhaust temperature (see Fig. 4e) and ultimately a reduction of HC emissions. The HC purification effect of DPF is better than that of POC. However, it should be noted that the original emission of HC is very low. The emission of HC is mainly concentrated at a higher speed (2200 rpm), and the changing trend of the contribution rate is akin to the specific emission. This is due to the uneven mixture of fuel and surrounding air when at high speed in the spray area, the mixture becomes too rich and then becomes leaner gradually. And a too lean mixture is formed at the periphery of the spray before it catches fire, causing the fuel to not burn completely. Fig. 5 shows the overall value-specific emissions obtained from the NRTC test. It is a weighted value of 90% hot start cycle and 10% cold start cycle. PM, NOX, HC and CO all meet the China IV emission requirements. And the specific emissions of DPF are lower than those of POC. However, from the perspective of conversion efficiency, except for the greater difference in PM conversion efficiency, other emissions conversion efficiencies are relatively close. For example, the conversion efficiencies of POC to PM, NOX, HC, and CO are 60.12%, 95.45%, 92.82%, and 79.51%, respectively, and those of DPF to PM, NOX, HC, and CO are 92.83%, 96.99%, 96.86% and 81.45%, respectively. Compared with NRSC, the conversion efficiency of NOX and HC under NRTC has a small decrease, and the conversion efficiency of CO and PM under NRTC has a small increase, which is mainly affected by the operating conditions. The original emissions of HC and CO in the steady state and transient state are relatively low and have met the emission regulation limit, so PM and NOX are our key research objects.However, it is worth noting that for the POC purification system, PM has not reached the latest emission limits of the US and EU, but is very close to them, especially the US limit. The main reason is that the China IV standards mainly refer to the EU IV standards, which are equivalent to the Tier 4 emission regulations of the US. Therefore, China's emission regulations are one or two stages later than those of the US and EU. In order to meet stricter emissions regulations, the mainstream research route for the non-road diesel engine is still carried out in two aspects: internal purification and external purification technology. The internal purification technologies mainly include low-emission combustion systems, adding EGR (Yin et al., 2022; Zhao et al., 2022), low temperature combustion technology (Xu et al., 2022), fuel injection system optimization (Zhang et al., 2018), variable intake system, variable turbocharging compressor, homogeneous compression ignition technology (Zhou et al., 2022), alternative fuels (Liu et al., 2022a; Sun et al., 2022), dual-fuel (Atelge et al., 2022; Liu et al., 2022b; Tripathi et al., 2022) including adding H2 (Şanlı et al., 2021; Bose and Banerjee, 2012); In our another dedicated study, injection strategies (including the smoke limit strategy) were discussed. External purification technology is mainly related to the catalyst (E et al., 2020a), such as honeycomb microstructure optimization and/or improvement of catalyst dispersion, catalyst precious metal exploitation, catalyst structure optimization and continuous regeneration (E et al., 2020b; Zhao et al., 2021). Actually, the POC microstructure effect on PM reduction was discussed in our another dedicated study.In the NRTC test, we found that the original emissions of HC and CO were lower than the China IV standards emission limits, and the emissions were lower after being equipped with an aftertreatment system. So, we focused on the PM and NOX emissions, and only described the HC and CO emissions as a whole. The average value of CO during the cold start of POC is 30 ppm, and the peak value is 502 ppm at 45 s. While the average value of CO during the cold start of DPF is 20 ppm, and the peak value is 605 ppm at 51 s. DPF efficiency is increased by 33.33% during cold start and only 6.72% during hot start. That is mainly due to the emissions of cold start is higher than the hot start. It is caused by lower aftertreatment temperature and higher air-fuel ratio (Iodice and Senatore, 2016). At the same time, the CO reduction rates of POC and DPF during hot start are 81.50% and 82.74%, respectively. Hu et al. found that the CO oxidation efficiency of DOC during hot start is 55% (Hu et al., 2021), so POC and DPF can further improve the CO reduction efficiency. HC decreases sharply with the increase of air-fuel ratio. When the air-fuel ratio exceeds ϕ a = 1, HC will drop to the lowest value. However, if the air-fuel ratio is too large, HC will rebound again for unstable combustion. The HC fluctuating range during most time of cold start and hot start is relatively small, and the average value of the two start modes is also relatively close. However, compared with POC, the HC purification effect of DPF at cold start is increased by 71.16% on average, but it is only increased by 47.60% at hot start.Opaque smoke can evaluate PM emissions to a certain extent, and hot start accounts for 90% of the weighting ratio, so we focused on the hot start opaque smoke and NOX emissions, as shown in Fig. 6. During cold start, DPF can keep the opaque smoke emission at a low level with small fluctuations. The change trend is akin to that of hot start and the emission is close to zero, but the POC has greater fluctuations at higher speed. Compared with cold start, the emission of hot start is reduced, the emissions of opaque smoke in hot start are shown in Fig. 6a. POC changes frequently from 0 s to 300 s, which is related to changes in engine speed and load. In addition, the conversion efficiency of POC is low, but DPF always has higher conversion efficiency. The main reason is that the filtration principles of ceramic POC and ceramic DPF are different. POC relies on the pressure difference between the direct flow channel and the closed-end channel to achieve the flow of airflow, and captures the particles through the wall. Unlike the previous plugging method, we added ceramic plugging at the outlet port of the POC (see Fig. 1a), and the inner wall is coated with a layer of catalyst containing Pt and Pd, which can oxidize PM into CO2, and reduce HC and CO at the same time. The strong oxidation performance of POC is mainly derived from residual oxygen and NO2. DOC can convert NO into NO2 and increase the content of NO2 in POC intake flow. NO2 has strong oxidizing properties and can oxidize soot. Both upstream and downstream of the DPF are plugged alternately and asymmetrically by ceramics (see Fig. 1a). Exhaust gas enters from the inlet port and then exits from the outlet port. The carrier is coated with a layer of catalyst containing precious metals (such as Pt, Pd), which is conducive to the adsorption of DPF.It can be seen from Fig. 6b that the opaque smoke of POC and DPF varies greatly from 725 s to 760 s, which is very similar to cold start. Pressure drop can well characterize the filtration efficiency (Konstandopoulos et al., 2012). So, let’s look into pressure drop evolution, the pressure drop of POC and DPF during 725–760 s has a significant decrease (see Fig. 6f), which indicates that the exhaust resistance has decreased, which is mainly related to the exhaust flow rate and soot layer thickness. Usually, PM is filtered and regenerated in POC at the same time. If the efficiency of capture and regeneration are in balance, then the POC will work normally. But with the engine speed and load increase, PM will increase (see Fig. 6b, at high speed and load). The accumulation phenomenon will cause the catalytic efficiency of the catalyst coating to decrease, and the regeneration process will be blocked. So, PM will be discharged along the channel, and the POC conversion efficiency will decrease. When the accumulated soot is oxidized, the POC conversion efficiency will increase again. So, the catalyst coating has a substantial effect on distribution of soot on the wall. The accumulation of soot will also change the porosity and soot distribution of DPF, which will result in increase of pressure drop. Therefore, the filtration efficiency of DPF will also change, but the DPF efficiency is not significantly reduced due to two-end alternated blocking in the DPF structure. Fig. 6e also finds that after 820 s, the overall pressure drop is lower, which indicates that the filtration efficiency of POC and DPF is reduced. It can be seen from Fig. 6a (within the black circle area) that the opaque smoke of DPF has risen slowly. In other words, the filtration efficiency of DPF is reduced, however, the filtration efficiency of POC remains constant. This may be related to the particle size of PM, which is discharged from the channel as the particle size becomes smaller, resulting in an increase in PM emissions. This phenomenon requires further work to continue to explore.In the combustion process of diesel engines, NO is mainly produced, and NO accounts for up to 95% of NOX, which is mainly converted into NO2 through DOC and DPF (Liang et al., 2019; Salman et al., 2018). Due to diesel engines’ large excess air coefficient ( ϕ a ), the NO2 content is generally between 5% and 15%, NO2 will affect the reduction ability of SCR and the regeneration process of PM in DPF (Tighe et al., 2012; Jiao et al., 2017). Although POC and DPF can reduce NOX emissions to a certain extent, the conversion rate is low. Note that the former DOC catalyzer oxidizes part of NO into NO2, but DOC does not reduce NOX emissions, and that only changes the NOX composition. After NO2 enters POC and DPF (Reijnders et al., 2016), the molecular bond of NO2 can be broken at a lower temperature (about 250 ℃), and the resulting O2 can be burned with the captured PM to generate CO2. Thus, PM can be effectively removed. NOX is mainly removed by SCR. The SCR catalyst used in this experiment is mainly V2O5/WO3/TiO2, which is characterized by better selectivity, wide temperature window and resistance to sulfur poisoning to NOX (Kang et al., 2019). It should be noted that the reducing agent in SCR catalytic reaction is ammonia (NH3). Since more than 90% of NOX emissions are NO, the main principle of reduction is shown as Formula (R-1). This reaction is also known as the "standard SCR reaction", and O2 is essential. The reaction rate of Formula (R-2) is 17 times faster than that of formula (R-1) at low temperature, which is called "fast SCR reaction". It is beneficial to improve the low temperature activity of catalysts and solve the problem of low exhaust temperature of the diesel engine. Practical studies show that when the NO/NO2 value is 1 (Nova et al., 2006; Ciardelli et al., 2007), the optimal NOX conversion efficiency can be achieved. When the NO2 ratio is too high, the conversion efficiency will decrease. That is mainly due to the increase of (R-3) reaction, but the reaction rate is very slow (Hu et al., 2018). Ko et al. (2019) similarly found that when the ratio of NH3/NOX is 1, SCR has the highest efficiency. Nevertheless, too high NO2 in POC and DPF would worsen the reduction of SCR. (R-1) 4NO+4NH3+O2=4N2+6H2O (R-2) 2NH3+NO+NO2=2N2+3H2O (R-3) 6NO2+8NH3=7N2+12H2O There is a clear cutoff point for NOX emissions during the cold start. From 0 to 400 s, there is a large NOX emission, and the NOX emissions from 400 s to the end of the test are very low, almost reaching zero emission. This is mainly related to the inlet temperature of SCR. The best working temperature of SCR is 250 °C–450 °C, and the inlet temperature of SCR is only close to 250 °C at 300 s during cold start. The hot-start emission of NOX is shown in Fig. 6c. The NOX conversion efficiency of DPF and POC is very close, and the trend of change is also very similar. For example, after 300 s, the NOX of DPF always remains at a low level, and the NOX of POC fluctuates frequently, but the peak value is relatively small. In order to have a clearer understanding of NOX emissions, we have partially enlarged the view of Fig. 6c, as shown in Fig. 6d. It can be seen that the NOX peak values of POC and DPF during hot start are 361 ppm at 96 s and 345 ppm at 94 s, respectively, and that they are much smaller than the peak emission of cold start. That is mainly due to the engine has a 20-minute hot soak period after cold start. Therefore, the SCR inlet temperature decreases, and the SCR purification ability decreases. So there will be a large NOX emissions peak. After the temperature gradually rises, NOX emissions have a significant drop, and the SCR plays a major role in purification. However, the engine load has an important influence on NOX emissions. In the black part of Fig. 6d, the large fluctuation of NOX is related to the sharp decrease of engine load. For example, in the red circle area, both cold start and hot start have similar changes, but when the load is dropped quickly from 97% to 0%, the speed is only slightly reduced, and the NOX emissions of POC and DPF both increase slightly, but POC has more NOX emissions.In this paper, the DOC + DPF/POC + SCR + ASC coupling aftertreatment system is adopted to reduce non-road emissions in view of the advantages of non-road diesel engines and increasingly strict non-road emissions regulations. It is necessary to save costs while improving the emissions of non-road diesel engines in order to reduce the pollution of non-road diesel engines. Therefore, the influence of a coupled aftertreatment system on non-road diesel engine emissions under NRSC and NRTC (including hot start cycle and cold start cycle) conditions was discussed in detail, and the performance and feasibility of aftertreatment were evaluated and analyzed. Some important findings are summarized below: 1) DOC + DPF/POC + SCR + ASC can effectively reduce the emissions of non-road diesel engines, but the two aftertreatment systems are significantly affected by engine speed and load. Under steady-state conditions, DPF can reduce PM, NOX, and HC emissions more than POC, without producing more CO2. At high speed and high load, more PM will be generated. DPF is more conducive to reducing PM than POC, and the conversion rate is 87% and 60% respectively. NOX and HC conversion rates of DPF and POC are relatively close, both are above 95%. DPF and POC also have higher CO conversion rates which are more than 75%, and POC conversion rate of CO is slightly higher than DPF. 2) Under transient conditions, the cold start cycle emissions of PM, NOX, HC, and CO are much larger than the hot start cycle emissions, and the engine load is the main influencing factor. DPF and POC can effectively reduce PM emissions, and the PM conversion rate is 92.83% and 60.12%, respectively. The transient PM conversion rate of DPF is 5.83% higher than that of steady state, but NOX and HC are reduced by about 3%. For HC emission, from 400 s to 1238 s, the DPF purifying effect of the cold start is improved by 71.16% compared with POC. But the DPF purification effect of hot start is improved by 47.60% compared with POC, and the peak of NOX emission usually occurs at the position of load mutation. The DPF conversion of CO is slightly higher than POC. 3) POC and DPF have similar filtering effects on NOX, HC, and CO emissions, and can meet the China IV standards emission limits. For PM, although the filtration efficiency of POC is lower than that of DPF, it can still meet the China IV standards emission limits. Moreover, DPF needs to accurately control the regeneration conditions and regeneration frequency, which are prone to increase back pressure, long development period, and high operating cost. Therefore, POC can save costs while reducing emissions, and POC has great potential to replace DPF. DOC + DPF/POC + SCR + ASC can effectively reduce the emissions of non-road diesel engines, but the two aftertreatment systems are significantly affected by engine speed and load. Under steady-state conditions, DPF can reduce PM, NOX, and HC emissions more than POC, without producing more CO2. At high speed and high load, more PM will be generated. DPF is more conducive to reducing PM than POC, and the conversion rate is 87% and 60% respectively. NOX and HC conversion rates of DPF and POC are relatively close, both are above 95%. DPF and POC also have higher CO conversion rates which are more than 75%, and POC conversion rate of CO is slightly higher than DPF.Under transient conditions, the cold start cycle emissions of PM, NOX, HC, and CO are much larger than the hot start cycle emissions, and the engine load is the main influencing factor. DPF and POC can effectively reduce PM emissions, and the PM conversion rate is 92.83% and 60.12%, respectively. The transient PM conversion rate of DPF is 5.83% higher than that of steady state, but NOX and HC are reduced by about 3%. For HC emission, from 400 s to 1238 s, the DPF purifying effect of the cold start is improved by 71.16% compared with POC. But the DPF purification effect of hot start is improved by 47.60% compared with POC, and the peak of NOX emission usually occurs at the position of load mutation. The DPF conversion of CO is slightly higher than POC.POC and DPF have similar filtering effects on NOX, HC, and CO emissions, and can meet the China IV standards emission limits. For PM, although the filtration efficiency of POC is lower than that of DPF, it can still meet the China IV standards emission limits. Moreover, DPF needs to accurately control the regeneration conditions and regeneration frequency, which are prone to increase back pressure, long development period, and high operating cost. Therefore, POC can save costs while reducing emissions, and POC has great potential to replace DPF. Renhua Feng: Writing – review & editing. Xiulin Hu: Validation. Guanghua Li: Writing – review & editing. Zhengwei Sun: Resources. Banglin Deng: Conceptualization, Methodology.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 work is jointly sponsored by the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2021A1515010481). The authors appreciate the anonymous reviewers and the editor for their careful reading and many constructive comments and suggestions on improving the manuscript.Supplementary data associated with this article can be found in the online version at doi:10.1016/j.ecoenv.2022.113576. Supplementary material. .

Non-road emission regulations are becoming increasingly rigorous, which makes it necessary for non-road engines to adopt aftertreatment systems. The commonly used aftertreatments mainly include diesel oxidation catalytic (DOC), diesel particulate filter (DPF), particle oxidation catalyst (POC), selective catalytic reduction (SCR) and ammonia purification catalyst (ASC). The purpose of this study is to investigate the effects of using an integrated system (DOC + DPF/POC + SCR + ASC) on non-road diesel engine emissions under steady-state and transient operating conditions, respectively. The major works are the comparison between POC and DPF from the viewpoint of emission reduction. The results show that both POC and DPF can effectively reduce particulate matter (PM) and nitrogen oxide (NOX) emissions under steady-state conditions, and DPF has better purification effect than POC, especially for PM. The PM conversion rate of DPF is up to 87%, while that of POC is only 60% under the non-road steady-state test cycle (NRSC). Both NOX and hydrocarbon (HC) conversion rates are high, exceeding 95%. The conversions of PM, NOX, HC, and carbon monoxide (CO) of DPF in the non-road transient test cycle (NRTC) are 92.83%, 96.99%, 96.86% and 81.45%, respectively, while those of POC are 60.12%, 95.45%, 92.82% and 79.51%, respectively. Both the POC and DPF systems can meet the emission regulation limits. As a result, POC has the potential to substitute DPF in non-road engines due to its lower product and maintenance costs. We hope that the comparison study will provide useful guidance for improving the emissions performance of non-road diesel engines.

Data will be made available on request. Data will be made available on request.Hydrogen is considered to be a fuel of the future. Pure H2 and the product of its burning are environmentally safe. H2 exhibits the highest energy capacity per weight unit among all known fuels (121 kJ/g). Nowadays, 90% of the H2 is produced via natural gas reforming or light oil fractioning during crude oil conversion [1–3]. However, these processes are not sustainable. The decrease of fossil fuel stocks along with the reduction of atmospheric pollution requires exploiting alternative feedstocks. Among the possible alternatives, ethanol has several advantages as a chemical source of H2. Ethanol exhibits extremely low toxicity and can be produced via fermentation of renewable biomass feed [4–6]. To emphasize the fact that ethanol is obtained from biomass sources, the term “bioethanol” is often used. A typical ethanol concentration in crude bioethanol is 7 – 12 vol.% [7].Ethanol steam reforming (ESR) is an attractive, sustainable, and environmentally friendly approach to H2 production. Based on the stoichiometry principle, the ESR reaction may be represented as follows: C 2 H 5 O H + 3 · H 2 O = 2 · C O 2 + 6 · H 2 The ESR reaction is endothermic. In a gas phase, enthalpy change of the reaction per mole of ethanol equals ∆H298 = 173.4 kJ/mol, whereas ∆H298 = 347.4 kJ/mol for the reaction in a liquid phase [8,9]. It is worth noting that the expensive distillation process, which is an essential part of the ESR-based technology, may be avoided if the crude bioethanol is directly converted into H2 by steam reforming. Highly diluted ethanol (12 vol.% of ethanol) offers a high steam/carbon ratio appropriate for the steam reforming process [10].The ESR reaction allows extracting H2 from both, ethanol (50%) and water (50%). Moreover, H2 production from bioethanol by ESR reaction does not affect the concentration of carbon dioxide in the atmosphere [11–13]. Considering the increasing demand for renewable energy, the industrial implementation of bioethanol steam reforming (BESR) as an alternative source of H2 is highly realistic [14]. H2 produced via steam reforming from the renewable feed may be efficiently utilized for electric energy production, especially for small-scale electricity supply. Coupling the steam reforming with a fuel cell establishes a new kind of technology for power generation. Using portable power systems based on fuel cell application simultaneously resolves the two issues of H2 use, particularly, safe hydrogen storage and its transportation [15–17]. In this respect, bioethanol is a promising feedstock for alternative energy production. Using diluted bioethanol significantly reduces the energy costs compared to pure ethanol reforming.In the catalytic ESR process, the choice of the appropriate catalyst is a key factor for ethanol conversion and high H2 selectivity. A wide range of catalysts was explored to obtain optimum reaction conditions [8,11,18–21]. Most studies have been performed over supported noble metals (Pt, Pd, Rh, Ru, Ir) and non-noble metals, e.g. Ni and Co. However, limited attention was paid to the oxide catalysts. Nevertheless, the spinel-type MnFe2O4 belongs to the best catalysts toward H2 yield (> 90%) [20]. This catalyst is relatively cheap and exhibits no transformation under reaction conditions in contrast to many other oxide catalysts.Various hydrogen generators, including portable devices, have been introduced in recent years. A small-size generator based on n-dodecane steam reforming for naval use with the H2 productivity of 1.5 Nm3/h has been reported [22]. A combined all-in-one portable generator with a PEMFC has been developed using sodium borohydride as a source of H2 [23]. An autonomous setup for H2 production by methane steam reforming coupled to catalytic methane combustor has been developed [24]. However, the reported devices imply the use of chemical hydrides, ammonia boran [25], and hydrocarbons as a feed for H2 generation. These chemicals are typically expensive and scarcely can be produced in large amounts suitable for large-scale application of the corresponding H2 generators. In this respect, creating a portable H2 generator based on bioethanol feed is a timely topic in the development of green technologies.Previously, we have studied the ESR over the MnFe2O4 catalyst [26]. However, the raw bioethanol typically contains small amounts of other alcohols, e.g. propan-1-ol, propan-2-ol, butan-1-ol, butan-2-ol, pentan-1-ol. Due to the presence of the impurities, the peculiarities of the ESR process may be quite different compared to pure ethanol reforming. Herein, we investigate the steam reforming of the mixtures of ethanol and higher alcohols over the MnFe2O4 catalyst. The optimum parameters of the ESR process are highlighted to create a conceptual design of an autonomous catalytic H2 generator. To this end, the goal of the present paper is to introduce a conceptual design of a portable autonomous catalytic H2 generator based on steam reforming of either diluted ethanol or ethanol/higher alcohols mixture over the designed noble metal free MnFe2O4 catalyst. Taking that into account, we used water/ethanol molar ratio equal to 19/1. On one hand, that ratio is close to the water/ethanol ratio in bioethanol obtained via fermentation of biofeed [7]. On the other hand, this ratio falls within the range of water/ethanol ratios typically utilized for hydrogen production by steam reforming of diluted ethanol because using high molar excess of water in the feedstock prevents coke formation [10].The preparation of the MnFe2O4 catalyst has been performed by two different procedures. According to the first procedure, MnFe2O4 was synthesized by solvothermal decomposition of the heteronuclear complex [MnFe2O(CH3COO)6(H2O)3]∙2H2O. A detailed description of the synthesis technique is presented in [27]. The use of this complex allows the formation of ferrite MnFe2O4 with the exact stoichiometric ratio of Fe/Mn = 2. The obtained ferrite was calcined at 250°C for 5 h and denoted as MnFe2O4-HC. Ferrite MnFe2O4 forms a stoichiometric chemical compound. The formation of an alloy during the synthesis of the catalyst and after the catalytic test was not observed, according to the XRD analysis [27]. The method of decomposition of the heteronuclear complex yields oxides of a given stoichiometric composition because the stoichiometry is a priori defined by the ratio of the metal ions in the framework of the conventional trinuclear complex with molar ratio Fe/Mn = 2, which is confirmed by atomic absorption spectroscopy [27].Another procedure admits the manganese ferrite, further denoted as MnFe2O4-CP, preparation by chemical co-precipitation. Accordingly, the ammonia-water solution was added to a solution of Fe(III) and Mn(II) nitrates with a molar ratio of 2:1 under stirring. The obtained reaction mixture was kept at 90°C for 5 h under stirring. The final precipitate was separated by magnetic decantation, washed with deionized water, ethanol, and diethyl ether, dried at room temperature and calcined at 400°C for 2 h. Commercial chemically pure grade and analytical grade reagents were used for catalyst synthesis without additional purification.X-ray diffraction measurements (XRD) were carried out with a Bruker D8 Advance diffractometer, with a Cu-anode, λ = 0.154 nm, step 2θ = 0.05 exposition time 5 s/step. The identification of crystalline phases was performed by matching with the ICDD files in the PDF-2 Version 2.0602 (2006) database. BET surface areas were measured by using a Sorptomatic 1990 instrument by adsorption of nitrogen at 77 K. Electron diffraction analysis (EDA) was conducted on a PEM-125K transmission electron microscope (Selmi, Ukraine) at 100 kV acceleration voltage.The ESR reaction was carried out in a fixed-bed tubular quartz reactor at atmospheric pressure. The procedure of catalytic tests and reaction mixture analysis were described elsewhere [28]. A catalyst, approx. 1 g, with a particle size between 1 and 2 mm, was placed in a quartz reactor between two layers of quartz grains of the same diameter. It was held in a reaction mixture at each temperature for 1 h, followed by gas chromatography analysis. Before a chromatographic analysis, condensable vapors, e.g. alcohols, aldehydes, ketones, and water, were trapped using a condenser at -15°C (258 K).The product analysis included three gas chromatographs (GC) equipped with three thermal conductivity detectors (TCD), a flame ionization detector (FID), and four kinds of packed columns: (1) Molecular sieve 5A, (2) Polysorb, (3) tris-cyano-ethoxypropane/Polysorb, and (4) Separon-BD. Analysis of H2 and CO was carried out using column (1), TCD, and Ar as a gas carrier. Column (2) was used for CO2 and N2 analysis (TCD, He as a gas carrier). The liquid samples were analyzed with column (3) (TCD, He as a gas carrier). Hydrocarbons were analyzed using column (4) with FID detector and He as a gas carrier.The sensitivity of the TCD and FID detectors to each analyzed compound (response factor) was determined periodically by their calibration against standards of gas and liquid mixtures of known compositions.The calculation procedure was as follows. The inlet flow rates of the components of the initial reaction mixture (alcohol, water, nitrogen), mol/h, were determined. Knowing the volume of the liquid sample accumulated over a certain time, it is possible to calculate the outlet feed of liquid products of the reaction, l/h. The flow rates of the liquid components (alcohols, aldehydes, ketones, water) of the reaction products mixture (mol/h) were determined using the results of the chromatographic analysis (in terms of the component molar concentration, mole/l) and outlet feed of liquid products. The flow rates of the gaseous components of the reaction products mixture (mol/h) were calculated using the results of the chromatographic analysis (in terms of component mole fraction) and outlet feed of gaseous products (nitrogen was the internal standard). For the reported runs, the carbon balance is defined as the ratio of the product moles to the consumed moles of ethanol, accounting for stoichiometry. The carbon balance error did not exceed 5%.The following experimental conditions were used: temperature 500 – 700 оС, inert gas-carrier (nitrogen) rate Fg  = 8.0⋅10−2 mol/h, liquid feed rate Fl  = 8.9·10−2 mol/h, which corresponds to WHSV equal to 4000 h−1, molar ratio H2O/C2H5OH = 19 (2.7 mol.% C2H5OH, 50 vol.% H2O, N2 balance). This ratio is close to the water/ethanol ratio in crude bioethanol and widely utilized for hydrogen production by BESR. To compare the catalytic performance towards different alcohol/water mixtures, the following diluted alcohols were used: propan-1-ol, propan-2-ol, butan-1-ol, butan-2-ol, pentan-1-ol (1% (mol.) of each alcohol), and 4.75% (mol.) of ethanol. After a catalytic test, the catalyst was cooled in the N2 atmosphere to room temperature and stored for characterization.Ethanol conversion, X, and selectivity of carbon-based reaction products, SCn , were evaluated according to the following expressions: (1) X = F C n F E t , i n · 100 % (2) S C n = n · F C n ∑ n · F C n · 100 % where n is the number of C-atoms in a product Cn; FEt,in is an inlet feed of ethanol, mol/h; FCn is a feed of corresponding product, mol/h.H2 selectivity was defined as 100% under the assumption that 6 moles of H2 are formed per 1 mole of converted C2H5OH. H2 selectivity was calculated as follows: (3) S H 2 = F H 2 6 · F E t , i n · 100 % where F H 2 is an outlet feed of H2, mol/h.H2 yield, Y H 2 , was calculated as follows: (4) Y H 2 = X · S H 2 It should be also emphasized that the experimental analysis of BESR over the developed catalysts revealed that ethanol decomposition, carbon formation reaction, and methanation reactions do not occur.After the BESR process, the catalyst samples obtained by decomposition of heteronuclear complexes and by co-precipitation are the manganese ferrite MnFe2O4 with a spinel structure, according to XRD analysis (Fig. 1 a and 1b). Fig. 1c presents a TEM image of the as-synthesized catalyst MnFe2O4-HC. The particle size distribution is rather sharp and centered at 8 nm. For the as-synthesized catalyst MnFe2O4-CP, reflections were not observed on the XRD pattern and this may be explained by the small size of the catalyst nanoparticles (< 3 nm). This suggestion is confirmed by the presence of ring reflections in the electron diffraction patterns of the sample, from which the crystalline phase of ferrite with a lattice of the cubic spinel-type was identified [26]. Therefore, the catalysts mainly consist of crystalline particles. BET surface area of as-synthesized MnFe2O4-HC and MnFe2O4-CP samples was 56 m2/g and 140 m2/g, respectively. A comparison of the surface area and the crystallite size for MnFe2O4-HC and MnFe2O4-CP indicate that a difference in the surface area for these two solids is mainly associated with a difference in a crystallite size that is strongly affected by a synthetic procedure.To design an autonomous H2 generator an appropriate catalyst should be used or developed. Investigation of the optimal conditions of the catalytic reaction is also a key feature for the successful implementation of the process of H2 synthesis into a portable power generator. Herein, we provide a theoretical analysis of the optimal BESR reaction condition and develop an effective catalyst for the corresponding process.The detectable products of the BESR are CO2 and H2 only. However, depending on the reaction conditions and catalysts used, the reaction network may be very complex. The following reactions are suitable for thermodynamics analysis of BESR over catalysts used in this study [3]. This set of reactions accounts for all possible products that are typically observed in various experimental studies over different catalysts [2,3,8,10–13]. (R1) C2H5OH ⇄ CH3CHO + H2 (R2) CH3CHO + H2O ⇄ 2 CO + 3 H2 (R3) CO + H2O ⇄ CO2 + H2 (R4) CH3CHO ⇄ CO + CH4 (R5) CH4 + H2O ⇄ CO + 3 H2 (R6) 2 CH3СНО ⇄ СН3СОСН3 + СО + H2 (R7) СН3СОСН3 + 2 Н2О = 3 СО+5 Н2 (R8) C2H5OH ⇄ C2H4 + H2O These reactions are typical for the most of experiments presented in the literature. A linear combination of these reactions allows one to obtain other reaction equations that are used in the literature for a description of the ESR process.Equilibrium yields of the reaction products may be calculated based on the thermodynamic approach. Using this approach is especially essential because the optimal reaction conditions may be theoretically estimated. Thermodynamic calculations are based on using the equilibrium constant, K p, as a function of temperature, T, for reactions (R1) – (R8): (5) ln K p = A · T − 1 + B · ln T + C · T + D · T 2 + E Numerical values of A, B, C, D, and E are presented in Table 1 . These values were calculated using thermodynamic parameters at 500, 700, and 900 K [29].Reactions (R5) and (R7) are not independable. They may be obtained as a sum of other reactions from that reaction network: (R5) = (R2) – (R4) (R7) = 2(R2) – (R6) Therefore, these reactions were not taken into account in the thermodynamic analysis.It is worth noting that some catalysts may exhibit coke deposition. Typical reactions leading to coke deposition are the Boudouard reaction, methane cracking, and reverse gasification reaction: 2 СО = CO2 + С СН4 = 2 Н2 + С СO + H2 = Н2O + С No coke deposition was found for the catalysts used in our study for the selected conditions. Therefore, these reactions were not taken into account.Ethanol conversion is not thermodynamically limited at any temperature. Thermodynamic calculations provide no information regarding the product distribution of ESR reaction under kinetic control. Nevertheless, a thermodynamic analysis may be useful to choose the optimal parameters of the process. These parameters allow preserving the kinetically controlled reaction regime under thermodynamically favorable conditions. Therefore, the catalyst for the ESR process should simultaneously exhibit high activity in reaction (R1), as well as low catalytic activity in methanation, WGS, and coke deposition reactions. Both catalysts used in the present study satisfy these conditions. Therefore, the thermodynamic analysis was performed using reaction network (R1) – (R4), (R6), and (R8) using Eq. (5) with parameters presented in Table 1 to calculate equilibrium constants at different temperatures. Fig. 2 presents the equilibrium H2 yield versus temperature. Based on the temperature for which the equilibrium H2 yield achieves YH2eq  = 50% during the steam reforming, higher alcohols may be placed in the following order: ethanol < propan-1-ol < propan-2-ol < butan-1-ol < pentan-1-ol < 2-methylpropan-2-ol. For temperatures higher than approximately 270°C, the equilibrium H2 yield achieves 100% for each alcohol. Therefore, the appropriate reaction temperature is above 270°C. Fig. 3 demonstrates the temperature dependence of the ethanol conversion and hydrogen selectivity during water/ethanol steam reforming over MnFe2O4-CP and MnFe2O4-НC catalysts. The selectivity toward each reaction product is presented in Table 2 .During BESR over MnFe2O4-СР catalyst, almost 100% ethanol conversion is achieved in the temperature range between 500 оС and 550 оС. Selectivity toward СО2 increases from 57% to 81% with the temperature increase from 500 °C to 600 оС, whereas selectivity toward H2 reaches the maximum (SH2  = 79.7%) at 600 °C. The main reaction products are hydrogen and carbon dioxide. Insignificant amounts of oxygenates, e.g. acetic aldehyde (0.1 – 1.4%), acetone (0.3 – 6.1%), and hydrocarbons, e.g. methane (1.8 – 3.1%), С2 – С3 hydrocarbons (0.9 – 1.3%) were also detected among the products. No detectable amount of carbon monoxide was found among the reaction products at 500 – 600 оС. However, at 650 оС carbon monoxide formation was detected due to the reverse reaction of the water-gas shift (WGS).The ethanol conversion over the MnFe2O4-НC catalyst in the temperature range of 500 – 650 °С is 90 – 100%. This catalyst demonstrates higher selectivity toward H2 at 600 – 650 °С. The selectivity reaches 94.6%. The formation of the carbon monoxide on the MnFe2O4-НC catalyst is observed only at 700°C. Therefore, the MnFe2O4 catalysts do not support CO formation at low temperatures. Perhaps, that is related to the peculiarities of the surface reactions when CO2 becomes the final product of the ethanol transformations over the surface of the catalyst while the probability of the CO formation on the surface is miserable due to high oxygen surface concentration. Moreover, our results indicate that both MnFe2O4-CP and MnFe2O4-НC do not catalyze the water gas shift reaction at low temperatures due to kinetic limitations. The water gas shift reaction becomes appreciable only at 650оС for MnFe2O4-CP and 700 оС for MnFe2O4-НC. Selectivity for CO over these two catalysts is lower compared to the thermodynamics (15.6% at 650 °C and 18.8% at 700 оС) which is an indication of kinetic limitations.The results of thermodynamic calculations regarding equilibrium product selectivity are given in Table 3 . Under equilibrium conditions, the formation of H2 and CO2 is the most favorable, as follows from the data presented in Table 3 where selectivity for each product is given at equilibrium. A comparison between the results of catalytic steam reforming of bioethanol (Table 2) and thermodynamic equilibrium (Table 3) shows the following. With increasing temperature, the reaction system gradually approaches equilibrium, while on the MnFe2O4-HC catalyst, the equilibrium and experimental hydrogen selectivity at 650 and 700°C are almost identical. At low temperatures, the equilibrium selectivity for hydrogen significantly exceeds the selectivity observed in experiments, especially in the case of the MnFe2O4-CP catalyst. That may be caused by a low rate of acetaldehyde and acetone conversion under experimental conditions. As a result, the selectivity observed in experiments for acetaldehyde and acetone far exceeds their equilibrium values leading to a decrease in the selectivity for hydrogen.Other features of the MnFe2O4-HC catalyst are associated with a high selectivity toward CO2 that reaches 92.5% at 650 °C and the absence of CO among the products at temperatures up to 650 °C. Carbon monoxide content may be affected by WGS reaction (R3). The data presented in Tables 2 and 3 demonstrate that the WGS reaction (R3) occurs at a relatively low rate on ferrite catalysts (CO and CO2 concentrations are far from equilibrium). As a result, carbon dioxide, not CO, is a primary product of a sequence of ethanol transformations over ferrite. The concentration of CO is either significantly lower than the equilibrium value or CO is absent in the reaction products as it occurs in the temperature range of 300 – 600 °C. Seemingly, carbon monoxide does not desorb into a gas phase during catalysis and reacts with the oxygen located on the surface of MnFe2O4 yielding in CO2 formation. The reacted surface oxygen is balanced by the dissociative adsorption of water.Our results are in good agreement with the experimental studies of the bio-ethanol catalytic steam reforming over supported metal catalysts that show zero selectivity for CO over various catalysts [1]. Moreover, a comparison of the present study with the studies presented in the literature indicates that not only metal-containing catalysts but also metal oxides may show zero selectivity for CO in BESR. Fig. 4 shows the H2 yield versus temperature for MnFe2O4-HC and MnFe2O4-CP catalysts. MnFe2O4-СР catalyst exhibits higher H2 yield in the temperature range between 450 °C and 550 °C compared to MnFe2O4-НC catalyst, as it follows from Fig. 4. In contrast, the highest H2 yield is observed on the MnFe2O4-НC catalyst at Т > 600 °C. The MnFe2O4 catalyst productivity in the temperature range between 550 °C and 600 оС is 300 – 600 ml Н2/(h∙gcat.).The major impurities in raw ethanol are C1, C3, and C4 alcohols, which represent up to 87% of all impurities, C3 and C4 aldehydes, acetone, acetic acid, glycerin, ethers, and nitrogen-containing chemicals. The presence of these impurities complicates the steam reforming process and results in either an increase or decrease of the H2 yield and catalyst lifetime. The results of the steam reforming investigations carried out using model mixtures of raw bioethanol and real bioethanol are summarized in ref. [8].The temperature dependence of the alcohol conversion and H2 selectivity for C3 – C5 alcohols steam reforming over the MnFe2O4-CP catalyst is presented in Fig. 5 . Temperature increase from 500 °C to 650 °C leads to the conversion increase from 88% to 100% and from 80% to 98% for propan-1-ol and propan-2-ol, respectively. Butan-1-ol and pentan-1-ol conversion is higher compared to propan-1-ol and remains 100% in the whole range of the used temperature.Selectivity toward H2 during steam reforming of C3 – C5 alcohols in the temperature range between 500 and 650 °C typically overreaches 90%. Moreover, a 150 °C temperature increase results in a decrease of the selectivity from 98% to 94% and from 98% to 90% for propan-1-ol and butan-1-ol, respectively. For pentan-1-ol, SH2  = 94% – 96%, whereas for propan-2-ol SH2 increases from 89% to 96%. Considerable decrease of the H2 selectivity during butan-1-ol steam reforming is associated with the presence of carbon monoxide in the reaction mixture, which enhances the water gas shift reaction.For the steam reforming of the ethanol and C3 – C4 alcohols mixture, the ethanol conversion reaches 97% at 500 °C. For other alcohols in the mixture, conversion varies between 83% and 90% depending on the alcohol nature, and H2 selectivity is approximately 80% (Fig. 6 ). Based on the conversion value, the alcohols may be placed in the following order: butan-1-ol < ethanol < propan-1-ol ≈ butan-2-ol < propan-2-ol (at 300 оС) and butan-1-ol < butan-2-ol < propan-2-ol < propan-1-ol < ethanol (at 500 оС). At 600 оС, ethanol conversion and H2 selectivity decrease.Reforming product distributions over the MnFe2O4 catalyst is quite different at 300°C and 400°C. Particularly, at a lower temperature, the only reaction products are H2 (92.6 mol. %), СН3СНО (7.1 mol. %), and C2 – C3 hydrocarbons (0.3 mol. %). The reaction temperature increases up to 400°C leads to an increase in higher hydrocarbons yield, as a consequence of an increase in the alcohol dehydration rate. The reaction products at 400°C are as follows: H2 (63.8 mol. %), CH3CHO (0.8 mol. %), CO2 (32.7 mol. %), CH4 (0.2 mol. %), and C2 – C3 hydrocarbons (2.5 mol. %).By comparing the results of the ethanol-C3/C4 alcohols mixture reforming and individual C2 – C4 alcohols reforming, the following peculiarities are obtained. For ethanol-higher alcohols mixture reforming, an insignificant decrease in the ethanol conversion (ХEtOH  = 97.2%) is observed at 500 оС in contrast to water-ethanol mixture conversion (ХEtOH  = 98.2%). The butan-1-ol conversion in the alcohol mixture is only 83.3%, which is considerably lower compared to the conversion of the individual alcohol. For the propan-1-ol, the conversion is identical for both, mixture reforming and pure propan-1-ol reforming. Propan-2-ol exhibits an increased conversion in the mixture compared to pure alcohol reforming. The main reaction products for alcohol mixture reforming are H2 and CO2. At 500 °C, productivity toward H2 of the steam reforming process is higher for the alcohol mixture in contrast to the water-ethanol mixture without higher alcohols. This difference in productivity is governed by an effective steam reforming of higher alcohols with the utilization of the water vapor on the developed catalyst.The design of the autonomous catalytic H2 generator was created under ISO 16110 Hydrogen generators using fuel processing technologies and IEC 62282-5-100 Portable fuel cell power systems requirements. In general, the essential parts of a portable hydrogen generator are fuel processing system, fuel cell module, fuel supply system, onboard energy storage system, and water treatment system. The fuel processing system provides fuel conversion into H2. In the fuel cell module, H2 is converted into electric energy and heat in an electrochemical way [30,31]. The obtained heat and electricity are further integrated into the energy generation system. The fuel supply system may be either built-in or as a separate container that is refueled on demand. The onboard energy storage system provides energy for the correct work of the fuel cell module. The water treatment system improves the quality of either regenerated or freshwater to make it suitable for use in a compact power plant. Thereafter, to develop a portable autonomous generator, all these parts should be as compact as possible. Also, the profound efficiency of the catalyst is required to reduce the reactor volume.For a successful design of an autonomous catalytic H2 generator, the following items should be carefully considered: the type of the fuel cell, its characteristics which define the chemical reactions in the fuel cell, catalysts, temperature range, and the required feed. In this paper, we treat the proton-exchange membrane fuel cell (PEMFC). This cell works at low temperature (< 80°C) and is not sensitive to H2 purity, e.g. CO2 additives [32]. For the calculations of the generator efficiency, we use the characteristics of the commercially available PEMFC produced by Horizon Fuel Cell Technologies Company. Particularly, to produce 1 kW of electricity PEMFC requires 13 l of technical H2 per minute. These values were utilized to evaluate the technological and financial aspects of the portable autonomous catalytic H2 generator design.The design of the fuel supply system is defined by the reaction pressure of BESR. Steam reforming reaction results in an increase in the overall number of moles in the reaction mixture. Therefore, the higher is the reaction pressure, the lower is the reaction rate in equilibrium conditions. In a compact power station, PEMFC is operated under atmospheric pressure in contrast to apparatus of the natural gas steam reforming that is typically operated at relatively high pressure (15 – 30 bar) [33]. As a consequence, BESR should be performed at atmospheric pressure. The low pressure enhances the H2 yield at a lower temperature. Therefore, the fuel supply system should provide the component dosage under the pressure required for homogeneous feed flow.According to the overall BESR reaction scheme, the stoichiometric molar ratio water/alcohol equals 3, which corresponds to 46% of alcohol by weight. The initial bioethanol (grout) contains approximately 9% (vol.) of alcohol, which corresponds to the molar ratio of water/alcohol equal to 19 [34]. For a three-column scheme of the bioethanol synthesis, the grout distillate with an average alcohol content of approximately 40% (vol.) is obtained in the first column. This alcohol content is equivalent to the water/alcohol molar ratio of 3.5. The higher is the water/alcohol ratio, the higher is H2 yield [35]. The water excess also prevents the catalyst deactivation by coke deposition. However, increasing the water/alcohol ratio results in a higher heat amount required for water vaporization. In this study, the water/alcohol ratio identical to 3.5 was used. This composition may be easily obtained using partial purification of bioethanol grout by rapid water vaporization.For the portable H2 generator, the fuel processing system should exhibit a maximum of technological simplicity. All of the existing BESR technologies imply several reactors combined differently, e.g. BESR reformer, HT-WGS, LT-WGS, and Met – purification from CO [36,37]. This approach is reasonable for large-scale H2 production because many sub-processes (methane steam reforming, WGS, CO removal) are well-known and applied in industry in different operating regimes. In this respect, H2 synthesis by the BESR method in various reactors allows optimization of the whole process technology. However, these complicated schemes are unsuitable for a portable autonomous generator. In this paper, we propose an approach to carry out a BESR reaction in a single fixed bed reactor using the developed ferrite catalyst.The portable autonomous H2 generator is operated under low temperature and requires no strict specifications of the feed content. A principal technological scheme of the H2 generator is shown in Fig. 7 . The applicability of a single-reactor scheme is defined by the developed MnFe2O4-НC catalyst. Using this catalyst in the BESR reaction, almost 100% bioethanol conversion may be achieved at a relatively low reaction temperature (650°C). Very small amounts of the reaction byproducts e.g. oxygenate and CO, are obtained. The amount of CO is only 50 – 60 ppm, which is suitable for PEMFC stable functioning [38]. The unreacted bioethanol and reaction byproducts (oxygenates) do not affect the PEMFC operating. No CO2 poisoning the Pt-Ru catalyst in the fuel cell occurs. Water is required for stable fuel cell functioning.Consider 1 kW PEMFC fuel cell module. The latter requires 13 l of H2 per minute or 70 grams of H2 per hour [39]. For the technical evaluation of the autonomous H2 generator concept, the BESR reactor temperature was considered to be 650°C. Using the experimental data obtained for the developed catalysts, the mass balance of the proposed technological scheme is calculated (Table 4 ).The experimentally verified average productivity of the catalytic BESR process is 450 ml H2/(gcat∙h). Therefore, to produce 13 l of H2 per minute, 1.3 kg of the catalyst is required. The bulk density of the catalyst is approximately 1 g/cm3. The reactor volume for this BESR process equals 2.5 l.With the knowledge about the mass balance, the energy consumption of the technological scheme may be evaluated (Table 5 ). At the first stage, 627.05 g/h of the water-alcohol mixture is heated to 650 oС. This requires 1.96 MJ/h of thermal energy. To initiate the endothermic ESR reaction, an additional 1.19 MJ/h of heat is needed. Therefore, for the production of 13 l/h of hydrogen, the overall heat consumption is 3.15 MJ/h. This amount of thermal energy is consumed to produce 3.6 MJ/h of electricity using 1 kW PEMFC. The energy recovery under given reaction conditions equals 0.45 MJ/h. The water/alcohol ratio equal to 5 is recommended by Ref. [40] to prevent coke deposition, as well as to increase H2 yield. However, for water/alcohol ratio = 5, 3.74 MJ/h thermal energy is required to produce 3.6 MJ/h electricity, which is economically unreasonable. Energy consumption for H2 production may be reduced only by optimization of the technological scheme and reaction heat utilization. The reaction products are cooled from 650 oС to the temperature of the fuel cell module (60 oС). The optimized technological scheme is demonstrated in Fig. 8 .The presented scheme admits the utilization of thermal energy of the reaction mixture for vaporization and heating the input water-alcohol mixture. According to this scheme, the heat exchanger is placed between the reactor and fuel cell module and serves as a pinch zone. Using the pinch methods [41,42] allows evaluating the energy balance of the presented scheme. The calculated amount of energy consumed by the system is 2 MJ/h. The thermal energy required for the synthesis of 13 l/h of H2 is compensated by the heat recuperation of the reaction products. Therefore, the difference between the produced energy (3.6 MJ/h) and consumed energy (2 MJ/h) equals 1.6 MJ/h, which is 3 times higher compared to the use of the ordinary linear technological scheme.The energy saving of the proposed design of the hydrogen generator for PEMFC is highlighted by the pinch analysis that has been performed for the heat exchanger. The results of the pinch analysis are presented in Fig. 9 . The obtained results give the value of thermal energy QREC =1 MJ/h that can be recovered in this process at the pinch point for the technological parameters used. Pinch analysis also indicates no need for external cooling for the proposed design of the hydrogen generator at temperatures above the pinch point that support the autonomous operation of the generator.Also, the pinch zone (Fig. 8) provides the flexibility of the water/alcohol ratio in the input mixture. The role of the water/alcohol ratio is crucial because this ratio significantly affects the alcohol conversion, H2 yield, and catalyst lifetime. Heat recuperation of the reaction products allows using the water/alcohol mixture with arbitrary alcohol dilution because the water excess serves as an ordinary heat carrier. The energy required for vaporization and heating the excessing water is recovered in the second recuperative heat exchanger.2 MJ/h energy required for BESR operation may be obtained in 3 different ways: (i) electric heating using the produced electricity; (ii) burning the input water/alcohol mixture (however, a major issue with this approach is associated with the fact that diluted alcohol mixture is inflammable); (iii) the use of different fuel, e.g. natural gas. For a compact power plant, we believe that the simplest and most effective approach admits the use of electric heating.Furthermore, the energy conversion efficiency (η) of the developed power plant is evaluated. The energy conversion efficiency is calculated as the ratio between the produced electricity and the energy which may be obtained by burning the alcohol amount used for electricity production: η = E Q · 100 % , Where Е is the electricity amount produced by the fuel cell in a time, Q is the amount of thermal energy obtained by burning the equivalent alcohol amount at the same time. The heat of the thermal ethanol burning is 30.6 MJ/kg. 267 g of ethanol feed is required to produce 13 l of hydrogen and 3.6 MJ of electricity. The heat of 267 g of ethanol burning is 8.1 MJ. Therefore, the energy conversion efficiency toward ethanol equals 44%.The obtained value of the energy conversion efficiency is higher compared to other approaches based on the utilization of renewable energy sources. For instance, the solar cells exhibit the maximum energy conversion efficiency of almost 39% [43]. The energy conversion efficiency obtained during biomass fermentation is only 26.6 % [44].The presented calculations concern the conceptual design of the hydrogen generator. To construct a physical device, knowledge about the size of the main components and the specific size and volume of the whole system with respect to the targeted applications is required. The ready-to-use generator should also contain pressure control, temperature control, and the controller for the content of the water-gas mixture. However, the engineering aspects of the relevant construction lie outside the scope of the paper. To start the generator, any power source is appropriate, e.g. electricity, battery, input alcohol, etc. The power source is chosen based on the generator's construction, power output, and operating conditions. For this purpose, electrocatalysis or photocatalysis may be also applied. However, in this case, the construction of the generator will be too complicated.Hydrogen is a perspective fuel that may remove traditional fossil fuels in the future. H2 may be produced from renewable feedstock, e.g. bioethanol derived from biomass, using steam reforming. In this case, H2 yield significantly depends on the catalyst used and process conditions. Therefore, to investigate the optimal parameters for the conceptual design of the autonomous catalytic H2 generator, the process of the steam reforming of either diluted C2 – C4 alcohols or ethanol/higher alcohols mixtures is studied that reflects the composition of the raw bioethanol. Steam reforming is performed over the noble metal free MnFe2O4 catalyst with spinel structure. Using this catalyst 98 – 100% ethanol conversion is achieved in the temperature range between 500 and 650 °C, whereas H2 yield reaches up to 94.6%. For the ethanol/higher alcohol mixture, ethanol conversion and higher alcohol conversion at 500 °C are 97% and 83 – 90%, respectively. Selectivity toward H2 is ∼80%. The catalyst productivity is 0.3 – 0.6 l of H2/gcat·h in the temperature range between 550 and 600 оС.H2 obtained by steam reforming may be converted into electricity via the application of the fuel cells. Existing H2 power plants are inappropriate for small-scale electricity production. To this end, a conceptual design of a single-stage autonomous catalytic hydrogen generator is introduced. The developed generator concept contains no H2 purification equipment and utilizes the heat of the reaction products. For a stable functioning, 1 kW fuel cell consumes 0.63 kg of water/alcohol mixture per hour with 50% ethanol content. This fuel cell consumes 780 l of H2 per hour resulting in an energy conversion efficiency of 44%.The authors declare no conflict of interests.This research is partially supported by the Target Program of the National Academy of Sciences of Ukraine “Development of scientific grounds for hydrogen production, storage, and use in autonomous energy supply systems”. This work was completed despite the unprovoked invasion of Ukraine by Russia, supported by Belarus. The authors are thankful to the Armed Forces of Ukraine for serving our country and protecting our freedom.

The conceptual design of a portable autonomous catalytic hydrogen generator is introduced. The generator is based on the bioethanol steam reforming over the developed ferrite catalyst. The generator admits the utilization of thermal energy of the reaction mixture for vaporization and heating the input water-alcohol mixture. Moreover, the generator is characterized by a simple single-stage design without a stage for hydrogen purification. The generator is capable to produce 1 kW/h of electricity with 0.63 kg/h water/alcohol mixture (50% ethanol) consumption. The energy conversion efficiency of the developed generator is 44%. The proposed hydrogen generator is suitable for various applications related to on-site hydrogen production.

Composite solid propellants are extensively used as one of the most important propulsion energy sources in the field of rocket launching and space vehicles carrying [1–3]. With the rapid development of aerospace technology and the increasing competition among different nations, higher requirements have been put forward for the performance of composite solid propellants. Developing composite solid propellants with high energy characteristics, high firing range and high survivability have become the mainstream research directions. It is known that composite solid propellants are mainly composed of fuel, oxidizing agent, polymer binder and other functional components. Ammonium perchlorate (AP), which is a kind of strong oxidizer [4,5], presents some unique characteristics, including high density, high oxygen content, high heat generation, large gas production rate and high stability. Owing to these excellent features, AP has been widely used as oxidizing agent in composite solid propellants [6]. In addition, AP accounts for 65–70 wt percent of the overall propellant, and in some formulations, the content can even be as high as 90%. It can be seen that the characteristics of AP have a decisive impact on the property of composite solid propellant [7–9]. Burning rate and energy performance, which can directly dominate the ballistic property of missiles and rockets, are two key factors in evaluating the property of composite solid propellant. The improvement of burning rate and energy performance of propellant can be gained by enhancing the thermal decomposition of AP.Therefore, it is necessary to take feasible technical measures to modify the thermal decomposition behavior of AP. Generally, there are two major methods to modify the thermal decomposition of AP, including physical method (super-refining treatment of AP), and chemical method (utilizing various catalysts). Super-fining treatment of AP is one of measures currently taken to promote the thermal decomposition of AP, which can be ascribed to the increased specific surface area and active contact sites by decreasing the particle size of AP. Nevertheless, the superfine particles tend to aggregate, which will reduce the effectiveness in practical use. Besides, the super-fining process of AP should be carried out under severe conditions to ensure safety [10,11]. Hence, many research works are concentrated on different catalysts on AP decomposition and the thermal decomposition of AP can be accelerated by adding a small amount of catalysts [12–16]. Utilizing a reasonable catalyst can reduce the thermal decomposition temperature of AP, and increase the thermal decomposition rate and the amount of heat release, which are beneficial to shortening the ignition time and increasing the combustion rate of propellant. Moreover, the pressure index of propellant can also be adjusted by rational design of catalyst. Consequently, designing and constructing of different catalytic materials with complex microarchitectures have raised a wide concern in recent years.In the last few decades, a variety of catalysts, such as metal powders [17], metal alloys [18], metal oxides [19–24], metal hydroxide [25–27], metal-organic chelates [28–31], carbon-supported composites [32–35], and so forth, have been extensively researched and demonstrated to be effective in modifying thermal decomposition behavior of AP. In the past five years, most research work has focused on transition metal oxides and carbon-supported transition metal oxides, due to their high reactivity, versatile structure, low cost and natural abundance. Although extensive research on the catalytic effect for thermal decomposition of AP in the presence of transition metal oxide and carbon-supported transition metal oxides has been performed, the high thermal decomposition (HTD) temperature, the amount of heat release, and kinetic parameters, remain as key factors to evaluate the catalytic activity. This paper provides a comprehensive summary on transition metal oxides and carbon-supported transition metal oxides as catalysts for thermal decomposition of AP in recent five years.It is well known that transition metal oxides (TMOs) can serve as active catalysts for AP decomposition [36–38]. When a small amount of TMOs are introduced, the thermal decomposition performance of AP can be regulated. Up to now, a variety of TMOs with different morphologies and versatile composition have been explored to catalyze AP [39]. Here, we classify TMO catalysts into three categories, including single transition metal oxide catalysts, binary transition metal oxide catalysts and composite transition metal oxide catalysts. The related reports on catalytic performance for the thermal decomposition of AP are summarized.Nowadays, single transition metal oxide catalysts, such as ferric oxides, cobalt oxides, nickel oxides, zinc oxides, and copper oxides and so on, have been extensively researched, due to their facile fabrication, tunable structure and high catalytic activity.The catalytic performance of ferric oxide is closely related to its morphology and average particle size. When the particle size decreases to nano size, the catalytic activity of ferric oxide will be greatly improved. Cao et al. [40] investigated the catalytic performance of nano-sized α-Fe2O3 with four different particle sizes (127 nm, 115 nm, 86 nm and 84 nm) using differential scanning calorimetric (DSC) method. DSC tests indicated that the temperature for high-temperature decomposition (HTD) of AP decreased by 40.7 °C, 42.9 °C, 50.6 °C and 53.4 °C with the addition of 2 wt% of four different α-Fe2O3, implying the catalytic activity of α-Fe2O3 on the thermolysis of AP increased when the average particle size of α-Fe2O3 decreased. For pure AP, the released heat (ΔH) during the process of thermal decomposition was calculated to be 864 J/g. When AP was mixed with 2 wt% of 127 nm and 84 nm α-Fe2O3, the values of released heat were increased to 984 J/g and 1235 J/g, which indicated that the thermal decomposition of AP could be improved by nano-sized α-Fe2O3. The authors also studied the kinetic analysis and the results further illustrated the decreased particle size of α-Fe2O3 could increase the efficiency of catalytic action. They thought that more active sites were exposed on the surface of smaller particles, which would result in higher catalytic activity. Mechanism was not proposed here.Hossein and his co-workers studied the thermal decomposition behavior of AP catalyzed by nano-sized α-Fe2O3 with spherical morphology [41]. They found that both the particle size and the content of α-Fe2O3 can affect the decomposition of AP. Small average particle size and high content of α-Fe2O3 can lead to low decomposition temperature and high decomposition enthalpy of AP. Authors also further investigated the variation tendency of kinetic and thermokinetic parameters, the apparent activation energy (Ea) and the activation enthalpy (△H ≠) are remarkably decreased in the presence of α-Fe2O3 NPs. Activation energy can be defined as the minimum energy that is required from the reactant molecule to the activated molecule in a chemical reaction. The smaller the activation energy is, the higher the reactivity is. The activation enthalpy (△H ≠) represents the reaction heat that the molecules absorbing or releasing from stable state to activated state. The decreased values of △H ≠ imply less energy is needed during the reaction process. Hence, the reactant activity of AP is improved in the presence of α-Fe2O3 NPs.Sharma et al. [42] investigated the catalytic performance of hexagonal cones structural α-Fe2O3 with average particle size around 400–500 nm. Adding 2% of α-Fe2O3 to AP can remarkably decrease the LTD and HTD temperature by 20 °C and 75 °C, respectively. A possible mechanism has been proposed by the authors according to electron transfer mechanism, as shown in Fig. 1 .Generally, the decomposition of AP undergoes three primary steps, including endothermic low-temperature crystal transformation, exothermic low temperature decomposition (LTD) and exothermic high-temperature decomposition (HTD). The LTD process acts as a controlling step and electrons transfer from ClO4 - to NH4 + during this process, while for HTD, the main reaction can be attributed to the transformation from oxygen (O2) to superoxide ion (O2 -). Due to the distinct morphology, remarkable photoelectric and conductive performance of α-Fe2O3 HCs, electron movement can be enhanced, which might accelerate electron transmission from ClO4 - to NH4 +. Besides, the accelerated electron flow would facilitate the translation of O2 into O2 - . Hence, the thermal decomposition of AP is considerably enhanced.Researchers are paying great attention to fabricating cobalt oxide (Co3O4) on account of its variety of application in the fields of catalysts [43], sensors [44], lithium sulfur batteries [45], super-capacitors [46] and so on. As an important member of transition metal oxide, Co3O4 present outstanding catalytic activity towards the thermal decomposition of AP [47]. Li and co-workers [48] introduced Co3O4 spherical microspheres to catalyze AP, the thermal decomposition of AP presents a quite different feature in comparison with that of pure AP. There is only one strong exothermic peak located at 325.4 °C with the addition of 2 wt% of Co3O4 microspheres. The thermal decomposition temperature of AP was 111 °C lower than that of pristine AP, indicating Co3O4 microspheres show outstanding catalytic activity on thermolysis of AP. Moreover, the heat release of the mixture was estimated to be 1312.9 J/g, which is 3.76 times higher than that of pure AP (349.0 J/g). The activation energy (Ea) and the pre-exponential factor (lnA) of the mixture were calculated to be 121.9 ± 2.87 kJ/mol and 4.40 ± 0.02 min−1, respectively. Whereas, the values of Ea and lnA for pure AP were calculated to be 280.5 ± 11.8 kJ/mol and 26.40 ± 0.04 min−1, respectively. With the catalytic effect of Co3O4 microspheres, the values of Ea and lnA for AP were significantly decreased, indicating the as-prepared Co3O4 microspheres possess highly catalytic efficiency in AP thermal decomposition.The performance of nano-sized materials is highly related to their average particle size (APS) and specific surface area (SSA). Hossein and his cooperator [49] systematically researched the effect of nano-sized Co3O4 with various APS and SSA on thermolysis temperature of AP. Solvent and non-solvent methods were utilized to fabricate AP/Co3O4 nanocomposites (2 or 5% of Co3O4 NPs in weight percentage). The specifications of three kinds of commercial Co3O4 nanoparticles (marked as A, B and C) with different APS and SSA are summarized in Table 1 and their catalytic performance on AP are listed in Table 2 .APS, SSA and the content of Co3O4 can directly affect the decomposition behavior of AP according to the results listed in Table .2. With the decrease of APS and the increase of SSA, the catalytic efficiency, including decreased decomposition temperature and enhanced decomposition heat, is remarkably improved. Authors have also illustrated that catalytic performance can be improved by increasing the content of Co3O4 on the thermal decomposition of AP.It is well known that the property of materials can be adjusted by tuning their microstructures. It is of great importance to design and construct various micro/nano materials with complex microstructures. Low-dimensional micro/nano structures, including zero-dimensional (0D) nanoparticles [50], one-dimensional (1D) nanowires [51], and two-dimensional (2D) nanosheets [52], have been extensively researched on account of excellent property such as small grainsize, exposed active sites, high specific surface area and shortened mass transfer distance. However, the nano structures tend to aggregate due to their high surface energy, which will inhibit practical applications. Some achievements have been made for the preparation of nano transition metal oxides, but the aggregation still remains a challenge for developing catalysts with high activity. In order to inhibit the aggregation of low-dimensional nanomaterials, an efficient measure can be taken by designing three-dimensional (3D) hierarchical micro/nanostructure. Investigations by Miao et al. [53] prove that different morphologies of Co3O4 have different impact on the thermal decomposition of AP. Different morphological 3D hierarchical Co3O4 micro/nanostructures (Fig. 2 , Sample 1#-5#) are introduced as catalyst for AP decomposition.DTA results indicate that different morphologies of Co3O4 micro/nano structures present different activity on thermal decomposition of AP. The thermolysis of pure AP undergoes two weight loss procedures, the initial decomposition temperature was about 283 °C and the final decomposition temperature was around 443 °C. When hierarchical Co3O4 micro/nano structures (Sample 1#-5#) are employed, the related initial decomposition temperatures are decreased to 235, 233, 234, 232 and 232 °C, and the final decomposition temperatures are decreased to 306, 308, 315, 296 and 301 °C, respectively. DSC tests confirm the LTD and HTD process are merged into one exothermic process and the decomposition temperature decreased significantly compared with that of pure AP. Whereas, the exothermic heat is enhanced and the values are increased to 1197, 994, 1228, 933 and 1123 J/g for AP with the addition Co3O4 catalyzers (Sample 1#-5#), respectively. These results clearly imply that Co3O4 present good catalytic activity and the catalytic performance can be controllably tuned by regulating the morphologies of Co3O4 nanoparticles.Besides, a multiple of Co3O4 with different morphologies and particle sizes have been fabricated, and the detailed information including method of synthesis, morphology, particle size, surface area, HTD temperature of AP with and without catalysts, have been summarized in Table 3 .According to the data in Table 3, it can be seen that HTD temperatures for pure AP utilized in different papers are varied. This phenomenon can be ascribed to the difference of the physical properties of AP utilized in different papers, such as particle size, particle size distribution, and morphology of AP. Besides, test conditions during the DSC period, such as types of gaseous condition (nitrogen atmosphere or oxygen atmosphere, gas flow et al.) and types of crucible (aluminum crucible or alumina crucible). All the above mentioned factors can affect the decomposition temperature of AP. Moreover, it can be also observed that morphology, particle size, surface area, and the contents of catalysts have great influence on the catalytic performance on Co3O4.Nickel oxide (NiO), as a p-type transparent semiconductor, has been widely used in electronic, magnetic and catalytic aspects. For thermal decomposition of AP, NiO nanoparticles have been drawn great attention due to their apparent catalytic activity. A comparative research on catalytic effects of two different morphologies of NiO was performed by Zhao and his co-workers [66]. NiO microflowers present higher catalytic activity than that of NiO nanorods, which could be attributed to the difference of specific surface area. Based on the experimental results in this paper, the specific surface area for NiO microflowers is calculated to be 41.725 m2/g, which is higher than that of NiO nanorods (38.077 m2/g). This means more effective and active sites of NiO microflowers would be exposed on the surface, which is helpful to gas adsorption reaction. Therefore, the catalytic activity of NiO micro flowers is better than NiO nanorods.Sharma et al. [67] reported a green and eco-friendly biosynthetic strategy to fabricate NiO nanoparticles (NPs) by using leaf extract of plant calotropis gigantea. The as-obtained NiO NPs present spherical morphology with uniformly distributed particle size about 20–50 nm. The catalytic results indicate NiO NPs prepared by biosynthetic method possess better catalytic activity than the NPs fabricated by chemical routes. Authors also studied the dependence of Ea on different extent of conversion (α) for AP and mixtures of AP with NiO NPs (Fig. 3 ). According to the results, Ea for pure AP are higher than those for AP mixed with NiO NPs at all values of α. The variation tendency between Ea and α manifests that themolysis of AP is a complicated interaction effect of multiple, competing process and the rate limiting process varies with the extent of conversion. At the initial stage of α, the high values of Ea may be dominated by nucleation and growth of nuclei. Whereas, the lessening of Ea observed at α > 0.15 is attributed to the transition from kinetically controlled decomposition process to the mass transfer controlled decomposition process [68].Among various transition metal oxides, zinc oxides are also active catalysts in the thermal decomposition of AP. Tian and his co-workers prepared hierarchical porous ZnO hollow microspheres by a facile template-free method in mild experimental conditions [69]. The as-obtained ZnO hollow microspheres were assembled by ZnO nanorods and exhibited exposed (001) facets on the external surface. Both ZnO hollow microspheres and ZnO nanorods show catalytic activity towards the thermal decomposition of AP. In the presence of ZnO hollow microspheres, the decomposition temperature of AP is reduced to 308 °C and the decomposition heat release can reach up to 1174 J/g. The maximum decomposition temperature and the decomposition heat for AP are estimated to 321 °C and 959 J/g, when ZnO dispersed nanorods are added. Kinetic study indicates the values of Ea are remarkably decreased to 63 ± 7 kJ/mol and 90 ± 11 kJ/mol with the catalytic effect of ZnO hollow microspheres and ZnO dispersed nanorods, respectively. Compared with that of ZnO dispersed nanorods, the catalytic activity of ZnO hollow microspheres is prominent in the thermal decomposition of AP. This may be caused by the structural difference between microspheres and nanorods, such as specific surface area, crystallinity and exposed facets. ZnO hollow microspheres possess a larger specific surface area than ZnO dispersed nanorods, which is beneficial for the adsorption and diffusion process of gaseous HClO4 and NH3 (Fig. 4 (b)). The exposed (001) facets positioned at the external surface of ZnO hollow microspheres can also accelerate the generation of active oxygen species from the adsorbed HClO4 which will further oxidize NH3 gas. Hence, the absorbed gases will be decomposed. While for ZnO dispersed nanorods, all of the (100) facets, (101) facets and (001) facets are exposed to the gaseous HClO4 and NH3 (Fig. 4 (a)). Although most gases were absorbed by the (100) facets, they will not be decomposed [70], which will affect catalytic performance. Therefore, the catalytic activity of ZnO hollow microspheres are enhanced compared with ZnO dispersed nanorods.Oxides of copper, as important transition metal oxides, have been extensively researched in the aspect of thermal decomposition of AP because of their prominent catalytic performance. Ke et al. [71] prepared three-dimensionally ordered microporous (3DOM) CuO and investigated its catalytic performance for the thermal decomposition of AP. DTA results illustrated that with the effort of 2 wt% 3DOM CuO, the HTD temperature decreased to 354.9 °C, and the heat-release of the apparent decomposition of AP increased from 950 J/g to 1453 J/g. The excellent catalytic activity can be ascribed to large surface area and good mass transfer performance of 3D unique structure. Xie and his co-workers [72] fabricated one-dimensional CuO nanofibers by electrospinning method. They investigated the catalytic performance on the thermal decomposition of AP by TG and DTA. The HTD temperature of AP/CuO nanofibers were decreased by 101.9 °C compared with pure AP, indicating CuO nanofibers possess excellent catalytic activity. They ascribed this phenomenon to the higher surface to volume ratio of beaded CuO nanofibers. Hossein et al. [73] prepared uniformly distributed CuO nano particles by calcination of copper carbonate. With the addition of 0.5%, 2% and 5% CuO NPs, the HTD temperature of AP was reduced by 69.8, 66.9 and 104.5 °C. The decomposition heat increased to 1356, 1512 and 1588 J/g with the catalytic effect of 0.5%, 2% and 5% CuO NPs, whereas, the decomposition heat for pure AP was only 728 J/g. The activation energy (Ea) was also decreased remarkably when CuO NPs were employed. With the addition of 5% CuO NPs, the value of Ea decreases to 178.9 kJ/mol, which is approximately 65% of the value for pure AP (280.3 kJ/mol). Authors explained the catalytic performance by electron transfer mechanism. In their opinion, metal oxides act as a bridge for the transportation of electrons, which speed up the electron transferring from ClO- 4 to NH+ 4, thus, the decomposition behavior of AP was enhanced.Luo et al. [74] investigated three different morphologies of Cu2O cubes (cubic aggregate, mono-dispersed cube and {100} planes etched cube) for the thermal decomposition of AP. According to the calculated kinetic parameters, the numerical values of E arrange in an ascending order of {100} planes etched cube (92.6 J/mol), mono-dispersed cube (103.1 J/mol), cubic aggregate (110.4 J/mol). These results indicate the average E for AP mixed with Cu2O cubes are less than half the average E of pure AP (280.2 J/mol), implying Cu2O cubes possess outstanding performance in catalyzing AP decomposition. Besides, the {100} planes etched cube presents the highest activity in the aspect of decreasing the apparent activation energy. They also investigated the complete decomposition time of AP mixed with Cu2O cubes varies with temperature. The catalytic activity for the three types of Cu2O cubes can be easily distinguished in predicting the isothermal decomposition of AP. The results disclose that {100} planes etched cube shows better catalytic performance in AP isothermal decomposition than the other two.Besides single transition metal oxides, binary transition metal oxides with spinel structures have drawn great attention for catalyzing AP decomposition, due to their superb catalytic activity caused by the synergistic effect between two different constituent parts [75,76]. Spinel crystal structures usually can be expressed by the formula of AB2O4, where A and B represent di- and trivalent metal cations, respectively.Xiao et al. prepared mesoporous ZnCo2O4 rods through oxalate co-precipitation combined with controlled thermal decomposition method without any template [77]. The oxalates precursor was calcined at settled temperature under a slow heating rate and the nano-scaled ZnCo2O4 crystallites were automatically gathered to generate mesoporous ZnCo2O4 rods. They found that the calcination temperature could not change the ultimate structures of ZnCo2O4 rods, but the specific surface areas are greatly influenced by the calcination temperature. ZnCo2O4 nano crystallites will grow rapidly and the pore network will collapse under high calcination temperature. The specific surface area of ZnCo2O4 will be decreased as calcination temperature arises. Authors have also demonstrated the effect of increasing specific surface areas on the thermal decomposition of AP which possessed accelerated catalytic activity by increasing the specific surface areas. ZnCo2O4 rods calcined at 250 °C possess the largest surface area (102.34 m2/g) and highest catalytic performance, which can significantly reduce AP pyrolysis temperature by 162.1 °C. The catalytic activity of ZnCo2O4 rods can be explained by electrons transferring mechanism. Briefly, ZnCo2O4 rods act as a bridge for electrons transferring from ClO- 4 to NH+ 4 and from O2 to O- 2. Owing to high specific surface area, great adsorption of the mesoporous ZnCo2O4 rods and positive synergistic catalytic effect of binary oxide, the decomposition behavior of AP will be enhanced with the addition of ZnCo2O4 rods.A comparative investigation on catalytic performance of spinel MnCo2O4 nanoparticles and unclaimed MnCo2O4 precursor on the thermal decomposition of AP was done by Juibari and his co-workers [78]. The results indicate MnCo2O4 NPs present promising catalytic activity in decomposing of AP, while, the unclaimed MnCo2O4 precursor has little effect on thermolysis of AP. When 2, 3, 4 wt% of MnCo2O4 NPs are employed, the released heat of AP increase to 1350, 1410 and 1480 J/g, meanwhile, the HTD temperature shift downwardly to 308, 297 and 293 °C, respectively. The results illustrate the catalytic performance can be tuned by changing the content. The kinetic parameters of thermal decomposition of AP further indicate the reaction rate increases with the effort of MnCo2O4 NPs. As a p-type semiconductor, MnCo2O4 possess active d orbital of Co3+ (3d 5) and Mn2+ (3d 5) [79], which can be contemporaneously involved in the process of electron transfer and speed the process by simultaneous exposure to NH4 + and ClO4 -: Co3+ + ClO4 -→ Co2+ + ClO4 The bivalent cobalt cation (Co2+) are unstable and will transform Mn2+ (3d 5) into the Mn3+ (3d 6) during another electron transfer process: Co2+ + Mn2+ → Co3+ + Mn3+ A synergistic effect is probable to take place between Co3+ and Mn3+, which in turn promotes the formation of active sites of Mn+ and Co4+. The active sites play an important role in accelerating the catalytic process.Copper chromite is also an active catalyst for modifying thermal decomposition behavior of ammonium perchlorate. Hosseini and his co-workers [80] prepared a pure phase of spinel copper chromite by a sol-gel method. Authors investigated the catalytic performance of different Cu–Cr–O. The results indicated that different samples with various morphologies presented different catalytic activity. With the catalytic effect, all of the exothermic peaks of AP decreased. Among these, the sphere-like CuCr2O4 NPs presented the highest catalytic activity in reducing the decomposition temperature of AP. The sphere-like morphology of CuCr2O4 can effectively prevent nanostructures from aggregating, resulting in decreased particle size and uniform distribution. Moreover, the more crystallization makes sphere-like CuCr2O4 NPs a pure phase. All these factors endow sphere-like CuCr2O4 NPs with the highest catalytic activity.Nano-structured composite materials (or hybrid materials) with extraordinary physico-chemical performance have been widely researched and applied in versatile fields, ascribing to synergistic effect among different composite components. Inspired by this, extensive studies are focus on designing and synthesis of various nano-sized composite transition metal oxide to enhance catalytic activity toward AP decomposition [81]. β-AgVO3/ZnFe2O4 nanocomposites were employed as catalyst for thermal decomposition of AP by Abazari and co-workers [82]. As a comparison, β-AgVO3 and ZnFe2O4 were also prepared, respectively. According to the DSC tests for pure AP and AP mixed with 3 wt% of β-AgVO3, ZnFe2O4, and β-AgVO3/ZnFe2O4 nanocomposites, the HTD temperatures shift from 432 to 402, 367 and 339 °C, respectively. Moreover, the heat release (ΔH) for pure AP, AP + ZnFe2O4, and AP+β-AgVO3/ZnFe2O4 were estimated to be 764.8, 1169, and 1487.3 J/g, respectively. The results indicate that β-AgVO3/ZnFe2O4 nanocomposites are more active than β-AgVO3 and ZnFe2O4.Paulose et al. [83] prepared copper oxide alumina composite by using block copolymer template assisted sol-gel method. Mesoporous copper oxide dispersed on alumina (MCO) with a series of rations of copper oxide and alumina were synthesized. When introducing MCO, the crystallographic phase transition temperature of AP remained unchanged and the LTD temperatures were not remarkably reduced, indicating mesoporous CuO–Al2O3 have a slight impact on the primary decomposition of AP into ammonia and perchloric acid. Whereas, all the samples of as-obtained CuO–Al2O3 can significantly influent the HTD temperature. The exothermic temperature in HTD process declined, illustrating MCO samples can accelerate the decomposition of AP at a lower temperature.Nanoparticles are easily agglomerated, which remarkably decreases their specific surface area and catalytic activity. In order to overcome this problem, carbon materials, such as graphene, nitrogen-doped graphene, graphitic carbon nitride, carbon nanotubes, carbon black and so on, can be employed as a substrate to decorate nano-sized transition metal oxides. Carbon-supported nanocomposites present the combinative merits of nano-sized transition metal oxides and carbon based materials to produce excellent catalytic performance.The catalytic activity can be remarkably enhanced when the particle size is in nanometer-scale. Bare Fe2O3 nanoparticles tend to aggregate and fewer active sites are exposed, resulting in the decrease of catalytic activity. Graphene, owing to its unique structure and performance, can be a promising substrate to disperse and stabilize nanoparticles. With this in mind, the catalytic activities of Fe2O3 nanoparticles have been improved considerably by utilizing graphene as substrate. Lan [84] and co-workers synthesized graphene/Fe2O3 aerogel via a facile sol-gel and supercritical carbon dioxide drying method, as shown in Fig. 5 .The Fe2O3 nanoparticles in graphene/Fe2O3 are spherical and well dispersed on the graphene sheets. The specific surface area of graphene/Fe2O3 aerogel (101 m2/g) is much larger than that of pure Fe2O3 nanoparticles (13 m2/g), which confirms that graphene could prohibit the aggregation of Fe2O3 particles. The exothermic peaks for low temperature and high-temperature shift to a lower position with the addition of graphene/Fe2O3 aerogel. The exothermic heat shows a rising trend with the increased contents of graphene/Fe2O3 aerogel. Yuan et al. [85] synthesized Fe2O3/graphene nanocomposite by hydrothermal method. Fe2O3 nanoparticles are homogeneously distributed on the wrinkled graphene sheets and the particle sizes are ranged from 50 nm to 80 nm. DSC tests indicate that both Fe2O3/graphene and Fe2O3 show high catalytic activity in the thermal decomposition of AP, and Fe2O3/graphene show higher catalytic activity than pure Fe2O3, which is related to the high theoretical surface area and good conductivity of graphene. Graphene could not only prevent the agglomeration of Fe2O3 but also provide accelerated electrons to enhance the decomposition of AP. Hence, the catalytic performance of Fe2O3/graphene is superior to pure Fe2O3. So, the support of graphene can effectively improve the catalytic properties of Fe2O3 nanoparticles.To improve the dispersity of CuO nanoplates in the graphene nanosheets, a facile one-step in situ method was employed to fabricate G/CuO nanocomposite according to Fertass and co-worker’s report [86]. On the basis of G/CuO nanocomposite, Al/G/CuO (Al: G/CuO = 82.18: 17.82) composite was also obtained by physical mixing of aluminum powder and G/CuO nanocomposite. SEM images show that some CuO nanocomposites are decorated on the surface of graphene nanosheets, while others are wrapped within the graphene nanosheets. For Al/G/CuO composite, the whole surface of aluminum powder is covered by G/CuO nanocomposites. In the presence of CuO, G/CuO and Al/G/CuO additives, the LTD and HTD peaks of AP all merged into one decomposition peak, which is consistent with the observed result in TG curves. The high decomposition temperature of AP blended with G, CuO, G/CuO and Al/G/CuO declined from 432 °C to 400 °C, 350 °C, 325 °C and 315 °C, respectively. Meanwhile, the activation energy are decreased from 129 kJ/mol to 123.41 kJ/mol, 85.12 kJ/mol, 71.47 kJ/mol and 56.18 kJ/mol, respectively. The order of catalytic performance for AP thermal decomposition is ranked as Al/G/CuO > G/CuO > CuO > G. The enhancement of AP decomposition is connected with the inherent characteristics of nano additives. Graphene nanosheets present large surface area and high electron transfer, which can accelerate the decomposition of AP. As a transition metal oxide, the d-orbitals of Cu2+ cations are partially filled, which can accept electrons generated from AP ions, thus, the electron mobility is promoted and the thermal decomposition of AP is accelerated. The as-prepared G/CuO presents higher catalytic activity than that of pure CuO, which can be attributed to the increased dispersity of CuO nanoplates in graphene nanosheets and more exposed active sites. The substrate of highly conductive graphene decreases the aggregation of CuO, whereas the highly active surface area of graphene remarkably improves the catalytic activity of G/CuO. The catalytic performance of Al/G/CuO is better than G/CuO, indicating the aluminum powder can increase the catalytic activity of G/CuO. The aluminum powder can improve the heat transfer and therefore enhance the chemical reaction process. Moreover, the Al/G/CuO composite can provide a large number of active sites to absorb the gases generated from the initial decomposition process of AP, sequentially, the second decomposition process of AP can be accelerated. Hence, Al/G/CuO shows the best catalytic activity among these additives.In addition to graphene, nitrogen-doped graphene is also attractive in the field of catalyst due to the combination of the three dimensional frameworks and the prominent performance of graphene.Hosseini et al. [87] reported a promising catalyst for AP decomposition, which contains CuO nanoparticles and nitrogen-doped graphene. CuO nanoparticles are uniformly distributed and directly decorated on three dimensional graphene-based frameworks (3D-GFs) with particle size around 20–30 nm. The catalytic properties of as-obtained CuO@3D-(N)GFs nanocomposite are related to its specific surface area. By using nitrogen adsorption-desorption analysis, the value of specific surface area for CuO@3D-(N)GFs nanoparticles is calculated to be 124.6 m2/g, while for CuO, the value is 15.6 m2/g. When 4% 3D-(N)GFs are employed, there is only slight effect on the thermal decomposition of AP. Whereas, with the addition of 1, 2 and 4% CuO@3D-(N)GFs, remarkably decrease of HTD temperatures of AP can be observed, which may be attributed to large specific area and more exposed active sites of CuO nanoparticles. Owing to the synergistic effect between 3D-(N)GFs and CuO, the enhanced exothermic heat for AP mixed with CuO@3D-(N)GFs is significantly improved compared with pure AP. The catalytic mechanisms are proposed according to electron transfer theory and proton transfer theory, respectively. On the basis of electron transfer theory, 3D-(N)GFs could provide accelerated electrons to promote the electrons transfer from ClO- 4 to NH+ 4 and the generation of superoxide (O2 -) from oxygen (O2). Besides, the positive hole provided by partially filled 3d orbit in Cu2+ can act as electron acceptor to decompose AP. Under the combined action of 3D-(N)GFs and CuO, CuO@3D-(N)GFs present excellent catalytic activity. When referring to proton transfer theory (Fig. 6 ), proton transfer happens between NH4 + and ClO4 -, the superoxide ions (O2 -) generated from AP decomposition or located on the surface of CuO nanoparticles can capture protons during the process [88]. As depicted in Fig. 6, the advantageous performance of 3D-(N)GFs, including large specific surface area and high thermal conductivity can facilitate the proton transfer from NH4 + to ClO4 - and adsorb more intermediate gas of HClO4 and NH3. As the temperature goes up, the adsorbed NH3 and HClO4 will desorb and react with each other in the gas phase. Moreover, the graphene based substrate can also participate in combustion reaction with HClO4, which will produce more CO2 and more exothermic heat will be produced. Furthermore, the 3D-(N)GFs as a substrate can inhibit the aggregation of CuO NPs, resulting in an increase of specific surface area and more active sites, which will further promote the catalytic process.Ni–Mn bimetallic nanoparticles decorated on three dimensional nitrogen-doped graphene-based frameworks by chemical co-reduction method has been also reported by Hosseini and his group [89]. They studied the catalytic performance of molar ratio of Ni: Mn, the weight ratio of Ni1Mn2@3D-(N)GFs, 3D-(N)GFs support and synergistic effect of Ni and Mn metals on thermal decomposition of AP in detail. The molar ratio of Ni and Mn contained in NiMn@3D-(N)GFs nanocomposites were 2:1, 1:1 and 1:2, respectively. The results indicate Ni and Mn with molar ratio of 1: 2 in NiMn@3D-(N)GFs nanocomposites present the best catalytic activity. The effect of the different weight ratio (3, 5, and 7 wt%) of Ni1Mn2@3D-(N)GFs nanocomposites toward AP decomposition were also studied. When 3 wt% of Ni1Mn2@3D-(N)GFs was employed, the LTD and HTD peaks shift downwardly from 389 to 430 °C to 281 and 335.14 °C, respectively, and the heat release increase from 509 J/g to 1411.78 J/g. With the addition of 5 and 7 wt% of Ni1Mn2@3D-(N)GFs nanocomposites, the LTD and HTD exothermic peaks were combined into one peak, which are centered at 329.43 and 287 °C, respectively. The overall heat release estimated for samples with 5 and 7 wt% additives were 1744.92 and 1331.17 J/g. The above results indicate the samples of 5 and 7 wt% show better catalytic performance than samples of 3 wt%. Compared with the catalytic effect of Ni1Mn2 NPs and Ni1Mn2@3D-(N)GFs nanocomposite, three-dimensional nitrogen-doped graphene act as an efficient support in improving the catalytic performance, which may be caused by the synergistic effect between 3D-(N)GFs and Ni1Mn2. Moreover, synergistic effect also exists in Ni and Mn metals. With the addition of Ni@3D-(N)GFs and Mn@3D-(N)GFs nanocomposites, there are two exothermic peaks ascribing to LTD and HTD process, respectively. When adding Ni1Mn2@3D-(N)GFs nanocomposites, there is a single exothermic peak, indicating the synergistic effect between two metals improves the catalytic performance.To date, two dimension graphitic carbon nitride (g-C3N4) as a narrow band gap semiconductor has been extensively concerned owning to its unique physical and chemical performance, such as high nitrogen content, excellence chemical and thermal stability, controllable electronic structure and eco-friendly. All these characteristics make g-C3N4 a prospective candidate for catalyst and catalytic substrate [90–92].Li et al. [93] reported g-C3N4 as an efficient and eco-friendly catalyst for thermal decomposition of AP by calcining the dicyandiamide. Bulk g-C3N4 displays 2D layered structures, which consists of several graphitic stacking layers. When g-C3N4 was introduced, the LTD and HTD process of AP were combined into a sole procedure with the exothermic temperature ranging from 384.4 to 390.1 °C. The result shows that g-C3N4 can accelerate thermal decomposition rate of AP. In the presence of 10 wt% g-C3N4, the decomposition temperature and activation energy (Ea) of AP are reduced by 70 °C and 119.8 kJ/mol, respectively. With the catalytic effect of 10 wt% g-C3N4, the exothermic heat of AP has a remarkable increase and the value can reach up to 1362.6 J/g, which is much higher than pure AP. The instinct of g-C3N4 is made up of triazine units linked by planar amino groups, which can be regarded as a Lewis base. Lewis acid-base interaction will be formed when HClO4 is absorbed on the surface of g-C3N4. The activation energy of AP decomposition can be decreased by the Lewis acid-base interaction, resulting in the enhancement of AP decomposition. Moreover, g-C3N4 is a kind of polymer semiconductor with a band gap and conduction band potential at 2.7 eV and −1.3 eV vs. RHE, respectively, which can be easily stimulated by external heat. When the energy of external heat surpasses the band gap energy, g-C3N4 will be excited to generate conduction-band electrons (e−) and valence band holes (h+) on the surface. In the decomposition process, HClO4 could be reduced by the conduction-band electrons to create a superoxide radical anion •O2−. Meanwhile, •O2− and h+ would further oxidize NH3 to produce H2O, NO2 and N2O. Thus, g-C3N4 presents catalytic activity on thermal decomposition of AP.On this basis of bare g-C3N4, Li also [94] successfully fabricated SnO2/g-C3N4 hybrids via one-pot calcining method. The catalytic results indicate SnO2 NPs/g-C3N4 hybrids display the best catalytic activity compared with SnO2 and g-C3N4, which may be ascribed to the synergistic effect between SnO2 NPs and g-C3N4. As stated above, e− and h+ could be formed on the surface of g-C3N4 under heat irradiation. Based on the synergistic effect of SnO2 NPs, the generated electrons on g-C3N4 would transfer to SnO2 (Fig. 7 ), thus increasing the separation efficiency and stabilization of the electron-hole pairs. Therefore, the synergistic effect of SnO2 NPs and g-C3N4 lead to the best catalytic performance among all the counterparts.Tan et al. [95] reported a direct precipitation method to prepare (g-C3N4/CuO) nanocomposites. The well-dispersed CuO nanorods with length of 200–300 nm and diameter around 5–10 nm were directly anchored on g-C3N4 by the ion-dipole interaction between cupric ions and long pair electrons on the nitrogen atoms of g-C3N4. In the presence of different catalysts, including pure g-C3N4, CuO, and g-C3N4/CuO (various content of CuO from 5 to 50 wt % versus g-C3N4/CuO), the catalytic performance is varied. The catalytic activities are ranked in ascending sequence of g-C3N4<CuO < g-C3N4/CuO, which imply the existence of synergistic effect between g-C3N4 and CuO nanorods.Rice-shaped manganese dioxides (MnO2) nanoparticles with width of around 5–10 nm and 10–30 nm in the length were successfully anchored on the surface of carbon nanotubes (CNTs) using redox reaction between potassium permanganate and CNTs according to Ling’s research [96]. The catalytic performance indicates the activity of rice-shaped MnO2/CNTs composite are superior compared with that of the mixture for MnO2 nanorods and CNTs, indicating the well distributed rice-shaped MnO2 with a larger surface area can expose more catalytic active sites to promote the thermal decomposition reaction of AP.Cui et al. [97] prepared CNTs/CuO composites by a complex-precipitation method. They also made a comparison between CNTs/CuO composites and the mixture of CuO-CNTs. With the addition of the mixture of CuO-CNTs, the high decomposition temperature decreases by 135 °C, while CNTs/CuO composites were employed, the high decomposition temperature decreases by 145 °C, demonstrating that the catalytic performance can be enhanced when CuO nanoparticles are deposited on CNTs.Carbon-coated metal nanoparticles possess distinct performance, such as core-shell structure, large specific surface area, good adsorption capacity, excellent conduction and heat transfer, which makes it a promising catalyst for the thermal decomposition of AP. An et al. [98] fabricated carbon-coated copper nanoparticles (CCNPs) via a detonating method, by which the mixture of microcrystalline wax, RDX, and copper nitrate hydrate were initiated through an electric detonator under nitrogen atmosphere and normal pressure. The as-obtained detonation products were spherical morphology with particle size ranging from 25 nm to 40 nm. The nanoparticles consist of a darker copper nanocrystal core and a lighter carbonous shell (3–5 nm thickness). With the catalytic effect of CCNPs, the peak at HTD stage decreases from 424.07 °C to 327.08 °C and the activation energy of AP are decreased by 25%, indicating CCNPs can accelerate the decomposition of AP under thermal stimulus.It is known that catalytic performance is related to conduction-band electrons from the irradiated ZnO. However, the electrons and holes tend to reunite, which will decrease the catalytic activity. Wang and co-workers decorated ZnO on carbon black (CB) via atomic layer deposition (ALD) method [99]. When ZnO nanoparticles are decorated on carbon black (CB), the electrons in conduction band will migrate to CB, thus, the recombination of electrons and holes will be prevented, which can further improve the catalytic performance. The ALD ZnO nanoparticles with mean particle size of 10 nm are deposited on the surface of carbon black through stable C–O–Zn bond, which were formed between oxygen-containing functional group on carbon black and Zn2+ ion. The catalytic effect of ZnO/CB hybrid was performed and the results indicate ZnO/CB hybrid present outstanding catalytic activity for the thermal decomposition of AP. The exothermic peak of AP catalyzed by ZnO/CB hybrids is located at 295 °C, which is lower than that of bare ZnO (311 °C). Moreover, the exothermic heat of AP catalyzed by ZnO/CB hybrids is increased from 376 to 1692 J/g, indicating ZnO/CB hybrid possess admirable catalytic performance. This phenomenon is caused by the appearance of C–O–Zn bond, which can promote electron transport from irradiated ZnO to CB, leading to high utilization rate of the electron than bare ZnO. Thus, the catalytic performance of ZnO/CB hybrids is enhanced.In the last few decades, numerous researches have been concentrated on the decomposition mechanisms of ammonium perchlorate and a number of possible decomposition mechanisms of AP have been proposed, but the decomposition mechanism remains a debatable issue. In this review, two possible mechanisms are briefly summarized.Bicromshaw and Newman proposed electron transfer mechanism in 1955 [100]. According to this theory, electrons transfer from ClO- 4 to NH+ 4, the decomposition procure occurs: ClO4 - + NH4 + →ClO4 + NH4 After harvesting an electron, decomposition of the ammonium radicals into ammonia and hydrogen atom takes place: NH4 → NH3 + H Hydrogen atom transfers over the lattice. Electron moves accurately in the same way over the anion sublattice: ClO4 + ClO4 - →ClO4 - + ClO4 HClO4 is generated owing to the reaction between ClO4 radical and H. The resultant HClO4 may continue reacting with H, hence: HClO4 +H → H2O + ClO3 The ClO3 radical acts as acceptors of electrons which can capture electrons. After harvesting an electron, the ClO3 radical is transformed into ClO3 − ion. Then, chlorite ion and ClO4 radical are decomposed and the products can react with NH+ 4 ions. Accordingly, secondary products including chlorine, nitrogen hemioxide and water are produced.In our previous study [101], we investigated the decomposition mechanism of AP catalyzed by CoFe2O4/RGO hybrids on the basis of electron transfer mechanism (Fig. 8 ). The catalytic performance of CoFe2O4/RGO indicates the hybrids can promote both the LTD and HTD process. The lattice of AP is usually assigned to the pair of ions (NH4 + and ClO4 -). As a controlling step, electrons transfer from ClO4 - to NH4 + 4 in LTD process. Meanwhile, oxygen (O2) would transform into superoxide (O2 -) in the HTD process. The partially filled 3d orbits in Fe or Co atom are beneficial for electron transferring. Moreover, positive hole in Fe or Co atom acts as electron acceptor for AP ion and its intermediate products, which can enhance the thermal decomposition of AP. Owing to the unique structure and excellent performance of graphene, the movement of an electron in graphene is fairly faster than in metal atoms and travel much longer distance without being scattered. CoFe2O4/RGO hybrids could provide accelerated electron flow to promote the controlling step. With the efforts of more active and accelerated electron flow, NH4 + and ClO4 - 4 can be easily transformed to NH3 and HClO4. Furthermore, HClO4 decompose into O2, which is subsequently transformed into superoxide (O2 -). The superoxide could help NH3 decompose into N2O, O2, Cl2, H2O and a little NO, completely.Jackobs [102] put forward proton transfer mechanism of AP thermal decomposition and this mechanism can be described as follows: NH4ClO4 ↔ NH4 +(a) + ClO4 -(a) ↔NH3 (g) + HClO4 (g) This mechanism comprises three steps. In step I, the pair of ions (NH4 + and ClO4 -) are involved in NH4ClO4 lattice. Step II involves proton transfer from the cation NH4 + to the anion ClO4 - via a molecular complex. In Step III, the molecular complex breaks down into ammonia and perchloric acid. The dissociation products of AP, NH3 and HClO4 molecules, either react in the absorbed layer over the surface of AP or interact by desorption and sublimation in the gas phase. The gaseous phase of NH3 and HClO4 react quickly, generating O2, N2O, Cl2, NO, and H2O as accessory products at a low temperature (<350 °C).During the reaction happening in the absorbed layer, the perchloric acid desorbed more quickly compared to ammonia, hence, resulting in incomplete oxidation of ammonia and forming a saturated atmosphere of NH3. As a result, the reaction in HTD process will decelerate and transform incompletely, thus forming NO, O2, Cl2, and H2O in the second exothermic decomposition procedure. Both LTD and HTD begin with proton transfer from NH4 + to ClO4 -. The difference between LTD and HTD is that the reaction in a low temperature takes place on the defects of AP crystal, whereas the slow reaction in a high temperature occurs in the lattice of the unreacted normal crystal. Ammonia and chloric acid will be absorbed and desorbed in HTD. Thus, proton transfer in the primary process plays an important role. When additives are introduced, the concentration of the protons will be changed, which will further affect ammonia.Zhang et al. [103] investigated the thermal decomposition behaviors of AP catalyzed m-g-C3N4/CuO and explained the reaction mechanism on the basis of proton transfer mechanism [98]. A solid-gas heterogeneous reaction happens in the LTD process, during which protons transfer from NH4 + to ClO4 -. Subsequently, NH3 and HClO4 are formed and then HClO4 will oxidize NH3 in the gas phase. In the primary stages, the adsorbed HClO4 located in the surface and pores of AP lattice act as a crucial chain carrier for AP decomposition, which can further accelerate the decomposition of AP. Besides, m-g-C3N4 can be classified into Lewis base, owing to its distinct morphologies and surface features. According to the Lewis acid-base theory, HClO4 absorbed on the surface of g-C3N4 would possess decreased activation energy to enhance the thermal decomposition of AP.As a semiconductor, the band gap for m-g-C3N4 is about 2.70 eV, which is beneficial for thermal decomposition of AP. When the excitation energy surpasses the banding gap energy, the valence band holes (h+) and conduction-band electrons (e−) can be formed on the surface of m-g-C3N4. The unique properties of m-g-C3N4, such as large specific surface area, high separation efficiency of electrons and pores, are beneficial for adsorbing HClO4 and NH3 on the surface. The gaseous HClO4 react with e− to form super-oxide radical anion (O2−), which could further react with h+ and NH3 to generate N2O, H2O and NO2. On the other hand, the low banding gap (1.68 eV) and band potential (0.46 eV) of CuO make it to be easily activated by heating. The excited electrons on CB of m-g-C3N4 could transfer to the CB of CuO and the h+ on the VB of CuO could transfer to the VB of m-g-C3N4, which inhibited the recombination of charge carrier. The high separation efficiency of e− and h+ improves the catalytic activity (Fig. 9 ).Although many efforts have been made to investigate the mechanism of thermal decomposition of AP catalyzed by different catalysts, the primary thermal decomposition mechanism is still not fully understood, so, there is still a long way to further investigate the mechanism by various technologies.In recent five years, there has been comprehensive research on preparation, modifications, characterization and performance for various kinds of catalysts in the thermal decomposition of AP. The easily tunable morphology of transition metal oxide and the unique structure of carbon-supported materials, as well as the excellent catalytic performance, makes transition metal oxide and carbon-supported transition metal oxide materials a promising candidate for AP decomposition. This review gives a summary of general strategies and recent process in developing novel and high efficient catalyst in thermal decomposition of AP, including transition metal oxide and carbon-supported transition metal oxide. The as-stated materials in this review present tunable catalytic performance by varying metal elements, adjusting morphology and compounding with carbon-supported analogue. This progress has illustrated that transition metal oxide and carbon-supported transition metal oxide are playing and will continue to play a significant part in modifying the decomposition behavior of AP.Although a remarkable improvement has been made for the development of high catalytic activity of transition metal oxide and carbon-supported transition metal oxide, there are still considerable challenges for further investigation. Most research works focus on the decomposition behavior of AP mixed with versatile catalysts, however, few works are concentrated on the combustion behavior of solid composite propellants when the novel nano-sized catalysts are employed. The combustion process of solid composite propellants is complicated and the internal ballistic property of solid motor related with the mechanisms is still not fully understood. Moreover, batch manufacturing of nano-sized catalysts have not been implemented, which hinders their practical application. So, it’s necessary to do some research about the catalytic effect on solid composite propellants in practical formulation and develop manufacturing technology for large scale production of catalysts. Motivated by the increasing demand of solid composite propellants coupled with critical performance, it can be envisioned that the utilization of novel nano-sized catalysts will be increased in the near future.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 financially supported by the Science and Technology project of Jiangsu province (BN2015021, XZ-SZ201819).

As a main oxidizer in solid composite propellants, ammonium perchlorate (AP) plays an important role because its thermal decomposition behavior has a direct influence on the characteristic of solid composite propellants. To improve the performance of solid composite propellant, it is necessary to take measures to modify the thermal decomposition behavior of AP. In recent years, transition metal oxides and carbon-supported transition metal oxides have drawn considerable attention due to their extraordinary catalytic activity. In this review, we highlight strategies to enhance the thermal decomposition of AP by tuning morphology, varying the types of metal ion, and coupling with carbon analogue. The enhanced catalytic performance can be ascribed to synergistic effect, increased surface area, more exposed active sites, and accelerated electron transportation and so on. The mechanism of AP decomposition mixed with catalyst has also been briefly summarized. Finally, a conclusive outlook and possible research directions are suggested to address challenges such as lacking practical application in actual formulation of solid composite propellant and batch manufacturing.

Concentration of the component i (mmol cm−3)Diffusion coefficient (cm2 s−1)Activation energy named i (J mmol−1)Molar flow at the reactor inlet of the component i (mmol min−1)Molar flow at the reactor outlet of the component i (mmol min−1)Kinetics constant number i (see reference [26])Preexponential factor number i (J mmol−1)CO2 hydrogenation equilibrium constant (atm−2)Adjust parameter (dimensionless)Reaction rate of the component i (mmol g−1 s−1)Average formation rate of CH4 (mmol g−1 s−1)Ideal gas constant (J K−1 mmol−1)Time (s)Time of the storage step (s)Time of the hydrogenation step (s)Temperature (K)Axial position (cm)CO2 conversion (%)H2 conversion (%)CH4 production (μmol g−1)Correction factor (dimensionless)Porosity (dimensionless)Covering factor of the component i (dimensionless)Density (g cm−3)Adsorption capacity (mmol g−1)Maximum adsorption capacity (mmol g−1)Dual Function MaterialPartial Differential EquationSynthetic Natural GasThe CO2 methanation reaction (Eq. 1), also known as Sabatier reaction, originates in 1902 [1]. The scientific interest on the Sabatier’s reaction has grown in recent years in the context of a massive implementation of renewable energies. One of the main drawbacks of renewable energies is their intermittent nature due to their dependence on atmospheric conditions. In an energy system based on renewable energies, periods of energy shortage or surplus can occur. Thus, the storage of energy to balance the energy demand is essential. In this context, the CO2 methanation has a practical application. In periods of energy surplus, the electric energy produced by renewables energies would be used to produce hydrogen by electrolysis, which then reacts catalytically with CO2 (captured from an industrial effluent) to produce CH4 named as synthetic natural gas (SNG). This process is also known as Power to Gas (PtG) technology, which aims to connect the electric grid and the gas grid to make the future energy system more robust [2]. (1) CO2+4H2 ⇆ CH4+2H2O Sabatier’s reaction is characterized by being strongly exothermic, and therefore equilibrium is favored at low temperatures. On the other hand, the number of moles of products is less than that of reagents, so the thermodynamic equilibrium is favored at high pressures. However, working at high pressure implies a high economic cost, so it is more convenient to work at atmospheric pressure [3]. In addition, it should be noted that the complete reduction of CO2 (oxidation state C+4) to methane (oxidation state C−4) implies the transfer of eight electrons, which implies to overcome a high kinetic barrier. Therefore, the use of catalysts is essential [4].In general, the catalysts used for this reaction must have thermal stability in the operating temperature range of 200–400 °C. Catalysts with different active phases (Ru, Ni, Fe, Co.) [5–9], and different supports (Al2O3, zeolites, SiO2, TiO2.) [10–14] have been used in recent years. Among the active phases, nickel shows high CO2 conversion and is one of the most used metals due to its abundance and low cost. However, nickel tends to sinter, and therefore, deactivate. On the other hand, ruthenium is very active, selective and stable towards methane formation even at low temperatures. Although Ru is more expensive than Ni, it has been reported in a large number of publications [9,15]. On the other hand, the support can influence the dispersion of the active phases, its reducibility and the formation of spinels that can reduce the activity of the catalyst [16]. Generally, basic mesoporous solids are used; in particular, alumina has been the most used support to disperse the active phase.In the early 2020 there were 38 methanation plants with a total capacity of 14.5 MW and that number is growing exponentially [17]. One of the main drawbacks of PtG technology is the high costs associated with the CO2 purification. One cost effective alternative is the utilization of a dual-function material (DFM) as catalyst. The DFM contains an alkaline or alkaline earth element that acts as an adsorbent and a noble metal that assists the methanation reaction [18,19]. The DFM allows the capture of CO2 and its direct conversion to methane, without the need of intermediate CO2 sequestration processes, which are energy intensive. The operation is carried out cyclically alternating steps of CO2 storage and hydrogenation. This novel operative strategy has recently been proposed by Duyar et al. [20]. CO2 is first captured onto the basic element of the catalyst until saturation. Afterwards, H2 is injected and favors a spillover phenomenon that conducts the chemisorbed CO2 to the metal site where the methanation takes place. Both the CO2 capture process and the CH4 production process can operate at a temperature of 250–400 °C. An effluent of a combustion process can easily reach this temperature. Thus, the PtG technology using DFMs can be directly applied, without the need for an external heat input [21,22].This novel cyclic operation strategy is very promising. However, it still needs further development for industrial implementation. Advances in the formulation and physyco-chemical properties of DFMs are required to boost the adsorption capacity and hydrogenation activity of the samples. In this sense, an intimate contact between the adsorbent and the metal is crucial [22]. On the other hand, the influence of the operational variables on the CO2 adsorption and hydrogenation performance has to be addressed. This experimental work is usually a very time consuming step if the number of studied variables is so large to cover a wide range study. One possibility to predict the influence of operational variables on the catalytic behavior is simulation. For that, it is required to build first a robust model able to predict accurately the evolution of reactants and products under a wide range of operational conditions.In own previous work [23–25], we reported a complete reaction scheme able to describe the CO2 adsorption and hydrogenation using DFMs with formulation x-Na2CO3/Al2O3 (x = Ru/Ni). Briefly, CO2 and H2O compete for the adsorption sites (Na2O), forming the corresponding carbonate (Na2CO3) or hydroxide (NaOH), respectively. During the adsorption step, a CO2 molecule can displace a previously adsorbed H2O molecule, forming the carbonate and releasing H2O to the gas phase. During the hydrogenation step, the as-formed carbonates are decomposed and hydrogenated on the metal site producing CH4 and H2O. Some fraction of the as-formed H2O is adsorbed onto the storage sites forming the hydroxide.Based on the elemental reactions that govern the process, we proposed a kinetic model, which accurately predicts the evolution of CO2, CH4 and H2O [26]. The kinetic equations of the model rely on the concentration of reactants and products and on the surface coverage of CO2 and H2O. The model was validated in a wide range of reactants concentrations, i.e. 1.4–10.9% CO2 during the adsorption step and 1.4–10.9% H2 during the hydrogenation step, and in the 250–400 °C temperature range.In this work, we shall use the previously developed and validated model to optimize by simulation modeling the duration of the CO2 storage step and the duration of the hydrogenation step, i.e. the CO2 adsorption and hydrogenation cycles timing (t CO2 /t H2). First, different simulations are performed with different t CO2 /t H2 to qualitatively observe how the pair of times influences on the temporal evolution of CO2, CH4 and H2O. The surface coverages of CO2 and H2O are also analyzed at this point. Then, catalytic parameters are defined with which the global performance of the catalyst can be evaluated at any given t CO2 /t H2. The conversion of CO2 and H2, CH4 yield and the average CH4 formation rate are analyzed in order to define an optimum CO2 adsorption and hydrogenation cycle timing (t CO2 /t H2). Based on the optimal cycle timing, a reactor configuration is proposed for the industrial application.A dual function material with formulation 4%Ru-10%Na2CO3/Al2O3 was prepared by wet impregnation. A detailed description of the preparation procedure and characterization of the catalyst can be found elsewhere [26]. Reactor tests were performed in a stainless steel tube placed in a vertical furnace. 3 g of pelletized (0.3–0.5 mm) catalyst was housed in the reactor. The catalyst was pre-treated with a gas stream composed of 10% H2/Ar at 350 °C for 45 min to favor the reduction of Ru.The CO2 storage and hydrogenation is carried out with cyclic feeding. During the CO2 storage step, a gas stream composed of 5.7% CO2/Ar was fed for 2.5 min. During the hydrogenation step, a gas stream composed of 5.7% H2/Ar was fed for 5 min. A purging step with argon was fed between adsorption and hydrogenation cycles to avoid mixing of both gas streams. The operation was carried out at 350 °C and the total flowrate was set at 1200 ml min−1, which corresponds to a space velocity of 15,000 h−1. The composition of the gas stream leaving the reactor was analyzed by FTIR (MKS MultiGas 2030) for quantitative determination of CO2, CH4 and H2O concentration.The CO2 adsorption capacity (Ω) during the storage period is calculated by Eq. (2). CH4 and H2O productions are calculated by Eqs. (3) and (4), respectively. (2) Ω mmol g − 1 = 1 W ∫ 0 t CO 2 F CO 2 in t − F CO 2 out t d t (3) Y CH 4 mmol g − 1 = 1 W ∫ 0 t H 2 F CH 4 out t d t (4) Y H 2 O ( mmol g − 1 ) = 1 W ∫ 0 t H 2 F H 2 O out ( t ) d t t CO2 and t H2 correspond to the duration of the CO2 storage and hydrogenation periods, respectively. F CO 2 in and F CO 2 out correspond to the CO2 molar flow at the reactor inlet and outlet streams, respectively. F CH 4 out and F H 2 O out are the molar flows of CH4 and H2O at the reactor outlet stream, respectively. W is the weight of the catalyst housed in the reactor.Two additional parameters will be used to evaluate the catalytic performance, i.e. the conversion of CO2 and the conversion of H2 during the hydrogenation period. (5) X CO 2 = ∫ 0 t CO 2 [ F CO 2 in ( t ) − F CO 2 out ( t ) ] d t ∫ 0 t CO 2 F CO 2 in ( t ) d t ⋅ 100 = ∫ 0 t H 2 F CH 4 out ( t ) d t ∫ 0 t CO 2 F CO 2 in ( t ) d t ⋅ 100 (6) X H 2 = ∫ 0 t H 2 F H 2 in t − F H 2 out t d t ∫ 0 t H 2 F H 2 in t d t · 100 = 4 ∫ 0 t H 2 F CH 4 out t d t ∫ 0 t H 2 F H 2 in t d t · 100 All the CO2 stored reacts to form CH4, i.e. unreacted CO2 is not experimentally observed during the hydrogenation period, as will be seen later. Thus, the amount of CO2 stored can be also evaluated as the amount of CH4 produced during the hydrogenation step: ∫ 0 t H 2 [ F CO 2 in ( t ) − F CO 2 out ( t ) ] d t = ∫ 0 t H 2 F CH 4 out ( t ) d t , provided that Sabatier´s reaction stoichiometry states 1 mol CO2:1 mol CH4. On the other hand, H2 conversion can be calculated based on methane formation. For that, it is considered that hydrogen consumption quadruples methane formation, following again the stoichiometry of the Sabatier´s reaction (Eq. 1).Dynamic one dimensional isothermal heterogeneous plug flow reactor model with axial dispersion is considered for the modeling of the CO2 capture and hydrogenation. The evolution of the concentration of CO2, CH4 and H2O is calculated by solving jointly the partial differential equation (PDE) for the gas phase (Eq. 7) and the ordinary differential equation (ODE) for the solid phase (Eq. 8). (7) Gas phase : ∂ C i ∂ t = − u ε ∂ C i ∂ x + D ε ∂ 2 C i ∂ x 2 + ρ r i ε (8) Adsorbent phase : ∂ θ j ∂ t = R i Ω max where ε is the void fraction, C i the concentration of the gas phase of species i, θ j is the surface coverage of species j, u the linear velocity of the gas, D the diffusion coefficient, ρ the density of the bed, x the axial coordinate of the reactor, Ωmax the maximum CO2 adsorption capacity of the catalyst, and R i the rate of formation of species i, calculated according to: (9) R i = ∑ k = 1 r k υ i , k where i is the index of the species considered, r k the intrinsic velocity of reaction k and ν i,k the stoichiometric coefficient of species i in reaction k.We reported in a previous work [24] the mechanism of the CO2 storage and hydrogenation to CH4 using a dual function material that operates in alternate cycles. The main reactions occurring during adsorption period are: (10) Na2O+CO2 ⇆ Na2CO3 (11) 2NaOH+CO2 ⇆ Na2CO3+H2O (12) NaOH+CO2 ⇆ NaHCO3 and during methanation period: (13) Na2CO3 ⇆ Na2O+CO2 (1) CO2+4H2 ⇆ CH4+2H2O (14) Na2O +H2O ⇆ 2NaOH Check out reference [24] for more details regarding the CO2 adsorption and hydrogenation mechanism. The kinetic equations used by the model are based on those reactions. A detailed discussion about the kinetic expressions adopted by the model can be also found in our previous work [26]. The kinetic expressions used to predict the reaction rates of CO2, CH4 and H2O during the adsorption, purge and hydrogenation period are collected in Table 1.The temporal evolution of the concentration of CO2, CH4 and H2O predicted by the model was obtained by integrating the mass balance for the gas phase (Eq. 7) and for the solid phase (Eq. 8). The model considers that CO2 and H2O can be adsorbed onto the storage sites (Na2O) leading to the formation of carbonates (Na2CO3) and hydroxides (NaOH). The presence of carbonates and hydroxides in the surface of catalysts containing a basic element, such as Na2O or CaO, has been confirmed by FTIR when exposed to gas phase CO2 or H2O [27]. The kinetic expressions adopted in the model for the estimation of the reaction rates rely on the surface coverage of CO2 (θ CO2) and H2O (θ H2O). The surface coverage of CO2 is defined as the amount of CO2 adsorbed in the storage sites (Ω) with respect to the maximum CO2 storage capacity (Ωmax). Thus, if the catalyst is saturated with CO2, the surface coverage of CO2 would be 1 (θ CO2=1). On the contrary, if the catalyst is fully regenerated, the surface coverage of CO2 would be 0 (θ CO2=0). Depending on the state of the catalyst, the covering factor takes values comprised between 0 ≤ θ CO2 ≤ 1. The surface coverage of H2O is defined as the amount of H2O adsorbed in the storage sites with respect to the maximum CO2 storage capacity (Ωmax). As observed for θ CO2, depending on the state of the catalyst, the covering factor of H2O takes values comprised between 0 ≤ θ H2O ≤ 1. Taking into account the adsorption stoichiometry of CO2 and H2O (Eq. 10 and Eq. 14) one molecule of CO2 or one molecule of H2O is adsorbed onto one molecule of the storage site (Na2O). As CO2 and H2O compete for the same adsorption sites, at any time 0 ≤ θ CO2 + θ H2O ≤ 1. Besides, the model also considers the formation of bicarbonate type species (Eq. 12) when CO2 and H2O coexist in the gas phase. Note that formation of bicarbonates means that one additional molecule of CO2 and H2O are adsorbed onto and already carbonated adsorption site (Na2CO3). A global reaction scheme for bicarbonates formation was proposed in our previous work [26], in which CO2 is delivered by a neighborhood adsorption site and H2O is adsorbed from the gas phase. Thus, the formation of bicarbonates is considered as an unstable reservoir for the storage of H2O without implication in the storage of CO2. The surface coverage of bicarbonates is defined as θ H2O/CO2. In the presence of bicarbonates, the sum of the surface coverages of CO2 (θ CO2), H2O (θ H2O) and (θ H2O/CO2) could exceed 1, as a H2O molecule is adsorbed onto an already carbonated site.As defined in the previous paragraph, the surface coverages of CO2 (θ CO2), H2O (θ H2O) and (θ H2O/CO2) are all defined as the amount of adsorbed specie divided by the maximum CO2 storage capacity (Ωmax). In order to experimentally calculate Ωmax, CO2 adsorption and hydrogenation cycles are carried out provided that the regeneration period is extended until complete regeneration of the catalyst. We consider full regeneration of the catalyst when carbon containing products (CH4) are not observed at the reactor outlet stream, i.e. concentration is below 5 ppm. The subsequent CO2 adsorption period is extended until the catalyst is saturated with CO2. Under this experimental conditions, i.e. full regeneration during the hydrogenation period and full saturation during the adsorption period, the maximum CO2 storage capacity (Ωmax) is calculated by Eq. (2).Due to the cyclic nature of the operation, which alternates among different feeding compositions during the adsorption or hydrogenation periods, it is important to model how the feed enters the reactor. We showed in our previous work [26], that a first order transfer function was able to describe the evolution of CO2 concentration at the reactor entrance when the CO2 concentration is changed in step mode from 0% to 5.7% and from 5.7% to 0%. A first order transfer function was also applied to model the feeding of hydrogen at the beginning and at the end of the hydrogenation period.To solve the PDE system, the axial coordinate of the reactor was discretized based on finite differences in 19 equidistant elements. Backward and central differences were applied for the evaluation of the first and second derivatives of the concentration with respect to the reactor length, respectively. To solve the system of resulting ordinary differential equations, a program was developed in Matlab.The model has already been validated in our previous work [26] in a wide range of reactants concentration and temperature. Even though, as an example, Fig. S1 shows the temporal evolution of experimental gas phase CO2, H2O and CH4 concentration together with that simulated by the model. The operation is carried out at 350 °C feeding a gas stream composed of 5.7% CO2/Ar during the storage period and 5.7% H2/Ar during the hydrogenation period. The duration of the storage period is 2.5 min and an Ar purge is continued for 2 min. Then the hydrogenation period is extended for 5 min and finally another Ar purge is performed for 1 min before starting a new cycle. As can be observed, the model accurately predicts the experimental evolution of gas phase CO2, H2O and CH4 during the CO2 adsorption and hydrogenation. Besides, the evolution of the covering factors at the reactor exist are also shown.The kinetic parameters that best fit the experimental data were calculated based on the least squares method, using the concentrations of gaseous species (CO2, CH4 and H2O) as experimental responses and the values are collected in the Table S1. Fig. 1 shows the temporal evolution of gas phase CO2, CH4 and H2O predicted by the model when the operation is carried out at 350 °C. A gas stream composed of 5.7% CO2/Ar is fed to the reactor during the CO2 adsorption period, which is extended for 80 s. After the purging period with Ar for 120 s, a gas stream composed of 5.7% H2/Ar is fed to the reactor. The hydrogenation period is extended for 160 s. The covering factors for the last axial coordinate (in which the reactor has been discretized for the integration of the system of ODEs) are also collected in Fig. 1d. The dynamics of the CO2 capture and hydrogenation can be explained based on the main reactions that govern the process [23,24].During the adsorption period, CO2 is stored onto the basic sites of the catalyst (Na2O) in the form of carbonates (Na2CO3) through Eq. (10). Alternatively, CO2 can be also stored onto the hydrated form of the adsorption sites (NaOH) to form the carbonate through Eq. (11). Note that this route implies the release of H2O to the gas phase, i.e. CO2 displaces adsorbed H2O to form the carbonate. Eqs. (10) and (11) are the main reactions describing the CO2 storage process. The CO2 storage onto the catalyst can be evidenced by comparing the CO2 concentration signal at the reactor inlet (dotted line in Fig. 1a) and outlet streams (red line). Note that the CO2 concentration is lower at the reactor outlet with respect to that observed in the inlet stream, which highlights the CO2 adsorption capacity of the catalyst. In fact, the area comprised between the CO2 concentration signal at the reactor inlet and outlet streams can be directly related with the CO2 storage capacity of the catalyst. See Fig. S2 in supplementary material for detailed mathematical procedure to calculate the CO2 storage capacity of the catalyst.At the beginning of the CO2 storage period, the adsorption of CO2 occurs through Eq. (10). Afterwards, once the sodium oxide (Na2O) sites have been completely carbonated, the storage of CO2 can proceed through Eq. (11). Following the reaction stoichiometry, one molecule of H2O is released to the gas phase when one molecule of CO2 is stored. Thus, the storage of CO2 through Eq. (11) can be evidenced by the presence of gas phase H2O. As can be observed in Fig. 1c, water concentration breakthrough is detected after 0.25 min (15 s) of the storage period. Thus, for storage times lower than 15 s, the storage of CO2 proceeds through Eq. (10) (without the release of H2O), and afterwards through Eq. (11), as evidenced by the increase of the H2O concentration. As the storage period proceeds, the covering factor of CO2 (θ CO2) increases and the covering factor of H2O (θ H2O) decreases, as can be observed in Fig. 1d. Eventually, all the storage sites of the catalyst become carbonated. Hence, the concentration of CO2 at the reactor outlet matches that of the inlet (Fig. 1a) and the CO2 covering factor reaches the value of 1 (Fig. 1d). When the catalyst is saturated with CO2 no additional H2O is released to the gas phase and the concentration of H2O progressively decreases (Fig. 1c). When H2O and CO2 coexist in the gas phase, formation of bicarbonates is possible through Eq. (12). Indeed, bicarbonates covering factor (θ CO2/H2O) describes a maximum and then decreases following the decreasing trend observed for H2O concentration. The evolution of CO2 and H2O concentration during the storage period is governed by the kinetic equations of the model.After the storage period, CO2 is removed from the feed stream and the catalyst is purged with Ar for two minutes, observing that the CO2 concentration decreases progressively to practically zero. A slight decrease in θ CO2 is observed (Fig. 1d), due to the desorption of part of CO2 that is weakly adsorbed. This process has been described by Eq. (13) and is modeled by Eq. (18) using a Temkin-type desorption kinetics. At the beginning of the purging period, the covering factor of water (θ H2O) is zero because water has been completely displaced from the adsorption sites due to CO2 adsorption. Meanwhile, bicarbonates decomposition is accelerated by the elimination of gas phase CO2, which reduces their stability. Thus, the covering factor of bicarbonates (θ CO2/H2O, Fig. 1d) is rapidly reduced to zero. The H2O formation rate during the CO2 storage period is also valid for the purging period (Eq. 20).Finally, the hydrogenation period begins admitting 5.7% H2 in the feed. The inclusion of hydrogen provokes the decomposition of adsorbed CO2, which is represented by Eq. (13) and is modeled by the first term of Eq. (21). This reaction pathway can be facilitated by the lower stability of carbonates in the presence of H2 or by a catalytic process involving the spillover of hydrogen ad-atoms to the adsorption sites [28]. In the presence of gas phase CO2 and H2, the Sabatier´s reaction (Eq. 1) proceeds and CH4 and H2O are produced. Thus, just from the beginning of the hydrogenation period CH4 formation is detected. The formation of CH4 is modeled with a potential kinetic equation recently reported by Falbo et al. [29] (Eq. 22). Note that during the whole hydrogenation period, gas phase CO2 is not observed, which highlights that the CO2 methanation rate (r CH4) is higher than the CO2 decomposition rate (expressed by the first term of Eq. 21). As the hydrogenation period proceeds, CH4 formation is observed while the covering factor of CO2 (θ CO2, Fig. 1d) decreases progressively. The progressive diminution of the CO2 covering factor indicates that the adsorption sites of the catalyst are being regenerated. Due to the progressive reduction of the CO2 covering factor, the carbon source to be hydrogenated is reduced, and consequently, CH4 formation progressively decreases in the last section of the hydrogenation period.Water formation is also observed at the outlet of the reactor during the hydrogenation period (Fig. 1c). According to the Sabatier´s reaction (Eq. 1), water formation should double CH4, and should present a similar concentration profile. However, this is not observed in Fig. 1. The reason is that water interacts with the adsorption sites and is adsorbed, as described by Eq. (14). The consequence is that water formation in the gas phase is retarded with respect to CH4. Due to water adsorption on the storage sites, the covering factor of water (θH2O) increases progressively during the hydrogenation period. Water formation during the hydrogenation period is modeled by Eq. (23).Up to now, the mechanism of the CO2 storage and hydrogenation has been presented together with the kinetic equations used to model the operation. Now, we will focus on the influence of the adsorption and hydrogenation periods timing on the dynamics of the dual operation. For that, we will use the temporal evolution of the concentration of CO2, H2, CH4 and H2O as predicted by the model. Fig. 2 shows the evolution of reagents and products concentration, together with the covering factors, for different adsorption and hydrogenation periods timing, i.e. t CO2 and t H2, respectively. The concentration of CO2 and H2 at reactor inlet (dotted line) is also displayed in the corresponding charts. We have selected three scenarios to understand the influence of the adsorption and hydrogenation periods timing. In Fig. 2a we have selected t CO2= 45 s and t H2= 300 s as representative of a short adsorption period and long hydrogenation period. In Fig. 2b we have selected t CO2= 150 s and t H2= 300 s as representative of long adsorption and hydrogenation periods. Finally, in Fig. 2c we have selected t CO2= 150 s and t H2= 60 s as representative of long adsorption period and short hydrogenation period. Depending on the adsorption and hydrogenation periods timing, large differences are observed in the evolution of gas phase CO2, H2, CH4 and H2O concentrations, which we will explain in detail below. Note that the evaluation of the CO2 adsorption and hydrogenation performance has to be done in the whole operation, considering the adsorption and hydrogenation performances. As this is a cyclic operation, alternating consecutive adsorption and hydrogenation periods, the state of the catalyst at the beginning of a given period depends on the state of the catalyst at the end of the previous period. For example, the CO2 adsorption performance will be dependent on the state of the catalyst at the end the previous hydrogenation period. The same is applied for the hydrogenation period, which performance also depends on the state of the catalyst at the end of the previous adsorption period.As can be observed in the upper chart of Fig. 2a, an adsorption period of 45 s does not achieve the saturation of the catalyst with CO2. Note that at the end of the adsorption period, the CO2 concentration at the reactor outlet is notably below the CO2 concentration at the reactor inlet. Besides, the CO2 covering factor depicted in the lower chart shows a value below 1, i.e. θ CO2 = 0.6. At this point, it is important to emphasize that the CO2 adsorption takes place following an adsorption front, which moves forward along the reactor length as the adsorption sites are spent or carbonated. To illustrate the previous statement, Fig. 3a shows the evolution of the CO2 covering factor (θ CO2) along the reactor length during the adsorption period. As can be observed, at the beginning of the adsorption period, the CO2 covering factor is 0 along the reactor length, which highlights that the catalyst has been fully regenerated in the previous regeneration period. Then, as the adsorption period proceeds, CO2 is captured by the catalyst and thus θ CO2 increases. Note that the CO2 is preferentially captured in the reactor entrance, leaving the adsorption sites located downstream empty and available for the CO2 capture. As the adsorption period continues, the covering factor at the reactor entrance gets more and more saturated, and thus, adsorption sites located downstream start to be filled. At the end of the adsorption period, the CO2 covering factor at the reactor entrance is 1 (meaning a complete saturation) but the CO2 covering factor at the reactor outlet is 0.6 (as can be also observed in Fig. 2). Thus, under this operating conditions (t CO2=45 s and t H2=300 s), the catalyst is not fully saturated at the end of the adsorption period.There is another phenomenon that should be studied during the CO2 adsorption period, i.e. the release of water displaced by the CO2 adsorption onto the storage sites to the gas phase. As already reported in the previous section, first CO2 is adsorbed onto the free adsorption sites (Eq. 10), and once those sites are occupied, the storage of CO2 proceeds with the displacement of water (Eq. 11). This is the reason why water detection (Fig. 2c) is retarded with respect to the beginning of the adsorption period. Then, water concentration starts to increase but the adsorption period finishes before reaching the maximum value. Again, we will rely on the evolution of the H2O covering factor along the reactor length during the adsorption period (Fig. 3b) to better understand the state of the catalyst. As can be observed, the H2O covering factor (θ H2O ) is not zero at the beginning of the adsorption period, because some of the adsorption sites are hydrated at the end of the previous hydrogenation period. Note that θ H2O is higher at the reactor outlet due to the dynamics of the regeneration, which will be explained later. As can be observed, θ H2O is hardly affected in the first 15 s of the adsorption, because the adsorption of CO2 is being taken place in the free adsorption sites. Afterwards, θ H2O starts to decrease at the reactor entrance, where the occupation of the adsorption sites by CO2 is higher (Fig. 3a). At the end of the adsorption period, water has been completely removed from the adsorption sites (θ H2O=0) at the reactor entrance but there is still water adsorbed at the rear of the reactor. This is because the CO2 adsorption front does not reach the rear of the reactor and consequently does not displace adsorbed water.During the hydrogenation period, the evolution of H2, CH4 and H2O is also observed (Fig. 2a). As can be observed, H2 concentration at the reactor outlet is, at any time, lower than that fed to the reactor (dotted line). This fact indicates that H2 is being consumed through the Sabatier´s reaction (Eq. 1) to produce CH4 and H2O. CH4 is immediately detected after the beginning of the hydrogenation period. As the hydrogenation period proceeds, the covering factor of CO2 is progressively reduced. Eventually θ CO2 reaches a value near 0 at the end of the regeneration period, which reveals a complete regeneration of the catalyst. In line with the complete regeneration of the catalyst, CH4 concentration is insignificant at the effluent of the reactor at the end of the regeneration period. As already explained in the previous section, water detection in the gas phase is retarded with respect to CH4 because water is adsorbed onto the storage sites. Thus, θ H2O increases with the hydrogenation time. In order to clarify the state of the catalyst during the hydrogenation, we will comment on the evolution of θ CO2 and θ H2O along the reactor length during the hydrogenation period (Fig. 3c and d). The first interesting phenomena to highlight is that the regeneration of the catalyst (decrease θ CO2 ) does not occur following a regeneration front. Opposite to that observed for the CO2 adsorption in Fig. 3a, the decrease of θ CO2 occurs homogeneously along the reactor length. At the end of the regeneration period, the occupation of the adsorption sites by CO2 is insignificant along the reactor length. With respect to θ H2O, it can be observed that the occupation of the storage sites with H2O is null in the whole reactor length at the beginning of the hydrogenation period. Afterwards, θ H2O increases due to the adsorption of water (produced through the Sabatier’s reaction). Water adsorption explains the retard observed in the detection of water in the gas phase with respect to CH4 (Fig. 2a). Water preferentially occupies the positions of the rear of the reactor because water produced at the reactor entrance is adsorbed in subsequent positions of the reactor axial coordinate. At the end of the regeneration period, the occupation of the adsorption sites by H2O is significant, specifically at the rear of the reactor.To sum up, under this operating conditions (t CO2=45 s and t H2=300 s), the catalyst is not fully saturated with CO2 at the end of the adsorption period and some water remains adsorbed, specifically in the adsorption sites located at the rear of the reactor. On the other hand, the catalyst is fully regenerated at the end of the hydrogenation period, i.e. almost no CO2 is adsorbed at any position of the reactor axial coordinate. However, a significant fraction of the storage sites is occupied by H2O.In Fig. 2b we will explain the performance of the catalyst when the adsorption and hydrogenation periods timing is t CO2/t H2= 150/300, representative of long storage and hydrogenation periods. As can be observed, the main difference during the CO2 adsorption period (with respect to a shorter storage period of 45 s, Fig. 2a) is that the CO2 concentration at the reactor outlet matches that of the inlet at the end of the adsorption period. This fact reveals a total saturation of the catalyst with CO2, which can be corroborated by the fact that θ CO2 reaches a value of 1. The evolution of the CO2 covering factor along the reactor length during the adsorption period (Fig. S3a) shows the same trend as in the shorter storage period (Fig. 3a). The unique difference is that a longer storage period of 150 s results in the total saturation of the catalyst with CO2 in the whole reactor length, as opposite to the partial saturation observed with a shorter duration of 45 s. The second difference is that water concentration peak is totally developed. In fact, at the end of the adsorption period, water concentration is negligible after peaking at 1.1% at 1 min of storage time. Besides, the covering factor of water is also zero. The evolution of the H2O covering factor along the reactor length during the adsorption period (Fig. S3b) shows that water is totally removed before the adsorption period is finished. No adsorbed water remains at the catalyst surface irrespective the reactor length.During the hydrogenation period, the performance of the catalyst running with t CO2/t H2= 150/300 s is similar to that shown in Fig. 2a (t CO2/t H2=45/300). The only difference is that with a longer storage time of 150 s the catalyst is fully saturated with CO2, and thus, during the hydrogenation period CH4 and H2O formation is slightly enhanced. Apart from that, the state of the catalyst at the end of the hydrogenation period is almost similar with both timings (t CO2/t H2=45/300 or t CO2/t H2=150/300). A long hydrogenation period of 300 s enables almost a total regeneration of the catalyst, and θ CO2 is almost zero irrespective the reactor length (Fig. S3c). As previously explained, the storage sites of the catalyst are partially occupied with H2O as can be observed in Fig. S3d.To sum up, under this operating conditions (t CO2=150 s and t H2=300 s), the catalyst is completely saturated with CO2 at the end of the adsorption period and no water remains adsorbed irrespective the reactor length. On the other hand, the catalyst is fully regenerated at the end of the hydrogenation period. No CO2 is adsorbed at any position of the reactor axial coordinate but a significant fraction of the storage sites is occupied by H2O.Finally, Fig. 2c shows the performance of the catalyst when the adsorption and hydrogenation periods timing is t CO2/t H2= 150/60, representative of a long storage period and a short hydrogenation period. Due to a shorter regeneration period, the catalyst is not completely regenerated (as will be seen later) and some CO2 remains adsorbed in the storage sites at the beginning of the adsorption period. In fact, the CO2 covering factor for the last axial position of the reactor is 0.5 at the beginning of the adsorption period. Consequently, the CO2 adsorption capacity of the catalyst is limited and the CO2 breakthrough is earlier detected with respect to the previous t CO2/t H2 timings of 45/300 or 150/300. The evolution of the CO2 covering factor along the reactor length during the adsorption period (Fig. S4a) reveals that θ CO2 is around 0.5 irrespective the reactor length. Afterwards, the CO2 adsorption front evolves (as observed in Figs. 3a and S3a). However, as the CO2 adsorption period begins with the catalyst partially occupied by CO2, the saturation is achieved at earlier adsorption times. Fig. 2c shows that the hydrogenation period finishes before CH4 concentration peak is totally developed. This information, together with the fact that θ CO2 is not cero, points out that only a partial regeneration of the catalyst has been achieved. The evolution of the CO2 covering factor along the reactor length (Fig. S4c) shows that 60 s of hydrogenation is not enough to complete the regeneration. All the positions of the reactor show a rather homogeneous occupation of CO2, with a slight tendency to increase θ CO2 with the reactor length. This means that the reactor entrance achieves a slightly higher regeneration. Due to the lower regeneration of the catalyst, less H2O is also produced through the Sabatier´s reaction, and thus, less H2O is adsorbed onto the catalyst (as can be observed in Fig. S4d).To sum up, under this operating conditions (t CO2=150 s and t H2=60 s), the catalyst is not fully regenerated and some CO2 together with H2O remain adsorbed onto the storage sites at the end of the regeneration period. This fact limits the CO2 adsorption capacity of the subsequent storage period and the CO2 breakthrough is earlier detected.Once the dynamics of the dual process have been analysed, now the global performance of the catalyst is evaluated based on the following parameters: CO2 conversion (Eq. 5), H2 conversion (Eq. 6) and CH4 production (Eq. 3). First, we will calculate the catalytic parameters for the CO2 adsorption and hydrogenation timings defined in the previous section. Then, we will extend the analysis for CO2 adsorption periods ranging from 10 to 150 s and hydrogenation periods ranging from 20 to 300 s. The evolution of CO2 conversion (X CO2, %), H2 conversion (X H2, %) and CH4 production (Y CH4, mmol g−1 cycle−1) will be shown as a function of the CO2 adsorption and hydrogenation periods timing in a 3D picture ( Fig. 4).The conversion of CO2 (Eq. 5) relates the percentage of CO2 stored onto the catalyst with respect to the amount of CO2 fed. When the operation is carried out with a CO2 adsorption and hydrogenation periods timing of 45/300 (Section 3.2.1, Fig. 2a) the CO2 conversion results in 57%. This high CO2 conversion is the result of a deep regeneration, which fully regenerates the adsorption sites of the catalyst. This fact enables a high CO2 adsorption performance at the beginning of the adsorption period. Besides, due to the short CO2 storage period, the catalyst does not reach saturation and the amount of CO2 leaving the reactor is limited. The CO2 conversion is significantly reduced to 12% when the CO2 adsorption and hydrogenation periods timing is 150/300 (Section 3.2.2, Fig. 2b). The longer duration of the CO2 storage period, results in the complete saturation of the catalyst. Extending the length of the adsorption period after catalyst saturation penalizes the CO2 conversion, as no CO2 is further adsorbed and all the CO2 fed to the reactor is emitted in the effluent. Finally, the CO2 conversion is further reduced to 4% when the CO2 adsorption and hydrogenation periods timing is 150/60 (Section 3.2.3, Fig. 2c). The short regeneration period does not obtain the full regeneration of the catalyst. Consequently, the catalyst is earlier saturated (the CO2 breakthrough is earlier detected) and the amount of CO2 emitted in the effluent is enhanced. The result is a further reduction of the CO2 conversion. Fig. 4a shows the evolution of the CO2 conversion (X CO2, %) as a function of the CO2 adsorption and hydrogenation periods timing. For a given t CO2, the CO2 conversion increases with t H2 due to a deeper regeneration of the catalyst. For a given t H2, the CO2 conversion decreases with t CO2 due to a higher fraction of CO2 emitted after the saturation of the catalyst. Maximum CO2 conversion of 95% is obtained with t CO2/t H2 of 10/300, i.e. very short adsorption period and long hydrogenation period.The conversion of H2 is defined by (Eq. 6). When the operation is carried out with a CO2 adsorption and hydrogenation periods timing of 45/300 (Section 3.2.1, Fig. 2a) the H2 conversion results in 23%. First, it should be noted that unreacted H2 is observed at the reactor outlet from the very beginning of the hydrogenation period, which reveals a slow CO2 desorption and hydrogenation kinetics. In order to fully regenerate the catalyst, long hydrogenation periods are required, as observed in the previous section. However, due to the slow hydrogenation kinetics, high amounts of hydrogen are emitted without being converted, which reduces H2 conversion. The hydrogen conversion is hardly affected when the CO2 adsorption time is extended from 45 to 150 s, i.e. t CO2/t H2 of 150/300 (Section 3.2.2, Fig. 2b). The only difference among those operating conditions is that a longer adsorption time of 150 s fully saturates the catalyst. Consequently, slightly higher amounts of carbonates are adsorbed on the catalyst surface, which enhances somewhat H2 conversion to 25%. Finally, hydrogen conversion is significantly promoted to 51% when the operation is carried out with a CO2 adsorption and hydrogenation periods timing of 150/60 (Section 3.2.3, Fig. 2c). Due to the short regeneration period of 60 s, the hydrogenation of carbonates occurs with a high local concentration of carbonates, which enhances hydrogenation kinetics, and thus, results in higher H2 conversion. Note that under these operating conditions, high H2 conversion is obtained but at the expense of a low CO2 conversion due to an incomplete regeneration of the catalyst. Fig. 4b shows the evolution of the H2 conversion (X H2, %) as a function of the CO2 adsorption and hydrogenation periods timing. For a given t CO2, the H2 conversion decreases with t H2 due to the progressive inefficient usage of hydrogen, as already explained. For a given t H2, the H2 conversion is promoted up to t CO2= 60 s, and afterwards, is maintained unaltered. H2 conversion is promoted in the t CO2 range (0–60 s) where a progressive extension of the storage period results in a higher amount of CO2 stored. Therefore, a higher population of carbonates promotes hydrogen consumption. Storage times longer than 60 s do not change the amount of CO2 stored (since the catalyst is already fully saturated) and consequently do neither modify H2 conversion. Maximum H2 conversion of 56% is obtained for t CO2/t H2 of 60/20, i.e. a storage time leading to a complete saturation of the catalyst and a very short hydrogenation period.The production of CH4 is defined by (Eq. 3). When the operation is carried out with a CO2 adsorption and hydrogenation periods timing of 45/300 (Section 3.2.1, Fig. 2a) the production of CH4 results in 219 μmol g−1. The long hydrogenation period guarantees the complete regeneration of the catalyst, and thus, CH4 production is promoted. A slightly higher amount of CH4 is produced (232 μmol g−1) when the CO2 adsorption and hydrogenation periods timing is set at 150/300 (Section 3.2.2, Fig. 2b). The extension of the adsorption period leads to the complete saturation of the adsorption sites, and consequently, CH4 production is slightly enhanced during the hydrogenation period. Finally, CH4 production is significantly reduced to 73 μmol g−1 when the CO2 adsorption and hydrogenation periods timing is set at 150/60 (Section 3.2.3, Fig. 2c). Due to a short regeneration period, the catalyst is not completely regenerated and CH4 concentration peak is not totally developed, as observed in Fig. 2c. Fig. 4c shows the evolution of the CH4 production (Y CH4, μmol g−1) as a function of the CO2 adsorption and hydrogenation periods timing. Maximum CH4 production of 232 μmol g−1 is obtained for t CO2/t H2 of 60/300, i.e. a storage time leading to a complete saturation of the catalyst and a very long hydrogenation period to promote the complete decomposition of adsorbed carbonates and their hydrogenation to CH4.It is obvious that during the CO2 adsorption and hydrogenation, conversion of CO2 and H2 along with CH4 production should be maximized. However, as already observed in the previous section, it is not possible to look for a unique CO2 adsorption and hydrogenation period timing (t CO2/t H2) to maximize jointly three catalytic parameters. In principle, results in Fig. 4 suggest adsorption times around 60 s (close to catalyst saturation) and moderate hydrogenation times, which produce a high amount of CH4 per cycle with a reasonable H2 conversion. To better select the optimal hydrogenation time, a more appropriate catalytic parameter should be the average formation rate of CH4 ( r ¯ CH 4 , μmol g−1 s−1). Fig. 5 shows the production of CH4 and the average formation rate of CH4, as function of the hydrogenation time. The adsorption time has been set at 60 s in Fig. 5. As explained above (Fig. 4c), the amount of CH4 produced increases with the hydrogenation time, having a greater slope for low times. On the other hand, the average formation rate has a maximum between 80 and 100 s of hydrogenation. However, 120 s is selected as the optimal hydrogenation time, because also present a high average formation rate of CH4 and would allow working with three identical beds in parallel, one operating in adsorption and two regenerating producing methane. Thus, under the optimum CO2 adsorption and hydrogenation periods timing of 60/120 the production of CH4 results in 148 μmol g−1 cycle−1 (1.2 μmol g−1 s−1) and a CO2 and H2 conversion of 25% and 43%, respectively.This operation strategy, with 3 catalytic reactors, one working in adsorption and the other two regenerating producing SNG, is shown in Fig. 6. In Scheme 1 of Fig. 6, the first reactor operates in adsorption and the second and third in hydrogenation. However, the reactors that operate in hydrogenation are out of phase. When the hydrogenation begins in the second reactor, it is fully saturated, while the third reactor is partially regenerated, having completed half of the period. In parallel, the first reactor begins the adsorption step fully regenerated. Next, Scheme 2 shows the period change in the first reactor (adsorption to hydrogenation) and in the third reactor (hydrogenation to adsorption), while the second continues to hydrogenation. Subsequently, in Scheme 3, the second reactor changes to adsorption and the third to hydrogenation. Once again, the reactors that work in hydrogenation, both in Schemes 2 and 3, are out of phase. Finally, from Scheme 3 it is changed to Scheme 1 and the operation continues cyclically alternating the schemes.The model used in this work allows predicting the temporal evolution of reagents and products during the dual operation of CO2 adsorption and methanation, considering that the adsorption sites can be occupied by CO2, H2O or simultaneously by both forming a weakly adsorbed bicarbonate. The evaluation of the CO2 adsorption and hydrogenation yield is carried out in the whole operation, considering the adsorption and hydrogenation performances. As this is a cyclic operation, the state of the catalyst at the beginning of a given period depends on the state of the catalyst at the end of the previous period. In simulations with a short adsorption period and a long hydrogenation period, the catalyst is not fully saturated with CO2 at the end of the adsorption period and some water remains adsorbed, specifically in the adsorption sites located at the rear of the reactor. On the other hand, the catalyst is fully regenerated at the end of the hydrogenation period and a significant fraction of the storage sites are occupied by H2O. In simulations with a long adsorption and hydrogenation periods, the catalyst is completely saturated with CO2 at the end of the adsorption period and no water remains adsorbed. By last, in simulations with a long adsorption period and a short hydrogenation period, the catalyst is not fully regenerated and some CO2 together with H2O remain adsorbed onto the storage sites at the end of the regeneration period. This fact limits the CO2 adsorption capacity of the subsequent storage period.The global performance of the catalyst is evaluated based on the CO2 conversion, H2 conversion and CH4 production. Maximum CO2 conversion of 95% is obtained with t CO2/t H2 of 10/300, i.e. very short adsorption period and long hydrogenation period. Maximum H2 conversion of 56% is obtained for t CO2/t H2 of 60/20, i.e. a storage time leading to a complete saturation of the catalyst and a very short hydrogenation period. In addition, maximum CH4 production of 232 μmol g−1 is obtained for t CO2/t H2 of 60/300, i.e. a storage time leading to a complete saturation of the catalyst and a very long hydrogenation period to promote the complete decomposition of adsorbed carbonates and their hydrogenation to CH4. Therefore, it is not possible to define a unique CO2 adsorption and hydrogenation period timing (t CO2/t H2) to maximize all the above catalytic parameters. Adsorption times around 60 s (close to catalyst saturation) and moderate hydrogenation times, which produce a high amount of CH4 per cycle with a reasonable H2 conversion, are appropriate. To better select the optimal hydrogenation time, a new catalytic parameter is set, the average formation rate of CH4 ( r ¯ CH 4 , μmol g−1 s−1). 120 s is selected as the optimal hydrogenation time, which enable to work with three identical beds in parallel, one operating in adsorption and two regenerating producing methane, with a high average formation rate. Thus, under the optimum CO2 adsorption and hydrogenation periods timing of 60/120 the production of CH4 results in 148 μmol g−1 cycle−1 (1.2 μmol CH4 g−1 s−1) and a CO2 and H2 conversion of 25% and 43%, respectively.By last, we are now doing further research to readjust the model to predict the operation in the presence of O2 and H2O during the adsorption period and simulate the new optimal operating conditions, on which we will report shortly. Alejandro Bermejo-López: Validation, Methodology, Investigation, Writing – original draft. Beñat Pereda-Ayo: Conceptualization, Methodology, Visualization, Writing – review & editing. José A. González-Marcos: Methodology, Software, Data curation, Supervision, Funding acquisition. Juan R. González-Velasco: Conceptualization, 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 financial support from the Economy and Competitiveness Spanish Ministry (CTQ2015-67597-C2-1-R and PID2019-105960RB-C21) and the Basque Government (IT1297-19) is acknowledged. The authors thank for technical and human support provided by SGIker (UPV/EHU Advanced Research Facilities/ ERDF, EU). One of the authors (ABL) also acknowledges the Economy and Competitiveness Spanish Ministry for his PhD grant (BES-2016-077855).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.cattod.2021.08.023. Figure S1 Supplementary material .

CO2 methanation could play a significant role in the future energy system. The excess of renewable electric energy can be transformed into storable methane to balance the energy demand when required. Moreover, the CO2 methanation can be performed alternating steps of CO2 storage and reduction, avoiding expensive CO2 purification steps. In this work, we will use a previously developed and validated model to optimize by simulation the CO2 adsorption and hydrogenation cycles timing (t CO2 /t H2). The performance of the catalyst is quantified by the CO2 conversion (X CO2, %), H2 conversion (X H2, %) and CH4 production (Y CH4, mmol g−1 cycle−1). Long adsorption and hydrogenation times result in high CH4 productions per cycle, however, low CO2 and H2 conversion. Therefore, adsorption times close to the catalyst saturation (t CO2=60 s) and moderate hydrogenation times are preferable. To better select the optimal hydrogenation time, a new catalytic parameter is set, the average formation rate of CH4 ( r ¯ CH 4 , μmol g−1 s−1). The optimal hydrogenation time is set at 120 s. In addition to having a high average formation rate of CH4, t CO2 /t H2= 60/120 cycle timing would allow to work with three identical beds in parallel, one in adsorption mode and two in regenerating mode. With the optimum cycle timing of 60/120 the production of CH4 results in 148 μmol g−1 cycle−1 (1.2 μmol CH4 g−1 s−1) and a CO2 and H2 conversion of 25% and 43%, respectively

Data will be made available on request.Recently, with the increasing depletion of fossil fuels and growing awareness of global warming, biomass has received increasing attention as an abundant and CO2-neutral renewable energy source [1]. Gasification is one of the most promising technologies for converting biomass waste into fuel gases comprising H2, CO, CH4 and CO2, given the high volatile matter content and low N, S and ash content of biomass feedstocks, as well as its operational simplicity and flexibility of use [2,3]. The fuel gas can be fed into gas engines or turbines to generate electricity and/or heat, or it can be further processed to produce hydrogen or syngas for fuel cells and Fischer-Tropsch synthesis, respectively [3]. However, during the gasification process, undesirable byproducts such as NH3, NO x , tar and fly ash are inevitably generated, with tar being one of the most troublesome contaminants [4]. Tar is typically a complex mixture of condensable hydrocarbons containing monocyclic to polycyclic aromatic hydrocarbons and oxygen-containing compounds that can condense on the available surface as operating temperatures fall below its dew point, resulting in clogging and corrosion problems [5,6]. Additionally, the tar content is highly influenced by the gasifier type and operating conditions, ranging from 0.5 to 100 g/m3, which usually exceeds the permissible maximum of various downstream devices [7]. For example, the acceptable tar content of gasification fuel gas fed into internal combustion engines and gas turbines is typically below 100 mg/m3 and 5 mg/m3, respectively [8]. Therefore, efficient removal or conversion of tar is critical for the use of the gasification fuel gas.Several methods for tar removal have been proposed, including mechanical separation, thermal cracking, and catalytic reforming. The use of mechanical separation could result in secondary environmental pollution and the loss of energy contained in tar [6]. In thermal cracking processes, high operating temperatures (>1000 °C) are typically required to achieve desirable performance, resulting in high energy consumption [9]. Although catalytic reforming can convert tar into valuable products at relatively low temperatures, its industrial application still faces two major challenges: firstly, high reaction temperatures (>600 °C) are still required, resulting in high energy costs; and second, rapid deactivation of the catalysts due to sintering and carbon deposition compromises processing stability [10].In addition to the aforementioned approaches, non-thermal plasma (NTP) technology is receiving increasing interest as a potential alternative for tar removal due to its ability to activate reactants under mild conditions [11,12]. In NTPs, the energetic electrons have a typical temperature of 1–10 eV, which is high enough to initiate chemical reactions, while keeping the gas temperature low [13]. Several types of NTP have been used for tar removal, including dielectric barrier discharge (DBD) [14–17], corona discharge [18,19], gliding arc discharge [20–22] and microwave discharge [23,24]. According to the literature, the advantages of using NTP for tar removal include high tar conversion, mild reaction conditions, and operational simplicity and convenience. However, the relatively high energy consumption and low selectivity toward the desired products may limit its industrial applications.The introduction of heterogeneous catalysis into NTP, known as plasma catalysis, provides a promising approach to addressing the aforementioned issues through catalyst functionalities such as lowering activation energy and tuning product selectivity [25,26]. There are two configurations for the combination of NTP and solid catalysts: one-stage and two-stage. The one-stage configuration means that the catalyst is placed directly in the discharge zone, partially or completely filling the discharge gap, whereas the two-stage configuration means that the catalyst bed is typically placed next to the plasma reactor, with the one-stage configuration being the most common for plasma-catalytic tar removal [11]. Great efforts have been made to couple different types of NTP with catalysts for tar removal, such as corona discharge coupled with Ni/SiO2 [27], DBD coupled with Ni/Al2O3 [28], Ni/ZSM-5 [29], Fe/Al2O3 [30], Mn@13X [17] and NiFe/(Mg, Al)O x [31], and gliding arc discharge coupled with Ni/Al2O3 [20] and Ni-Co/Al2O3 [32], among others. When compared to the plasma-only system, the coupling process significantly improves tar conversion, selectivity and yield of target products and energy efficiency. Moreover, comparable performance can be achieved at low temperatures when compared to catalyst-only cases. Clearly, plasma catalysis is a promising alternative for achieving effective tar conversion under mild conditions.The most attractive advantage of plasma catalysis is the potential to generate a synergistic effect by integrating NTP and catalyst, whereby the reaction performance achieved in the coupling system is better than the sum of those achieved in plasma-only and catalyst-only modes. Mei et al. combined gliding arc discharge and Ni-Co catalysts for steam reforming of mixed tar model compounds (toluene and naphthalene), and a synergistic effect was successfully obtained in terms of tar conversion, energy efficiency, and yield and selectivity for H2, CO2 and CH4 [32]. In our previous study, toluene removal was carried out in a DBD reactor coupled with Ni catalysts using a simulating gasification gas. At 400 °C, the highest toluene removal of 91.7% was achieved, which was significantly higher than the sum (58.1%) of those obtained in the catalyst-only and plasma-only processes [33]. The synergistic effect is resulted from the complicated interactions between NTP and solid catalysts, and a detailed understanding of the synergistic effect is critical to facilitate the design and optimization of plasma reactors and catalysts, thus achieving better performance at a lower energy consumption. However, there has been very little research into the synergistic effect of plasma catalysis, particularly for tar removal. For instance, the relationship between the synergistic effect and key factors such as operating conditions and catalysts is unclear.In this work, plasma-catalytic steam reforming of tar was carried out using a DBD reactor. Toluene was selected as a model tar compound as it is one of the main compounds with high thermal stability in tar products [34], and Ni/γ-Al2O3 was used as a catalyst because of its high activity and low cost. The effects of three key factors (reaction temperature, calcination temperature of catalysts and relative permittivity of packing materials) on reaction performance and synergistic effect were investigated. Moreover, the characteristics of both the discharge and the catalyst were investigated using various approaches to gain a better understanding of the synergistic effect in the plasma-catalytic reforming of tar.The Ni/γ-Al2O3 catalysts used in this work were prepared using a wetness impregnation method. Before use, the commercial strip-shaped γ-Al2O3 support (diameter × length: 3 × (4 – 10) mm, specific surface area: 169 m2/g, Jiangsu Jingjing New Materials Co., Ltd, China) was calcined in air at 550 °C for 3 h before being crushed and sieved to particle sizes of 40–60 mesh. An appropriate weight of γ-Al2O3 was added to an aqueous solution of Ni(NO3)2·6H2O and impregnated overnight at room temperature. Following impregnation, the catalyst precursor was dried at 120 °C for 10 h and before being calcined in an air atmosphere for 4 h at different temperatures (450, 500, 550 and 600 °C). The as-prepared catalysts were labeled as NA(x), where x represents the calcination temperature. The accurate Ni loading was determined to be 8.9 wt% using the inductively coupled plasma optical emission spectroscopy (ICP-OES).The specific surface area and pore volume of the catalysts were determined by N2 adsorption/desorption isotherms at −196 °C using a surface area analyzer (ASAP 2010, Micromeritics). Prior to the measurement, the samples were degassed at 200 °C for 10 h under vacuum.Powder X-ray diffraction (XRD) measurements were performed on an X-ray diffractometer (PANalytical, X’pert Pro MPD) equipped with a Cu Kα (λ = 0.154 nm) radiation source (40 kV and 40 mA) in the scanning range of 10-80°. The average crystallite size of Ni nanoparticles (NPs) was calculated by Scherer’s equation [35]: (1) D Ni = K × λ / ( β × cos θ ) where the dimensionless shape factor K is 0.9, and β is the full width at half maximum of the Ni (200) peak at 51.7°.H2-temperature programmed reduction (H2-TPR) measurements were carried out on a TPR instrument (ChemStar, Quantachrome). Before the measurement, 50 mg of sample was preheated in a He stream at 300 °C for 30 min before being cooled to room temperature. The reaction chamber was then filled with 50 mL/min of 10 vol% H2/Ar gas while the temperature was raised from 40 to 900 °C at a heating rate of 10 °C/min.The CO2 and NH3 temperature programmed desorption (CO2-TPD and NH3-TPD) were performed on a TPD instrument (ChemStar, Quantachrome). Prior to the adsorption, 150 mg of sample was reduced at 650 °C in 10 vol% H2/Ar (50 mL/min) for 1 h and then cooled to 50 °C in He flow. The sample was subsequently heated to 50 °C or 100 °C for CO2 or NH3 adsorption, respectively. The adsorption of CO2 or NH3 was conducted by flowing pure CO2 or 7.9 vol% NH3/He (50 mL/min) through the sample, respectively. After adsorption for 1 h, the sample was purged with the He flow until baseline stabilization, and then heated up to 800 °C with a heating rate of 10 °C/min in the He flow (50 mL/min). The corresponding TPD spectra were obtained by monitoring the desorbed CO2 or NH3 using a thermal conductivity detector.The pulse chemisorption of CO was carried out on a chemisorption apparatus (AutoChem II 2920, Micromeritics). Prior to the measurement, 0.5 g of sample was reduced at 650 °C for 1 h in a 10 vol% H2/Ar (50 mL/min) atmosphere and then cooled to 50 °C in He flow. The CO chemisorption was operated by injecting 0.5 mL of 8 vol% CO/He and repeating the procedure every 6 min until the CO peaks became identical. The CO uptake was measured by a thermal conductivity detector and used for the calculation of the Ni metal surface area using the following equation [36]: (2) S A Ni m 2 /g-catal. = X × SF × N × RA where X is the CO uptake in moles per gram of catalyst (mol/g-catal.), SF is the stoichiometric factor (1), N = 6.023 × 1023 Ni atoms/mol, and RA is the atomic cross-sectional area of Ni (0.0649 nm2).In addition, the dispersion degree (%D) and the average particle size (d Ni ) of Ni were calculated by the following equations [37]: (3) % D = 1.17 × X / W × f (4) d N i n m = 97.1 / % D where W is the weight percentage of nickel, and f is the reduction degree.Thermogravimetric analysis (TG, STA409PC, NETZSCH) combined with a mass spectrometry (MS, QMS403, NETZSCH) was used to characterize the spent catalysts. The samples were heated from 40 °C to 900 °C at a heating rate of 10 °C/min and an air flow rate of 30 mL/min.The FTIR spectra of spent catalysts were recorded by an infrared spectrometer (INVENIO-S, Bruker) in the range of 400–4000 cm−1 with a resolution of 4 cm−1. Before the measurement, 1 mg of each sample was mixed with 100 mg of KBr (purity > 99%, Aladdin), and the mixtures were pressed into wafers with a diameter of 13 mm.Catalyst surface analysis was performed on an XPS instrument (ESCALAB 250Xi, Thermo Fisher) equipped with an Al (Kα) (hv = 1486.6 eV) X-ray radiation source. All binding energies were calibrated based on the C1s hydrocarbon peak at 284.6 eV. Fig. 1 shows a schematic diagram of the experimental setup. The DBD reactor consists of a cylindrical corundum ceramic tube (i.d. 19 mm, o.d. 25 mm) wrapped with a 50-mm-long stainless-steel mesh as the outer electrode. A stainless-steel rod (diameter 16 mm) is placed along the axis of the tube as the inner electrode. Hence, the discharge gap is 1.5 mm with a corresponding discharge volume of ∼4.1 mL. The catalysts were held in place by a stainless-steel sieve attached to the end of the inner electrode, and quartz sand (40–60 mesh) was used to fill the region between the lower edge of the discharge zone and the stainless-steel sieve. In the experiments, 0.4 g of catalyst (about 0.8 mL), 1 mL of packing material (quartz, corundum, zirconia ceramics or silicon carbide) and 3 mL of quartz sand, all having the same particle size of 40–60 mesh, were placed at the discharge zone after being fully mixed. Then, plasma catalysis and catalyst-only modes can be achieved by turning the plasma on and off, respectively. In addition, the catalyst can be replaced by quartz sand with the same particle size to evaluate the performance of a plasma-only mode. The DBD reactor was placed inside a tubular furnace with a temperature range of room temperature to 750 °C. The reaction temperature was measured using a K-type thermocouple located on the outside reactor tube wall at the midpoint of the discharge zone after the reaction reached a stable stage. The catalysts were reduced in situ in a flowing 10 vol% H2/N2 at 650 °C for 1 h before the experiments. After each experiment, the reactor was cleaned by heating it to 700 °C for 1 h in an air atmosphere to remove carbon deposition and other contaminants formed during the reactions.Toluene and H2O were pumped into the mixing chamber by two syringe pumps (LSP01-1A, Longer Pump) with a flow rate of 5.196 μL/min and 11.64 μL/min, respectively, to attain a constant steam/carbon (S/C) molar ratio of 2. Subsequently, toluene and H2O were vaporized and mixed with 133 mL/min carrier gas (N2) in a mixing chamber with a temperature of 250 °C before being fed into the DBD reactor. The produced gas stream passed through two absorption bottles, which were connected in-line and placed in an ice water bath. The former one contained 50 mL of n-hexane or isopropanol solvent to collect unconverted toluene or condensable byproducts, respectively, while the latter was left empty to collect entrained droplets. To avoid condensation of water vapor, toluene and liquid products, the pipeline between the mixing chamber and the inlet, as well as the pipeline connecting the outlet to the absorption bottle and the vent were heated to 200 °C during the experiments.The plasma was generated by an AC high voltage power supply (CTP-2000 K, Nanjing Suman) with a peak voltage of 30 kV and a frequency of 5–20 kHz. The frequency was kept at 7.5 kHz in this work. The applied voltage (V) of the DBD reactor was measured by a high voltage probe (P6015A, Tektronix). The charge (Q) and current were obtained by measuring the voltage drops on a capacitor (0.1 μF) and a resistor (200 Ω), respectively. These signals were recorded by a digital oscilloscope (DPO2024B, Tektronix). In this work, the discharge power was determined by multiplying the area of the V-Q Lissajous diagram with the frequency and was fixed at 13 ± 0.5 W.The unconverted toluene and by-product (benzene) collected by the n-hexane-containing bottle in 5 min were analyzed using gas chromatography (GC, GC-2014, Shimadzu) equipped with a capillary column (AE-PEG-20 M, ATEO) and a flame ionization detector. The gas products were analyzed by an online gas chromatography system (GC, Micro GC490, Agilent) equipped with two thermal conductivity detectors, as well as a Molsieve 5A and PoraPLOT Q column. After a 1 h reaction, liquid products were collected by the isopropanol-containing bottle and analyzed by an off-line gas chromatography-mass spectrometry instrument (GC–MS, Thermo Fisher, Trace 1300-ISQ) equipped with a DB-5 ms column (Agilent). Further details on the GC and GC–MS measurements are presented in Table S1.The toluene conversion X toluene and energy efficiency E were determined by following equation: (5) X toluene ( % ) = [ T ] in - [ T ] out [ T ] in × 100 (6) E ( g / kWh ) = [ m ] removal P × 60 / 3600000 where [T]in and [T]out represent the molar concentration of toluene at the inlet and outlet, respectively, and P represents the discharge power in watt and [m]removal represents the grams of toluene removed per minute.Note that the external heat power was not taken into account in the calculation of energy efficiency, in consistence with previous works [29,30,38].The yield Y and selectivity S of the products, and the total gas yield Y T were calculated by equations 7–12. As we cannot measure the conversion of H2O in this study, the selectivity of H2 cannot be determined. (7) Y H 2 ( % ) = [H 2 ] out 4 × [ T ] in + [ H 2 O ] in × 100 (8) Y CO x ( % ) = [ CO x ] out 7 × [ T ] in × 100 (9) Y C x H y ( % ) = x × [ C x H y ] out 7 × [ T ] in × 100 (10) S CO x ( % ) = [ CO x ] out 7 × ( [ T ] in - [ T ] out ) × 100 (11) S C x H y ( % ) = x × [ C x H y ] out 7 × ( [ T ] in - [ T ] out ) × 100 (12) Y T ( % ) = [ H 2 ] out + [ CO ] out + [ CO 2 ] out + [ CH 4 ] out + [C 2 ] out + [ C 3 ] out [ T ] in + [ H 2 O ] in × 100 where [H2]out, [CO x ]out, [C x H y ]out are the molar amounts of H2, CO x (CO and CO2) and C x H y (CH4, C2H2, C2H4, C2H6, C3H6, C3H8 and C6H6) at the outlet, respectively, and [H2O]in is the molar amount of H2O at the inlet. C2 contains C2H2, C2H4 and C2H6, and C3 includes C3H6 and C3H8.The carbon balance B C of the plasma catalytic process was determined by equation (13). (13) B C ( % ) = ∑ S C x H y ( x = 1 , 2 , 3 , 6 ) ( % ) + S CO x ( x = 1 , 2 ) ( % ) The synergistic capacity SC was used to evaluate the intensity of the synergistic effect between plasma and catalysts and calculated by equation (14). (14) S C ξ ( % ) = ξ p+c - ξ p - ξ c ξ p + ξ c × 100 Where ξ can be the toluene conversion, and the yield and selectivity of gas products. The subscripts, p + c, p and c, represent the performances obtained by plasma catalysis, plasma-only and catalyst-only, respectively. Fig. 2 a shows the toluene conversion obtained in the plasma-only, catalyst-only and plasma catalysis modes at different reaction temperatures. Quite different removal behaviors are observed among these processes. In the catalyst-only system, the conversion of toluene increased progressively with increasing reaction temperature, reaching a maximum of 45.4% at 450 °C. However, in the plasma-only system, the toluene conversion gradually decreased from 96.5% at 200 °C to 67.1% at 450 °C. In the plasma catalysis process, the conversion of toluene reached a maximum of ∼100% at 200 °C, then remained at ∼95% in the temperature range of 250 °C to 350 °C, then declined to 79.4% at 400 °C, followed by a rise to 87.1% at 450 °C. Consistent with the change trend of the conversion, the energy efficiency of plasma catalysis first decreased and then increased with increasing temperature, ranging from 16.6 to 20.8 g/kWh (Fig. 2b). Fig. 2c shows the effects of reaction temperature on total gas yield. In contrast to the decreasing trend observed in the plasma-only mode, an increase in total gas yield with temperature is observed in both catalyst-only and plasma catalysis modes, especially above 350 °C, indicating that the efficient production of gas products is strongly dependent on catalysts. At 450 °C, the maximum gas yields for catalyst-only and plasma catalysis were 39.2% and 72.6%, respectively. As a result, the combination of NTP and catalysts at 450 °C achieved 87.1% toluene conversion, 72.6% total gas yield and 18.2 g/kWh energy efficiency, outperforming the plasma-only and catalyst-only modes.The effect of reaction temperature on the selectivity and yield of gas products and benzene is presented in Figs. S1, S2 and S3. In the plasma-only system, the selectivity and yield of gas products and benzene maintain below 14% within the temperature range of 200–400 °C, with CO and H2 being the main products. When the temperature increased to 450 °C, the selectivity and yield of CO, H2, and C2 dropped significantly to 2%, becomes the dominating component in produced gases. In the catalyst-only case, due to the increasing catalytic activity, at higher temperatures all gas products obtain higher selectivity and yield values, which are much higher than those in the plasma-only mode. H2, CO, CO2 and benzene were the major products, and particularly, the selectivity of benzene was even up to 45%, which is an unsatisfactory result considering the steam reforming pursuing the production of combustible gases. In plasma catalysis, the selectivity and yield of CO, CO2 and H2, at temperatures below 350 °C, maintain less than 10%, and then increase markedly with rising temperature, obtaining the maximums, most of which are higher than those obtained by catalyst-only. Meanwhile, the plasma catalytic process greatly lowers the selectivity of benzene, CO, H2 and CO2 being the main products. For instance, at 450 °C, the CO selectivity is up to nearly 50% with a corresponding yield of about 45%, together with the selectivity of benzene of less than 10%.At temperatures below 350 °C, plasma catalysis shows satisfactory results in converting toluene (Fig. 2), but its poor performance in generating gas products indicates that most of the toluene is converted into liquid products and/or carbon deposition. This carbon deposition is an unwanted byproduct that can reduce the catalytic activity by blocking active sites and lowering the discharge power, leading to a decline in plasma catalysis performance [28]. To gain a better understanding of reaction products, liquid products and carbon deposition produced at temperatures of 200 °C and 450 °C were analyzed. Researchers conducted GC–MS analysis of liquid products and TG-MS, FTIR, and XPS analysis of spent catalysts, and the results are shown in Fig. S5, Table S2, and Fig. S6. At 200 °C, most of the toluene was converted into carbon deposition, which was mainly composed of aliphatic carbon and easier to eliminate. At 450 °C, the increasing catalytic activity improved the oxidation reactions of carbonaceous species, leading to a decrease in the amount of carbon deposits and a corresponding increase in CO and CO2 production. Additionally, at both temperatures, a small portion of the removed toluene was converted into liquid products with a molecular weight greater than benzene. The number and relative area of O- and N-containing compounds were higher at 200 °C than at 450 °C. These liquid products were formed through reactions between intermediates and fragments or radicals, such as CN and OH radicals [39].Previous studies have shown that changes in reaction temperature can affect the physical properties of plasma and influence its chemistry [40,41]. Thus, we investigated the discharge characteristics of the DBD at different temperatures. The packed-bed effect in this study resulted in a combination of filamentary discharge and surface discharge. The introduction of packing material pellets reduced the available discharge volume, leading to the formation of filaments only in the void between the pellets and the reactor wall. Furthermore, an increased electric field, due to polarization effects and charge accumulation, was found to be mainly located around the contact points between the pellets, where surface discharge was formed and propagated along the surface [42,43]. Fig. 3 shows the electrical signals of the DBD operating at different temperatures with a fixed discharge power. The current signal of the discharge was quasi-sinusoidal with multiple superimposed current pulses per half-cycle of the applied voltage. When the reaction temperature increased from 200 °C to 450 °C at the same discharge power, the magnitude of the current pulses decreased, but the displacement current increased correspondingly, as well as the applied voltage decreasing from 10.4 kVpk-pk to 8.8 kVpk-pk (as shown in Fig. 3a and b). These changes indicate that filamentary discharge weakened while the component of surface discharge increased at high temperatures [42,44]. Fig. 3c shows the Lissajous curves of the DBD at different reaction temperatures while keeping the discharge power constant. As the temperature increases, the shape of the curve transforms from a parallelogram to an oval, indicating changes in the discharge characteristics. The Lissajous curve enables us to determine the the onset voltage (Uon) and the effective capacitances of the dielectric barrier (Cd) and the total system (Ctot). Using these parameters, we can calculate the capacitance of the gas (Cg), the breakdown voltage in the gas gap (Ub), and the average reduced electric field (E/n). The calculation process is detailed in section 6 of the Supporting Information. Table S3 summarizes the estimated parameters at different temperatures. Fig. 3d shows the E/n decreases from 96.8 Td to 80.7 Td as the temperature increases from 200 °C to 450 °C. This phenomenon has also been observed in plasma-assisted cellulose reforming [40] and plasma CH4 reforming [41].Furthermore, the mean electron energy at different E/n values can be calculated using the Boltzmann equation and BOLSIG+ [45–47], as shown in Fig. 4 a. The mean electron energy increases with rising E/n, however, as temperature increases in the range of 200–450 °C, it progressively declines from 1.95 eV to 1.50 eV. This decline in mean electron energy weakens the plasma chemistry trigger, which negatively impacts the reaction performance of plasma. In the plasma toluene steam reforming process, important active species such as excited N2 molecules, OH and O radicals initiate and drive reactions [38,48]. Hence, the rate coefficients of the electron impact reactions leading to the formation of these species were calculated using BOLSIG+, and are shown in Fig. 4b. The rate coefficient of all reactions increases with increasing E/n, implying that higher rate coefficients can be achieved at lower temperatures. This result suggests that higher reaction temperatures are not favorable for generating excited species and radicals that can effectively decompose toluene.The analysis presented above allows us to draw several conclusions regarding reaction performance. In plasma catalysis, the destruction of toluene depends heavily on plasma intensity at low temperatures. The decreasing E/n caused by rising temperatures lowers the mean electron energy, reducing the production of active species, and subsequently leading to a decrease in conversion. The higher E/n values are more favorable for ring cleavage of aromatic intermediates and toluene, which is mainly initiated through reactions with energetic electrons and excited N2 [49,50]. This could explain the higher aliphatic nature of the carbon deposits formed at low temperatures. At high temperatures, the increased catalytic activity plays a crucial role in toluene destruction, reversing the declining trend in conversion. Furthermore, the efficient formation of gaseous products is strongly dependent on catalysis, and a significant increase in gas production can only be observed at high temperatures where catalytic activity has notably increased, accompanied by a correspondingly significant decrease in carbon deposits.The synergistic effect of the process was evaluated by analyzing the toluene conversion and total gas yield. Fig. 5 a displays the values of synergistic capacities at different reaction temperatures. Synergistic capacities were calculated to evaluate the synergistic effect for toluene conversion and gas production. The results showed that the synergistic effect can only be achieved at temperatures below 350 °C for toluene conversion with a capacity of about 4%. However, the synergistic capacity decreases with increasing temperature from 350 to 450 °C. On the other hand, for gas production, the synergistic capacities remained negative at around −50%, at temperatures between 200 and 350 °C. The synergistic effect increased with temperature, reaching about 65% at 450 °C. Fig. S8 presents the synergistic capacities calculated using selectivity and yield of gas products. No synergistic effect was observed in terms of both selectivity and yield of all gas products at temperatures lower than 350 °C. At higher temperatures, the synergistic effect was concentrated in the yield of the main gas products (H2, CO and CO2), and the synergistic capacity significantly increased with temperature.The relationship between reaction temperature and the synergistic capacity in terms of toluene conversion and gas production is shown in Fig. 5b. The temperature dependence curve can be divided into two parts based on a threshold temperature of 350 °C. Below 350 °C, although the linear fitting method cannot achieve a satisfactory result, the low slope of the fitted straight line suggests a weak temperature dependence of the synergistic capacity in this temperature range. Above 350 °C, however, the relationship between temperature and synergistic capacity is linear and significant. The synergistic capacity in toluene conversion and gas production shows negative and positive temperature dependence, respectively.The catalytic performance of the catalyst and discharge characteristics of the DBD suggest that increasing the temperature from 200 °C to 450 °C enhances the formation of surface discharge and decreases the E/n. By contrast, the activity of the catalyst increases significantly above a threshold temperature of 350 °C (Figs. 2 and S2). The threshold temperature for the rapid increase in catalytic activity is consistent with the threshold temperature for the change in synergistic capacity. This suggests that the variation in catalytic activity plays a more important role in the generation of the synergistic effect compared to the discharge characteristics.In summary, the synergistic effect of plasma catalysis in the steam reforming of toluene is largely determined by the catalytic activity of the catalyst and is therefore greatly influenced by the reaction temperature. The synergistic effect is most pronounced at temperatures above 350 °C and is particularly noticeable in terms of gas production. Above this threshold temperature, there is a strong linear relationship between the synergistic capacity and the reaction temperature, with a negative correlation in toluene conversion and a positive correlation in gas production.The effect of calcination temperature on the performance of catalysts and the synergistic effect in steam reforming of toluene is discussed in this section. The experiments were conducted at 450 °C, which is the most suitable operating temperature for the plasma-catalytic process in this study. It is worth mentioning that the results obtained in the plasma-only mode in this section are equivalent to the results obtained at 450 °C in the previous section. This was achieved by replacing the catalyst with quartz sands to create a plasma-only mode in the study. Fig. 6 a shows the effect of different calcination temperatures on toluene conversion. It can be seen that an increase in calcination temperature leads to a decrease in both the catalyst-only and plasma-catalytic modes of conversion. For example, when the packing material is changed from NA(450) to NA(600), the conversion in the catalyst-only mode drops from 84% to 27% and in plasma catalysis it decreases from 100% to 43%. This is accompanied by a significant decrease in energy efficiency, from 20.9 g/kWh to 9.0 g/kWh (Fig. 6b). The 100% toluene conversion and 20.9 g/kWh energy efficiency achieved with NA(450) as a catalyst is a competitive result, especially in DBD systems, compared to similar works listed in Table S5. The total gas yield is presented in Fig. 6c. In the catalyst-only mode, the yield ranges from 25% to 40% and NA(500) and NA(600) give the maximum and minimum, respectively. In plasma catalysis, the total gas yield decreases significantly with the increase in the calcination temperature of the catalyst, from 85% with NA(450) packing to 35% with NA(600) packing. The results suggest that the use of a catalyst with a low calcination temperature is favorable for both toluene conversion and gas production in plasma catalysis.The results show that the selectivity and yield of gas products and benzene are influenced by the calcination temperature of the catalysts used (Figs. S11 and S12). As the calcination temperature of the catalysts increases, the selectivity of CO and benzene increases (Fig. S11), while the selectivity of CO2 achieves the highest value with NA(500). The yield of CO, CO2 and H2 decreases with the increasing calcination temperature of the catalysts used. The selectivity and yield of CH4, C2, and C3 is less than 2% and is not significantly influenced by the change of catalysts packed. In plasma catalysis, the selectivity of CO is kept at 50–60% and the benzene selectivity is significantly reduced compared to the catalyst-only process. The yield of the main gas products decreases with increasing calcination temperature of the catalysts.Apparently, the calcination temperature of catalysts strongly influences the reaction performance of plasma catalysis in terms of removal capacity and gas production. To better understand this effect, various characterization techniques, such as BET, XRD, H2-TPR, CO2–/NH3-TPD and CO pulse chemisorption, were employed. As shown in Table 1 , with increasing calcination temperature from 450 °C to 600 °C, the surface area of the catalyst decreases slightly from 141.0 m2/g to 134.8 m2/g, while the mean pore size rises from 11.5 nm to 11.9 nm. The XRD patterns of the reduced catalysts are presented in Fig. 7 a. The main peaks at 44.3°, 51.6° and 76.3° correspond to metallic nickel, and the calculated nickel particle sizes (6.1 and 6.8 nm) are slightly influenced by the calcination temperature. Fig. 7b shows the H2-TPR profiles of the catalysts calcined at different temperatures. The three main peaks, the low-, medium- and high-temperature peak, correspond to free NiO species, the NiO specie with stronger interactions with the support, and stable nickel aluminate with a spinel structure, respectively [51]. The increase of calcination temperature results in a shift of the low- and medium-temperature peaks to higher temperatures, as well as an increase in the intensity of the high-temperature peak. It indicates that a high calcination temperature strengthens the interaction between NiO species and the support, which is unfavorable for the reduction of NiO species during the activation treatment and leads to a decreased reduction degree of the catalysts. The basic and acidic properties of the catalysts were characterized by CO2– and NH3-TPD, and the results are shown in Fig. 7c and d. Clearly, the CO2 desorption curves show two broad peaks, corresponding to the desorption of weakly and strongly adsorbed CO2. Similarly, the NH3 desorption curves show three peaks, associated with weak and medium/strong acid sites [52]. Table 2 summarizes the base/acid site distribution and density of the catalysts after the curves were deconvoluted. Interestingly, the variation in calcination temperature did not significantly affect the acidic and basic properties of the catalysts. This is likely because the basicity and acidity of the Ni/Al2O3 catalysts primarily arise from the alumina support [35,53]. As the γ-Al2O3 support used in this study was already calcined at 550 °C before use, calcining the precursor within the 450–600 °C range did not induce a notable or regular change in the basic and acidic properties of the catalysts. CO pulse chemisorption analysis was used to determine the Ni surface area, dispersion and particle size, and the results are presented in Table 3 . The Ni surface area decreased with increasing calcination temperature, from 1.22 m2/g-catal. of NA(450) to 0.55 m2/g-catal. of NA(600). Notably, the lowest dispersion was obtained with NA(500) instead of NA(450). Despite this, increasing the calcination temperature appeared to enhance the metallic dispersion. On the other hand, the Ni particle sizes had an opposite trend to the dispersion with sizes ranging from 21.1 to 28.9 nm.As previously mentioned, increasing the calcination temperature resulted in only minor variations in pore structure, base and acid properties, with a linear decrease in Ni surface area and nonlinear changes in dispersion and Ni particle size. In the catalyst-only case, the progressively decreasing conversion observed with increasing calcination temperature can be attributed to the decrease in Ni surface area, which leads to a reduction in available active sites, limiting toluene destruction. On the other hand, the change in total gas yield induced by the calcination temperature of the catalysts can be explained by Ni particle size (or dispersion), given the similar change trend. It is well-known that the size or dispersion of metal particles significantly influence catalyst selectivity and, therefore, the product distribution [54,55]. Additionally, in the plasma-catalytic process, the similar and decreasing trend in both conversion and total gas yield implies that the Ni surface area plays a crucial role in determining the reaction performance of plasma catalysis. Fig. 8 a shows the synergistic capacities calculated using toluene conversion and total gas yield for the different catalysts. No synergistic effect is observed in terms of toluene conversion. Catalysts calcined at higher temperatures tend to have lower synergistic capacities, except for NA(450). However, a clear synergistic effect is achieved in gas production regardless of the catalyst, with the synergistic capacity decreasing in the order of NA(450) > NA(500) > NA(550) > NA(600). The synergistic capacities for gas product selectivity and yield are presented in Fig. S9, showing that the synergistic effect in selectivity is mainly concentrated in NA(450), while for yield, a noticeable but weakening synergistic effect is observed at higher calcination temperatures. No synergistic effect is obtained for benzene and light hydrocarbons, except for CH4.As discussed earlier, the intensity of the synergistic effect appears to be closely related to the Ni surface area, which is supported by the negative correlation observed between the synergistic capacity in gas production and the calcination temperature of the catalysts. To investigate this relationship further, the correlation between Ni surface area and the synergistic capacity was analyzed for both toluene conversion and gas production, and the results are presented in Fig. 8b. Notably, the synergistic capacities obtained using NA(450) were not considered, as toluene was not detected at the outlet, making it difficult to estimate the actual values of toluene conversion and total gas yield. The results show a clear positive and linear correlation between Ni surface area and the synergistic capacity, suggesting that a higher Ni surface area is associated with a stronger synergistic effect. This finding can be explained by the fact that a higher Ni surface area provides more active sites, which increases the probability of generating a synergistic effect.The effect of the relative permittivity of packing materials was also investigated. Four packing materials were employed: quartz, corundum, zirconia ceramics, and silicon carbide, all of which were calcined at 950 °C for 6 h prior to use. Table S6 lists the composition and relative permittivity of these packing materials. The relative permittivity increases in the order of quartz < corundum < zirconia ceramics < silicon carbide, with silicon carbide having the highest relative permittivity of 200.3, which is much greater than that of the other materials.The conversion obtained with different packing materials is presented in Fig. 9 a. In catalyst-only experiments, the conversion remains at about 45% regardless of the packing material used, indicating that these materials have little thermal catalytic activity for toluene steam reforming. In the plasma-only mode, the use of high relative permittivity packing materials results in a decrease in conversion from about 67% with quartz packing to about 40% with silicon carbide packing. However, packing high relative permittivity materials in the plasma catalysis system leads to a slight increase in toluene conversion, and accordingly, the energy efficiency slightly increases from 18.2 g/kWh with quartz packing to 20.0 g/kWh with silicon carbide packing, as shown in Fig. 9b.Taking into consideration the limited catalytic activity of the packing materials in the catalyst-only process, their effect on product generation is minimal and not considered significant. The impact of packing materials on the total gas yield is shown in Fig. 9c. In plasma-only experiments, the use of silicon carbide as the packing material results in the lowest total gas yield, while the other materials have similar values. However, under plasma catalysis, a noticeable difference in gas production is observed, particularly in the case of silicon carbide, which exhibits the highest total gas yield of about 90%, compared to the values of less than 80% obtained with other materials. Overall, the use of packing materials with high relative permittivity has both detrimental and promoting effects on the reaction performance of the plasma-only and plasma catalysis modes, respectively.The effects of packing materials on the selectivity and yield of gas products and benzene in the plasma-only and plasma catalysis modes are illustrated in Figs. S14 and S15. In the plasma-only mode, there is only a slight variation in the selectivity and yield of gas products and benzene among quartz, corundum, and zirconia ceramics, while the use of silicon carbide leads to a significant decrease in the selectivity and yield of produced gases, along with a corresponding increase in benzene selectivity and yield. In the case of plasma catalysis, the differences in gas production among packing materials are more pronounced, particularly for silicon carbide, which results in higher CO and CH4 selectivity, as well as higher H2, CO, CO2, and CH4 yields, while also lowering the benzene selectivity and yield.To further understand the effect of the relative permittivity of the packing materials, the discharge characteristics of the DBD packed with different materials are also characterized. Fig. 10 shows the electrical signals of the DBD packed with different materials and operated at a fixed discharge power. With the exception of silicon carbide, which shows the highest applied voltage of 9.2 kVpk-pk, other materials have values of around 8.8 kVpk-pk. As the relative permittivity of the packing materials used increases, the magnitude of current pulses gradually decreases, but the displacement current increases accordingly, especially in the case of silicon carbide. This phenomenon indicates that the use of packing materials with high relative permittivity could increase the component of surface discharge in DBD [42]. Higher relative permittivity materials are more effectively polarized, resulting in a stronger locally enhanced electric field, especially around the contact points between pellets [56]. Thus, surface discharge on the surface of pellets is easily ignited when using high relative permittivity packing material. Fig. 10c exhibits Lissajous curves of the DBD with different materials packing at a constant discharge power. Quartz, corundum, and zirconia ceramics have almost identical Lissajous curves, but that of silicon carbide is quite different. Table S4 summarizes discharge parameters calculated through different Lissajous curves, and the E/n is shown in Fig. 10d. Quartz, corundum, and zirconia ceramics have almost the same E/n values, which are much higher than that obtained with silicon carbide. Obviously, the three materials that possess close values of relative permittivity have close discharge parameters, but due to the large difference in relative permittivity, silicon carbide gets quite different values.Based on the discharge characteristics results presented, it appears that in the plasma-only mode, the decrease in toluene conversion and total gas yield when using silicon carbide can be attributed to a decline in E/n. However, in the plasma catalysis mode, the results show that the use of silicon carbide leads to higher toluene conversion and total gas yield. This improvement is likely due to the increased surface discharge in the presence of silicon carbide. Fig. 11 a shows the synergistic capacities on conversion and total gas yield using different packing materials. The results indicate that all packing materials exhibit a synergistic effect in terms of gas production, with silicon carbide showing the highest synergistic capacity at around 120%, while the other materials have values around 70%. However, for toluene conversion, the synergistic effect is only observed in the case of silicon carbide, while the other materials show negative synergistic capacities around −15%.We also investigated the effect of packing material on the synergistic capacities calculated based on the selectivity and yield of gas products, as shown in Fig. S10. The results indicate that the synergistic effect is observed in the selectivity of CO, as well as in the yield of H2, CO, and CO2. In particular, silicon carbide exhibits higher and lower synergistic capacities in the yield of main gas products and benzene, respectively, compared to other packing materials. Moreover, for light hydrocarbons the synergistic effect is mainly observed in the selectivity and yield of CH4 and C3 hydrocarbons.We further examined the correlation between the relative permittivity of packing materials and the synergistic capacity in terms of toluene conversion and gas production, as shown in Fig. 11b. The results indicate that the synergistic capacity exhibits a positive and strong linear correlation with the relative permittivity of the packing materials. This finding suggests that the use of a packing material with high relative permittivity could enhance the synergistic effect.The observed correlation may be attributed to the increased surface discharge in the case of packing materials with high relative permittivity. The greater surface discharge component implies that more area on the catalyst surface is covered by the discharge, as reported in previous studies [43]. Additionally, the active species generated by discharge and involved in surface reactions through Langmuir-Hinshelwood or Eley-Rideal mechanisms are considered key drivers of the synergistic effect [26]. However, most of the active species produced by discharge activation have a short lifetime [57]. Therefore, increasing the discharge-covered area on the catalyst surface could increase the probability of active species participating in surface reactions, thereby intensifying the synergistic effect.In summary, the choice of packing material has a significant impact on the synergistic effect in plasma-catalytic processes, affecting the selectivity and yield of gas products. Moreover, using a packing material with high relative permittivity could lead to a stronger synergistic effect due to the increased surface discharge and higher probability of active species participating in surface reactions.In this study, we investigated the performance of plasma-catalytic steam reforming of toluene in a DBD plasma reactor combined with Ni/γ-Al2O3 catalysts. The results showed that the toluene conversion and gas production were affected by the reaction temperature, catalyst calcination temperature, and packing material relative permittivity. At low reaction temperatures, the toluene conversion mainly depended on the intensity of the plasma, while gas production was limited. However, at high reaction temperatures, the increased catalyst activity promoted toluene conversion and enhanced the oxidation of carbonaceous species, leading to a greater production of gas products. The process achieved a high toluene conversion of 87.1%, a total gas yield of 72.6%, and an energy efficiency of 18.2 g/kWh at 450 °C. Furthermore, we found that the synergistic capacity of plasma catalysis was positively correlated with the metal surface area and relative permittivity of the packing materials, and negatively correlated with the reaction temperature in terms of toluene conversion. However, gas production had a positive correlation with reaction temperature. These findings suggest that using catalysts with lower calcination temperatures and packing materials with higher relative permittivity can improve the process efficiency. Overall, this work highlights the potential of plasma-catalytic steam reforming of toluene for sustainable hydrogen production and provides insights into optimizing the process parameters.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 financially supported by the National Natural Science Foundation of China (Grant No. 52106282), the National Key R&D Program of China (Grand No. 2019YFB1503902), the Strategic Priority Research Program of Chinese Academy of Sciences (Grand No. XDA21060600), the Science and Technology Program of Guangzhou (Grant No. 202102020292, 201904010098 and 202002030126) and the Natural Science Foundation of Guangdong Province of China (Grant No. 2019A1515011535). X. Tu thanks the support of the British Council Newton Fund Institutional Links Grant (No. 623389161). N. Wang thanks the University of Liverpool and the Chinese Scholarship Council for funding his PhD.Supplementary data to this article can be found online at https://doi.org/10.1016/j.cej.2023.142696.The following are the Supplementary data to this article: Supplementary data 1

In this study, steam reforming of toluene was carried out in a dielectric barrier discharge (DBD) plasma reactor combined with Ni/γ-Al2O3 catalysts. The effect of reaction temperature, calcination temperature of catalysts, and relative permittivity of packing materials, on the reaction performance and synergistic effect of plasma catalysis was investigated. The results showed that toluene conversion decreased initially and then increased with increasing temperature, due to a decreasing average reduced electric field and increasing catalytic activity at higher temperatures. At 450 °C, the process achieved a high toluene conversion of 87.1%, a total gas yield of 72.6%, and an energy efficiency of 18.2 g/kWh, demonstrating the potential of this approach for sustainable hydrogen production. Catalysts prepared at lower calcination temperatures or with higher relative permittivity packing materials perform better, owing to the larger Ni surface area available for catalytic reactions and the higher surface discharge facilitating the occurrence of surface reactions. In addition, the synergistic capacity in terms of toluene conversion and gas production exhibited a positive relationship with the metal surface area of catalysts and the relative permittivity of packing materials, while the relationship between reaction temperature and toluene conversion was negative.

Energy crisis and environmental pollution have increasingly limited the development of human society, so seeking and developing high-efficiency, eco-friendly and recycling new energy has been imminent [1–5]. The natural-born merits of high energy density, strong renewable ability, convenient transportation, carbon-free emission make hydrogen as a ''Holy Grail'' of new energy to replace the traditional fossil energy which is imminent depletion and nonrenewable [6,7]. Because of the relatively low energy consumption, green and clean production technology, safe and simple operation process, the new way for acquiring hydrogen by water electrolysis has received extensive attention [8,9]. Water splitting cannot occur spontaneously, but in theory, it can be achieved by applying a voltage of 1.23 V [10]. However, the limitation of the slow OER dynamics occurring at anode, the efficiency of water electrolysis is not ideal. The required driving potential for the formation of O–O bond in actual conditions is much higher than that in theoretical conditions [11,12]. Therefore, it is of great significance to find appropriate means to reduce the anode reaction potential for improving the economic benefits of electrolytic water and promoting industrial application technology. In general, there are two feasible methods that have been widely accepted for reducing the driving potential. On the one hand, change the type of anodic oxidation reaction, which means replacing OER with other feasible anodic reactions with much lower theoretical oxidation potentials; on the other hand, explore and prepare robust, stable and low-cost oxygen-evolving electrocatalysts to accelerate the reaction rate and improve the reaction efficiency [13–16].The type of anodic electrooxidation reaction depends mainly on the choice of electrolyte. KOH solution is usually served as the electrolyte for OER, and rich OH− environment is conducive to accelerating the formation of O2. Recently, some readily oxidized nucleophile reagents are suitable candidates for replacing OER because of their superior oxidation thermodynamics, such as alcohols, aldehydes, amines and urea [17–20]. Among these nucleophiles, urea stands out for its superior stability, high energy density, low toxicity and abundant storage. Moreover, hydrogen production and wastewater degradation can be realized simultaneously by urea electrolysis [21,22]. Compared with the single KOH electrolyte, when a certain proportion of urea is added in the process of water electrolysis, HER (6H2O + 6 e - → 3H2 + 6 OH − ) still occurs at the cathode, while UOR (CO(NH2)2 + 6 OH − → N2 + 5H2O + CO2 + 6 e - ) occurs at the anode (seek supporting information for the specific reaction process of urea electrolysis) [23]. Notably, the UOR process shows more favorable chemical reaction kinetics compared with the sluggish OER process, of which theoretical oxidation potential (0.37 V) is much lower than that (1.23 V) of the OER [24]. Nevertheless, the complicated steps of six-electron transfer make the gas release process difficult, so, the actual UOR process needs to be further optimized [25,26].Precious metals and their derivatives are excellent catalysts for water electrolysis, while they subjected to high price, scarce storage and weak stability [27,28]. In recent years, researches have made some achievements on non-precious metal-based electrolytic water catalysts. For example, transition metal oxides [29,30], hydroxides [31,32], sulfides [33,34], phosphides [35,36], selenides [37,38] and nitrides [39,40], which not only avoids the defects of precious metal-based catalyst, but also gain on the former level in performance constantly. Among them, transition metal sulfides (TMSs) such as MxSy (M = Ni, Co, Mo, Cu, Zn, etc.) are well-known for the outstanding electrocatalytic activity, resulting from its rich valence states and certain structural defects [41,42]. In particular, the complex of two or more TMSs (e.g., Co9S8@Ni3S2 [43,44], MoS2@Ni3S2 [45,46], Cu2S@Co9S8 [47], Ni3S2@Co9S8@MoS2 [48], CdS@Co9S8@Ni3S2 [49], etc.) possesses higher conductivity, larger specific surface area and more active sites due to the metal synergistic effect, defective heterointerface and hierarchical structure, which are not available in single-component TMSs. Although much progress has been made in the exploration of the mixed TMSs, it is still important but also challenging to further improve their catalytic activity for water electrolysis. Elaborate design and synthesis of mixed TMSs with appropriate structures is an effective strategy to improve their electrocatalytic property. As a new star in the family of porous crystal materials, the metal–organic framework (MOF) has attracted great attention since its appearance in 1995 [50]. On the one hand, the adjustable framework structures, unique porous characteristics and clear crystal distributions contribute to abundant active sites; on the other hand, the in-situ growth of the material eliminates the use of adhesive, which significantly reduces contact resistance [51,52]. Therefore, MOF materials are widely worked as various advanced electrodes, for instance, Tang et al. synthesized a novel hybrid nanostructure of CeOx nanoparticles dotted the Zeolitic imidazolate framework (ZIF) derived hollow CoS (CeOx/CoS) by means of interfacial engineering strategy for boosting the alkaline oxygen evolution, which only requires a low overpotential of 269 mV to delivers the current density of 10 mA cm−2 [53]. Zhou et al. carbonized MOFs on conductive support nickel foam (NF) in a few minutes by advanced laser-induced annealing technology to obtain an excellent water electrocatalyst, of which the high activity results from the remarkable adsorption of intermediates by the nickel-doped Fe3O4 overlayer formed during laser treatment [54].Intrigued by these above-mentioned studies, in this work, we firstly successfully synthesized MOF-derived ZCNS with hollow NSAs on NF through a facile two-step hydrothermal method. Keeping the total moles of Zn2+ and Co2+ ions constant, a set of parallel samples ZCNS-r (r = 1/3, 1/2, 1) were obtained by adjusting the molar ratio of Zn2+ and Co2+ ions to optimize the catalytic performance of materials. The introduction of Zn2+ ion in the first hydrothermal process directly created the unique sword-like MOF structure and the Niδ+ (δ = 2 or 3) ion released from the corroded NF during the second sulfuration process also generated the Co9S8@Ni3S2 heterostructure unexpectedly. Meanwhile, in order to study the effect of Zn ion on the configuration and activity of the catalyst, a control sample (without Zn ion, called CNS) was synthesized by the same method. Finally, the close-knit and hollow ZCNS-1/2 NSAs were obtained, the material not only have an optimum activity but also display an excellent stability for catalyzing both water and urea electrolysis. It's worth noting that the ZCNS-1/2 material display superior electrocatalytic performance to deliver a certain current density for HER (97 mV@10 mA cm−2, 215 mV@100 mA cm−2), OER (1.463 V@20 mA cm−2, 1.537 V@100 mA cm−2), UOR (1.264 V@20 mA cm−2, 1.316 V@100 mA cm−2), water electrolysis (1.522 V@10 mA cm−2, 1.721 V@100 mA cm−2) and urea electrolysis (1.314 V@10 mA cm−2, 1.506 V@100 mA cm−2). This research will provide certain reference to design and synthesize MOF-derived trimetallic sulfides as efficient and stable electrocatalyst for enhanced water and urea electrolysis.Concentrated hydrochloric acid (HCl, 12 mol/L), acetone (CO(CH3)2, >99%), Cobalt nitrate hexahydrate (Co(NO3)2·6H2O, >99%), Zinc nitrate hexahydrate (Zn(NO3)2·6H2O, >99%), 2-Methylimidazole (C4H6N2, >98%), Thioacetamide (TAA, CH3CSNH2, >99%), urea (CO(NH2)2, >99%), ethanol (CH3CH2OH, >99%) and potassium hydroxide (KOH, >99%) were bought from Sinopharm Chemical Reagent Ltd and no further purification was required before use. Nickel foam (NF, 1.0 mm in thickness) was worked as substrates as well nickel source of materials with pretreatment before use. Furthermore, sufficient deionized water (DIW) was prepared throughout the experiments. The dosages of relevant reagents were recorded in Table S1.i) Pretreatment of NF. NF (3 cm × 7 cm) was ultrasonically treated in 100 mL 3.0 M HCl solution and 100 mL acetone for 20 min, respectively, to remove the oxide layer and oil stain from its surface. Then, it was rinsed subsequently with ethanol and DIW several times and dried under vacuum at 50 °C for 6 h to ensure a clean and dry surface. ii) Preparation for the mixed solutions of zinc nitrate and cobalt nitrate (denoted as solution A). The total moles of Zn2+ and Co2+ ions were kept at 2 mmol, and then different masses of zinc nitrate and cobalt nitrate were dissolved in 40 mL DIW respectively according to the molar ratio (1/3, 1/2 and 1/1) of Zn2+ and Co2+ ion. iii) Preparation of 2-methylimidazole solution (denoted as solution B). 1.25 g 2-methylimidazole was dissolved in 40 mL DIW with vigorous stirring to obtain an orange transparent solution. The solutions A and B were poured into a 100 mL Teflon-lined stainless-steel autoclave to form a blue-violet mixed solution, and then a piece of pretreated NF was transferred into the resulting solution. After a tight sealing, the autoclave was heated at 70 °C for 4 h. When the reaction was completed, cooled it to room temperature naturally. Taking the materials out and washed subsequently with DIW and ethanol several times, and then dried under vacuum at 50 °C for 6 h to obtain a series of ZnCo MOFs materials. If zinc nitrate was removed from solution A, Co MOF would be synthesized as a control sample by the same steps.First, 200 mg TAA was added into 80 mL ethanol and stirred vigorously for 30 min to form a clarified solution C. Then, all the MOFs and four copies of solution C were transferred to a 100 mL Teflon-lined stainless-steel autoclave respectively, sealed and kept at 150 °C for 2 h. Finally, take out the materials, washed it repeatedly and dry thoroughly to obtain the parallel group samples Zn–Co–Ni–S-r (ZCNS-r, r represents molar ratio of Zn2+ and Co2+ ion, r = 1/3, 1/2, 1) and controlled sample Co–Ni–S (CNS).The hollow ZCNS-1/2 NSAs were prepared by a simple and facile two-step hydrothermal method, involving coordination precipitation reaction and sulfuration process, respectively (Fig. 1 ). The first reaction was a low temperature hydrothermal process, involving the trapping of metal ions by organic ligands and the growth of ZnCo MOFs on the substrate. In the first step, 2-methylimidazole was acted as an organic ligand, zinc nitrate and cobalt nitrate provided Zn and Co sources, respectively. NF was selected as the substrate due to its foam-like 3D porous structure and the innate high electrical conductivity. A blue-violet solution was formed swiftly after the pink solution A mixed with the orange solution B, which was the result of the rapid and sufficient capture of metal ions by the ligand reagent. When the reaction was completed, the color of the NF changed from silver-gray to blue-violet and the color gradually deepened with the increase of the molar ratio of Zn2+ and Co2+, demonstrating that the ZnCo MOFs material were successfully grown on the NF. In the second step, the ethanol solution of TAA provided the S source in the sulfuration process, it's worth noting that it would cause some etching on the NF substrate to form nickel sulfide [43,55]. While the ligand in the ZnCo MOFs were gradually substituted with S2− during the sulfuration process, leading to the formation of MOF-derived Zn–Co–Ni sulfides (ZCNS) with smaller solubility product and the color of NF also turned black by degrees.The chemical composition and crystal information of the material were obtained by X-ray diffraction (XRD) analysis. Under the influence of the material substrate, the three strong peaks located at 44.51°, 51.85° and 76.37° are ascribed to NF (JCPDS #04–0850) (Fig. 2 a). The peaks seated at 28.32ᵒ, 47.99ᵒ and 56.29ᵒ are assigned to (111), (220) and (311) crystal plane of ZnS (JCPDS #05–0566), while the peaks located at 29.91ᵒ and 73.38ᵒ are corresponding to (311) and (731) crystal plane of Co9S8 (JCPDS #19–0364). Notably, the peaks at 21.75ᵒ, 31.11ᵒ, 37.78ᵒ, 49.73ᵒ, 50.12ᵒ, 55.21ᵒ, 55.46ᵒ can be attributed to (101), (110), (003), (113), (211), (122) and (300) crystal plane of Ni3S2 (JCPDS #44–1418), which results from the corrosion effect on NF during the sulfuration process. Moreover, the enlarged XRD patterns shows that ZnS only existed in the parallel samples rather than the control sample (Fig. 2b), indicating that the synthesized parallel samples were ternary Zn–Co–Ni sulfides (ZCNS) and the control sample was binary Co–Ni sulfides (CNS). Both scanning electron microscope (SEM) and transmission electron microscope (TEM) can be used to analyze the morphology of materials and the detailed crystal information of the single hollow ZCNS-1/2 NS can be further obtained by high-resolution TEM (HR-TEM). The morphology of CNS shows a simple combination of dispersed particles and irregular clumps (Fig. S1a), which are speculated to be Co9S8 and Ni3S2, respectively. With the introduction of Zn, the metal skeleton structure is basically formed and ZCNS generally displays a hollow nanosword (NS) structure, of which the inner wall thickens with the increase of molar ratio of Zn2+ and Co2+ ion. Furthermore, the inner wall of ZCNS-1/3 is too thin to maintain the structure, resulting in a certain degree of fragmentation (Fig. S1b1-4). It's worth noting that the inner wall of ZCNS-1 is so thick that causes the longitudinal growth of the surface and the combined growth of the bottom (Fig. S1d1-4). Only ZCNS-1/2 equips with an array structure (Fig. 3a1-4 and Fig. S1c1-4) and the uniform distribution results in a larger specific surface area, which is conducive to the catalytic reaction. In addition, TEM images from different angles (Fig. 3b1-4 and c1-4) reveals the unique hollow sword-like nanostructure of ZCNS-1/2, which is in line with the SEM observation (Fig. 3a1-4). Fig. 3b4 shows the HR-TEM of a single NS from the front view, where the three distinguished lattice fringes are assigned to the (311) crystal plane of Co9S8 with 0.298 nm, (110) crystal plane of Ni3S2 with 0.287 nm and (200) crystal plane of ZnS with 0.271 nm. It's worth noting that the heterostructure of Co9S8(111)@Ni3S2(101) and the (111) crystal plane of ZnS with 0.312 nm simultaneously exist in the HR-TEM from top view (Fig. 3). The existence of heterostructure is beneficial to the rearrangement of local positive and negative charges, thus accelerating the charge transfer rate of the material [56,57]. Considering that the top of the NS is the intersection of all planes, where the overall elements distribution can be better observed, so the element mapping of the NS is emphasized (Fig. 3d and e). The detected elements of Zn, Co, Ni and S mainly concentrates at the junction of each surface while less distributes on the surface, which may explain why some NSs have holes in certain planes. Energy-dispersive X-ray (EDX) spectrum (Fig. 3f) demonstrates that Zn, Co, Ni and S elements coexist in ZCNS-1/2 and their specific weight distribution is about 1.46: 1: 1.24: 1.77. The catalyst powder scraped from NF surface was analyzed by Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES) to obtain the real content and proportion of each metal element, and the measured molar ratio of Zn and Co is approximate to the added counterpart (Table S2).X-ray photoelectron spectroscopy (XPS) is usually used to obtain the information of surface elemental valence state from the as-prepared samples. As indicated in the XPS survey spectra (Fig. S2a), Co, Ni and S elements simultaneously exist in CNS and ZCNS-1/2 with almost the same peak position except for the extra Zn element in ZCNS-1/2, further verifying that the Zn2+ was successfully introduced into CNS to form ZCNS. As shown in Fig. 4 a, the XPS spectrum of Zn 2p is mainly fitted with two peaks, among which the peak at 1047.6 eV is Zn 2p1/2 orbit and the other peak at 1023.3 eV is Zn 2p3/2 orbit [58,59]. The XPS spectra of Co 2p and Ni 2p can be well deconvoluted into two spin–orbit doublets and two shakeup satellites. Because of the existence of the electron transfer and electronic coupling between ZnS and Co9S8 as well between ZnS and Ni3S2, both Co 2p and Ni 2p in ZCNS-1/2 have a slight negative deviation compared with that of CNS. For Co 2p spectrum of ZCNS-1/2 (Fig. 4b), the peaks located at 797.5 eV and 782.1 eV are assigned to the Co2+ of Co 2p1/2 and Co 2p3/2, while those located at 795.8 eV and 780.1 eV are ascribed to the Co3+ [60,61]. Likewise, for Ni 2p spectrum of ZCNS-1/2 (Fig. 4c), the peaks seated at 874.2 eV and 856.6 eV come down to the Ni3+ of Ni 2p1/2 and Ni 2p3/2, while those seated at 872.4 eV and 854.9 eV are boiled down to the Ni2+ [45,62]. Significantly, compared to CNS, the ratio of Co2+/Co3+ increases while the ratio of Ni2+/Ni3+ decreases in ZCNS-1/2, indicating that charge transfer and electron rearrangement may occur in the heterostructure of Co9S8(111)@Ni3S2(101) after the addition of Zn2+. What is more, the increase of Co2+ content can promote the generation of more active intermediates (CoOOH) as efficient active sites during the catalytic process, which is beneficial to improving the capture of OH− ions and the release of O2, giving rise to an accelerated reaction kinetics [53]. The XPS spectra of S 2p and O 1s can be well deconvoluted into three set of peaks. Contrary to Co 2p and Ni 2p, both S 2p and O 1s in ZCNS-1/2 have a slight positive deviation compared with that of CNS. For S 2p spectrum in ZCNS-1/2 (Fig. 4d), the peak at 168.5 eV, 161.6 eV and 162.4 eV are corresponding to the S–O bond, S 2p1/2 and S 2p3/2, respectively [41,55]. For O 1s spectrum in ZCNS-1/2 (Fig. S2b), the peak Oa located at 532.9 eV is corresponding to physicochemical water adsorbed on the surface of catalysts, the peak Oi seated at 531.5 eV is matching to oxygen ions, the peak Om situated at 529.7 eV is referring to the metal–oxygen bond [43,63]. When Zn2+ is introduced and the total amount of metal ions remains unchanged, the adsorption of water in ZCNS-1/2 increases, while oxygen ions and metal–oxygen bond are relatively reduced, suggesting that MOF-derived hollow NSAs endow ZCNS-1/2 with relatively stable structure, so its surface is less susceptible to oxidation relative to CNS.The performance parameters of all samples in HER process were mainly measured by using a three-electrode system in 1 M KOH solution at room temperature and recorded in Table S3. Linear sweep voltammetry (LSV) curves in HER process show that the performance of the parallel group samples (ZCNS-r, r = 1/3, 1/2, 1) is generally better than that of the control group sample (CNS). The ZCNS-1/2 performs best in the parallel group samples (Fig. 5 a), suggesting that the addition of Zn2+ ion can greatly enhance the catalytic activity of CNS and achieve the best electrocatalytic performance by adjusting the molar ratio of Zn2+ and Co2+ ions. HER activity of ZCNS-1/2 is also significantly better than that of NF, it's worth noting that a certain gap exists at low current density while surpasses at high current density electrode compared to the Pt/C (Fig. S3a), far beyond other HER catalysts recently reported (Table S4). The LSV curve of ZCNS-1/2 are almost the same with or without urea, illustrating that urea may have little effect on cathodic hydrogen evolution (Fig. S3b). Double-layer capacitance (C dl ) is obtained by fitting the data on the cyclic voltammetry curves (CV), and is applied to subsequently estimating the ECSA of materials. CV curves of all samples for HER progress was plotted in Fig. S4. As expected, the C dl value of ZCNS-1/2 (54.69 mF cm−2) overtops that of ZCNS-1 (45.26 mF cm−2), ZCNS-1/3 (32.81 mF cm−2) and CNS (21.64 mF cm−2), suggesting that ZCNS-1/2 possess the maximal ECSA resulted from the hollow NSAs (Fig. 5b). The reaction kinetics of catalysts in HER process can be well revealed by analyzing the corresponding Tafel slopes. There are generally two main steps for the HER in alkaline solutions, that is, the Volmer electrochemical hydrogen adsorption: H2O + e →H ads + OH−, and the Tafel reaction (chemical desorption: Had + Had→H2) or Heyrovsky process (chemical desorption: Had + H2O + e→H2+OH−). As shown in Fig. 5c, a lower Tafel slope of 42.24 mV/dec indicates that ZCNS-1/2 undergo the Volmer–Heyrovsky reaction process with a fast catalytic kinetics (H2O + Hads + e → H2 + OH−), while a higher Tafel slope of 190.51 mV/dec indicates that CNS undergo the Volmer reaction process with a sluggish reaction kinetics (H2O + e → Hads + OH−) [64]. The resultant elertrocatalysts display different Tafel slopes for the HER, demonstrating the rate-determining steps and reaction pathways was different for water reduction reactions over these catalysts, which is in accord with previously reported Ni-based HER catalysts. Impressively, ZCNS-1/2 requires the smallest overpotential of 97 mV and 215 mV to drive the same current density of 10 mA cm−2 and 100 mA cm−2, respectively (Fig. 5d). In order to better evaluate the activity of catalyst, the turnover frequency (TOF) was also considered (see supporting information for details). The TOF of ZCNS-1/2 reaches up to 0.021 s−1 at the overpotential of 250 mV, which is over 35-fold larger than that of the CNS, explicitly evidencing the prominent intrinsic activity of ZCNS-1/2 (Fig. S5). Electrochemical impedance spectroscopy (EIS) was measured at an open circuit potential of −1.2 V in the high frequency range from 10−1 to 105 Hz to evaluate the charge transfer rate of materials. A smaller nyquist semicircle endows ZCNS-1/2 with better electrical conductivity, which in turn leads to the excellent electrocatalytic HER activity (Fig. 5e). Stability test of ZCNS-1/2 for cathodic hydrogen evolution in 1 M KOH solution was measured by chronopotentiometry method and the result shows that ZCNS-1/2 can catalyze stably for more than 15 h under a constant current of 100 mA cm−2 without drastic voltage fluctuations (Fig. 5f). The difference of physicochemical properties of ZCNS-1/2 before and after the reaction was researched by means of XRD characterization (Fig. S6). All the above analyses demonstrate that the ZCNS-1/2 material can be used as advanced electrode to catalyze efficiently and steadily HER process.The performance parameters of all samples in OER process were mainly measured by using a three-electrode system in 1 M KOH solution at room temperature and recorded in Table S5. LSV curves was measured by the reverse scanning method to eliminate the influence caused by the forward oxidation peak (Ni2+ to Ni3+) at low current density, and the negative reduction peak near 1.3 V caused by reverse scanning corresponds to the conversion of Ni3+ to Ni2+ (Fig. 6 a) [23,65]. In addition, under an identical current density, the driving potential of ZCNS-1/2 is always smaller than that of ZCNS-1/3, ZCNS-1 and CNS, which is also much less than the counterpart of NF and RuO2/NF (Fig. S7), naturally making it become one of the best OER catalysts recently reported (Table S6). CV curves of all samples for OER process were plotted in Fig. S8 and the resulted C dl value of ZCNS-1/2 achieves the largest (42.84 mF cm−2), which is more than twice that of CNS (19.85 mF cm−2) (Fig. 6b). As can be seen in Fig. 6c, a lower Tafel slope of 100.69 mV/dec demonstrates that the ZCNS-1/2 material suffers from faster reaction kinetics for anodic oxygen evolution. For OER, ZCNS-1/2 still applies the smallest overpotential of 233 mV and 307 mV to attain the same current density of 20 mA cm−2 and 100 mA cm−2, respectively (Fig. 6d). As indicated in Fig. S9, ZCNS-1/2 attains the largest TOF of 0.014 s−1 at the overpotential of 300 mV, which easily exceeds those of ZCNS-1/3 (0.0063 s−1) and ZCNS-1 (0.0056 s−1), let alone the feeble CNS (0.0035 s−1). Nyquist curves are not perfect semicircles, but the fitted radius of ZCNS-1/2 remains minimum (Fig. 6e), manifesting that it still equips with a preferable conductivity in catalyzing the OER process. The anodic oxidation potential of ZCNS-1/2 increased from 1.546 V to 1.564 V after chronopotentiometry measurement 15 h under a constant current of 100 mA cm−2 (Fig. 6f). All the above analyses reveal that the ZCNS-1/2 material can be used as advanced electrode to catalyze efficiently and steadily OER process.The performance parameters of all samples in UOR process were mainly measured by using a three-electrode system in the electrolyte of 1 M KOH with 0.5 M urea at room temperature and recorded in Table S7. Compared with OER process, the distribution trend of measured LSV curves in UOR process behaves almost the same, but the catalytic activity of each sample was enhanced dramatically. Taking the most active ZCNS-1/2 as an example, the improved activity is digitized to a reduced potential of 221 mV at the current density of 100 mA cm−2 (Fig. 7 a), which possesses remarkable UOR activity and surpass most UOR catalysts reported so far (Table S8). CV curves of all samples for UOR process were plotted in Fig. S10 and the resulted C dl value of each sample enlarges obviously relative to the counterpart in OER process (Fig. 7b), inferring a larger ECSA may be obtained with the addition of urea. On the contrary, the Tafel slope of catalyst for urea oxidation is always lower than that of water oxidation under a similar logarithmic current gradient range, further verifying the robust reaction kinetics of UOR process (Fig. 7c). As demonstrated in Fig. 7d, the potentials of as-prepared samples was required for UOR at different current densities of 20 mA cm−2 and 100 mA cm−2: CNS (1.315 V, 1.358 V), ZCNS-1/3 (1.299 V, 1.343 V), ZCNS-1/2 (1.264 V, 1.316 V), ZCNS-1 (1.285 V, 1.332 V), and the improved activity compared to OER was recorded in Fig. S11: CNS (218 mV, 237 mV), ZCNS-1/3 (216 mV, 231 mV), ZCNS-1/2 (199 mV, 221 mV), ZCNS-1 (196 mV, 229 mV), concluding that the higher the current density, the higher the activity promotion. Not surprisingly, the TOF of ZCNS-1/2 is still the most prominent at the same potential of 1.35 V, and is 7 times higher than that of urea-free counterpart (Fig. S12). Nyquist curves on the positive Y-axis are not a semicircle but a semi-ellipse in alkaline urea solution. Meanwhile, under the same open-circuit voltage of 0.4 V, the EIS of ZCNS-1/2 from urea oxidation can even be surrounded that from water oxidation (Fig. 7e), speculating that the introduction of urea may accelerate the charge transfer rate of the catalyst. After chronopotentiometry measurement of 15 h under a constant current of 100 mA cm−2, the potential fluctuation caused by urea oxidation (10 mV) is much smaller than that caused by water oxidation (18 mV), which is mainly affected by the type of anodic oxidation and applied voltage. All the above analyses demonstrate that the ZCNS-1/2 material has better activity and stability as an advanced electrode for UOR process.Given that remarkable activity of ZCNS-1/2 in HER, OER and UOR process, it can be assembled as a favorable electrode couple (ZCNS-1/2//ZCNS-1/2) to catalyze both water electrolysis and urea electrolysis. Relevant tests were carried out with a two-electrode system in the electrolyte of 1 M KOH with or without 0.5 M urea at room temperature (Fig. S13). As illustrated in Fig. 8 a, CV curve was split into two LSV curves, among which the solid line represents the LSV curve with obverse scanning, while the dashed line represents the LSV curve with reverse scanning. It should be noted that the LSV curves of urea electrolysis almost coincide and the result of reverse scanning is better than obverse scanning, while the LSV curves of water electrolysis varies distinctly and the result of reverse scanning is worse than obverse scanning. The cell voltages of ZCNS-1/2//ZCNS-1/2 was regularly read from the relatively poor LSV curve and plotted in Fig. 8b. To reach different current densities, the required cell voltages for water electrolysis: 1.522 V@10 mA cm−2, 1.721 V@100 mA cm−2, 1.788 V@200 mA cm−2 and urea electrolysis: 1.314 V@10 mA cm−2, 1.506 V@100 mA cm−2, 1.567 V@200 mA cm−2, sequentially concluding that the difference value of cell voltage between the water electrolysis and urea electrolysis becomes larger with current density increases. Comparisons of the catalytic ability of ZCNS-1/2//ZCNS-1/2 with some representative recently reported electrode couples highlights its superior activity for both urea electrolysis and water electrolysis (Tables S9–10). The difference value (ΔV) of potential between independent anodic oxidation and independent cathodic reduction was contrasted with the cell voltages of assembled electrode couple in the same electrolyte. Whether in water electrolysis or urea electrolysis, the two are very similar at different current densities (Fig. 8c), which preliminarily verifies the stability of ZCNS-1/2. Chronoamperometry measurement ran for 15 h at 1.5 V for urea electrolysis and 1.6 V for water electrolysis without apparent current attenuation (Fig. 8d), further testifying the stability of target catalyst under a constant cell voltage. Apart from that, multi-current steps measurement ran for 5 h at a rising current density from 25 mA cm−2 to 125 mA cm−2 with a gradient of 25 mA cm−2 for urea electrolysis and from 10 mA cm−2 to 50 mA cm−2 with another gradient of 10 mA cm−2 for water electrolysis without obvious voltage fluctuation (Fig. 8e), as always proving the stability of target catalyst under different current densities.The difference of physicochemical properties of ZCNS-1/2 before and after the reaction was researched by means of some characterization methods, such as XRD, SEM and XPS. The reaction here mainly refers to the anodic oxidation reaction in the two-electrode system with chronoamperometry measurement for 15 h, including the UOR in urea electrolysis at 1.5 V and OER in water electrolysis at 1.6 V. After the reaction, the peaks of ZnS, Co9S8 and Ni3S2 are still appeared in the XRD patterns. The difference is that all the three sulfide peaks in ZCNS-1/2 after OER 15 h are almost consistent with that in fresh ZCNS-1/2, while the peak intensity of ZnS in ZCNS-1/2 after UOR 15 h decreases significantly (Fig. S14), speculating that the catalyst has been corroded during the long time UOR process. In terms of the SEM images, the hollow NSAs from ZCNS-1/2 after OER 15 h are still maintained (Fig. 9 b), of which the surface becomes rough due to corrosion, while the NSAs from ZCNS-1/2 after UOR 15 h collapses to certain degrees and the single NS shrinks in size (Fig. 9c). Combined with ICP-OES result analysis (Table S11), it may be ZnS that suffers from the corrosion effect, which is the core element of the hollow NSAs structure. The loss of ZnS in the material after UOR 15 h outdistance that after OER 15 h, leading to different changes of morphology, which all are assigned to the varying degree of current erosion on ZCNS-1/2. Although the setting value of cell voltage in urea electrolysis process is lower than that in water electrolysis process, the corresponding current density of the former (100 mA cm−2) is four times that of the latter (25 mA cm−2). In view of this, the XPS spectrum of ZCNS-1/2 after UOR 15 h is the focus of study. All the peaks in recovered survey are almost the same as fresh survey (Fig. S15a). It can be seen from the XPS fine spectrum that the main change exists in Zn 2p, S 2p and O1s peaks rather than Co 2p and Ni 2p peak before and after the reaction. The peak area of recovered Zn 2p only accounts for one third of that of fresh Zn 2p (Fig. 9d). The subtle change in Co 2p and Ni 2p peaks result from the charge transfer at the heterointerface, not only in microscopic view leads to the slight conversion of corresponding element valence states but also in macroscopic view alleviates the erosion of current to a certain extent (Fig. 9e and f). For S 2p orbital peak (Fig. 9g), the extended area of S–O peak accompanied with the reduced area of S 2p3/2 peak reveals that a small amount of S2− turned into SO caused by surface oxidation. For O 1s orbital peak, the adsorbed water molecules on the catalyst surface decreases while the metal–oxygen bond increases (Fig. S15b), suggesting that part of the metal sulfides were oxidized to metal oxides. To sum up, the loss of ZnS and surface oxidation bring about a certain collapse in the morphology of ZCNS-1/2 after the reaction, but ZCNS-1/2 still shows excellent performance in catalyzing water electrolysis at low current density and meets the needs to catalyze urea electrolysis at high current density as well.Density functional theory (DFT) calculation was performed to figure out the adsorption of water by each sulfide in ZCNS-1/2, which contributes to identify the real active site and better estimate the water electrolysis performance of target catalyst. The optimize ball-and-stick model of ZnS, Co9S8 and Ni3S2 is plotted in Fig. 10 a- c with single H2O molecular absorbed on their surface. The calculated water adsorption energy (ΔGH2O) of Co9S8, ZnS and Ni3S2 are −0.48 eV, −0.53 eV, −0.56 eV, respectively (Fig. 10d). It can be deduced that most water molecules are adsorbed on the surface of Ni3S2 in the catalytic process because of the superior adsorption feature as well as the larger proportion compared to Co9S8 and ZnS in ZCNS-1/2. As illustrated in Fig. 10e, the density of states (DOS) of Co9S8 near Fermi level reaches the maximum (8.57), which is 1.3 times that of Ni3S2 (6.51) and 2.4 times that of ZnS (3.63), indicating that Co9S8 equips with a better intrinsic metallic property. The existence of heterostructure realizes the strong combination between the larger ΔGH2O of Ni3S2 and the preferable metal activity of Co9S8, which greatly enhances the integral catalytic performance of ZCNS-1/2. Moreover, the partial electronic DOS (PDOS) of each element for ZnS, Co9S8 and Ni3S2 was plotted in Figs. S16–18. In particular, the distribution of the total state density mainly comes from the p orbitals of S and the d orbitals of Zn, Co, Ni for Co9S8, ZnS and Ni3S2. The formation of Zn–S, Co–S, Ni–S bonds stem from overlapping of p-orbital of Zn, Co, Ni, and the p-orbital of S (i.e., p–p hybridization). Combined with the previous morphology and stability analysis, small doses of ZnS are mainly used to construct and stabilize the hollow NS structure, while large doses of Co9S8 and Ni3S2 play the key role in catalytic activity and the introduction of ZnS can improve the performance of ZCNS-1/2 to a certain extent.Nevertheless, under the OER conditions, the surface composition of MOF-derived Zn–Co–Ni sulfides would be changed to amorphous oxide. The water adsorption energy of NiOOH and Zn–Co–NiOOH have also been provided (Fig. S19).We firstly reported the synthesis of a MOF-derived hollow ZCNS-1/2 NSAs on 3D porous nickel foam by dint of a facile two-step hydrothermal method. The as-obtained target catalyst affords outstanding activity and stability in HER and OER process. The water electrolysis process can drive current densities of 10, 100 and 200 mA cm−2 with cell voltages of 1.522, 1.721 and 1.788 V, respectively. Furthermore, ZCNS-1/2 also exhibits excellent UOR activity, achieving the improved activity of 199 and 221 mV at current densities of 20 and 100 mA cm−2, respectively, compared to OER process. The corresponding urea electrolysis process requires a very low cell voltage of 1.506 V to drive 100 mA cm−2, which is 215 mV less than that of water electrolysis process. Morphology and stability analysis reveals that the formation and maintenance of MOF structure mainly depend on the introduction of ZnS, while DFT calculation demonstrate that the overall electrocatalytic activity largely rely on the synergy between Co9S8 and Ni3S2. The structure design and performance optimization of ZCNS-1/2 in this experiment play an exemplary role in exploring efficient and stable catalyst for both water electrolysis and urea electrolysis.There are no conflicts to declare.This work was financially supported by the National Science Foundation of China (Grant No. 21802126).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.2021.09.007.

Water electrolysis is a promising technology to produce hydrogen but it was severely restricted by the slow oxygen evolution reaction (OER). Herein, we firstly reported an advanced electrocatalyst of MOF-derived hollow Zn–Co–Ni sulfides (ZnS@Co9S8@Ni3S2-1/2, abbreviated as ZCNS-1/2) nanosword arrays (NSAs) with remarkable hydrogen evolution reaction (HER), OER and corresponding water electrolysis performance. To reach a current density of 10 mA cm−2, the cell voltage of assembled ZCNS-1/2//ZCNS-1/2 for urea electrolysis (1.314 V) is 208 mV lower than that for water electrolysis (1.522 V) and stably catalyzed for over 15 h, substantially outperforming the most reported water and urea electrolysis electrocatalysts. Density functional theory calculations and experimental result clearly reveal that the properties of large electrochemical active surface area (ECSA) caused by hollow NSAs and fast charge transfer resulted from the Co9S8@Ni3S2 heterostructure endow the ZCNS-1/2 electrode with an enhanced electrocatalytic performance.

The sources of fossil fuels such as oil are gradually exhausted due to large consumption in field of energy over the years. Meanwhile the human need for energy is increasing continuously with the passage of time, therefore the development of new energy sources is utmost need to meet this requirement [1,2]. Most of the energy is obtained from fossil fuels, which led to the rising concentration of carbon dioxide in the atmosphere and thus resulted into global warming and serious climate change [3,4]. Hydrogen has the advantage of high energy and no pollution to the environment over other energy sources. Therefore, it has great potential to replace existing fossil fuels [5–9]. At the same time, the storage of hydrogen is a challenging task for researchers [10,11]. Ammonia borane (AB), a chemical compound, has high chemical hydrogen storage potential. It is solid under environmental conditions with excellent stability and ultra-high hydrogen content (19.6 wt%) [12–14]. Recently, a lot of research works have been done in hydrolyzing and dehydrogenation of AB to obtain free hydrogen from it [15–17]. The hydrolysis of AB is expressed as follows: (1) NH3BH3 + 2H2O → NH4 + + BO2 − + 3H2↑ The hydrolysis and dehydrogenation of AB is a slow process in absence of any catalyst, so the selection of appropriate catalyst can boost the release of hydrogen from it. Metal catalysts including transition metal nanoparticles, such as Pt, Rh, Ru, Pd, Co and Ni, etc. are appropriate options to enhance the hydrogenation process of AB. The noble metals such as Pt [18], Rh [19], Pd [20] and Ru [21–23] catalysts show relatively high catalytic hydrogen generation rate but its high cost and scarcity make it limiting for a wide range of practical applications. The non-noble metal such as Fe [24,25], Co [26,27], Ni [28,29], Cu [30,31], etc. can be used as an alternate source of precious metals. In recent years, the hydrogen production from AB by use of Co based catalysts has attracted great attention of researchers. Among the non-noble metal catalyst system, the Co catalysts express the best catalytic activity under the same preparation conditions as used for Ni and Cu based catalysts [32]. However, the aggregation of cobalt nanoparticles reduces their catalytic activity. Carbon is the best carrier for active catalysts due to its high chemical interactivity, especially in high alkaline and acidic environments and good interactions with active metals [33]. Wang's group has reported the one-step synthesis of Co nanoparticles in porous N-doped carbon (Co@N-C) and the catalytic stability of AB hydrolysis [34]. Meanwhile Zhang's group has reported a simple and efficient in-situ mosaic strategy for the preparation of mesoporous carbon catalysts co-doped with non-noble metals and nitrogen [35]. The study of Lin's group described the rapid synthesis of a catalyst encapsulated into graphitized nitrogen-doped carbon nanotubes by cobalt pyrolysis by one-pot pyrolysis [36]. Catalyst support affects the catalytic activity and stability of metal NPs [37]. Titanium dioxide (TiO2) has good photocatalytic performance, with its non-toxic nature and low cost. TiO2 is one of the catalysts suitable for environmental applications because of its strong oxidation capacity and high corrosion resistance [38]. TiO2 can be used as the carrier matrix of metal nanoparticles to improve the catalytic activity or stability of metal NPs [39].In this article, we propose an effective strategy for highly dispersing active component by the carrier and then encapsulating it in the carbon layer. The catalyst is treated in air to regulate its active components. A series of catalysts are obtained by adjusting the molar amount of Co. COTC-II exhibits the best performance in the production of hydrogen from AB. The prepared catalyst in our study has advantages of low cost, high activity and stability over other previous reported catalysts. Its magnetic property facilitates its recycling. The interaction between the support and used metal resulted in synergistic effect of Co and Co3O4, exhibiting the significant catalytic activity during the hydrolysis of AB.Resorcinol (0.64 g) was dissolved in anhydrous ethanol (3 mL), above solution was added into another anhydrous ethanol (60 mL), then ethylenediamine (0.58 g) and titanium butoxide (3.04 g) were added in order and stirred to form the mixture A. Formaldehyde (1.04 g) was added to deionized water (120 mL) and stirred to form mixture B. Mixture A was added to mixture B drop by drop to form solution C. Cobalt nitrate (2.6 g) was mixed into deionized water (20 mL) and added to solution C. Solution C was sonicated for 30 min, the mixture was stirred for 24 h in a 30 °C constant temperature water bath. The formed sol–gel was directly dried at 50 °C to obtain powder products. Powder products was heated to 800 °C at 2 °C min−1 under the protection of N2 for 1 h, and cooled naturally to room temperature. A sample of black powder was obtained. The sample was recorded as CTC-n (n = I, II, III) by adding 1.3 g, 2.6 g and 5.2 g cobalt nitrate respectively. COTC-n (Co-CoOx/TiO2@N-C) (n = I, II, III) was obtained by activating the above materials in air at 250 °C for 22 h. Co@N-C was obtained without the addition of titanium butoxide in a similar preparation process of CTC-II. COC (Co-CoOx@N-C) was obtained after air treatment of Co@N-C in air at 250 °C for 22 h.Resorcinol (0.64 g) was dissolved in anhydrous ethanol (3 mL), above solution was added into another anhydrous ethanol (60 mL), then ethylenediamine (0.58 g) and titanium butoxide (3.04 g) were added in order and stirred to form the mixture A. Formaldehyde (1.04 g) was added to deionized water (140 mL) and stirred to form mixture B. Mixture A was added to mixture B drop by drop to form solution C. Solution C was filtered and dried (50 °C) after stirring at 30 °C for 24 h. The powder obtained after drying was heated to 800 °C at 2 °C min−1 under the protection of N2 for 1 h. TiO2@N-C was prepared after natural cooling to room temperature.The crystalline phases of the prepared materials were characterized by X-ray powder diffraction (XRD, Bruker/D8-Advance, Cu Kα, λ = 1.5418 Å) in the 2θ range from 5° to 80°. The Raman spectrum was recorded on an HR Evoltion Raman Spectrometer (Horiba Scientific, France) with excitation from the 514 nm line of the Ar-ion laser at a power of about 5 mW. X-ray photoelectron spectroscopy (XPS) is recorded on a PHI Quantum SXM spectrometer (with Al Kα = 1486.6 eV excitation source), and the binding energy is calibrated by reference to the C 1s peak (284.8 eV) to reduce the charge effect of the sample. The morphology of catalysts was studied by using transmission electron microscope (HRTEM, FEI Tecnai G2 F20 S-TWIN electron microscope, operating at 200 kV). The N2 sorption isotherms were measured on surface area and pore size analyzer (ASAP2420-4MP, Micromeritics, USA) at 77 K. From the adsorption branch of isotherm curves in the P/P0 range between 0.05 and 0.35, the specific surface areas (S BET) of COTC-n (n = I, II, III) were calculated by the multi-point Brunauer–Emmett–Teller (BET) method. The pore size distribution was evaluated by the non-localized density function theory (NLDFT).Hydrogen generation was studied with the typical water displacement method. Catalyst (20 mg) is placed in a round-bottom glass flask. Then the aqueous NaOH (1 M, 10 mL) solution of AB (86 mg) was injected through constant pressure drop funnel. The flask was placed on a magnetic stirrer. In self-stirring mode, only catalyst and reaction mixture were loaded in the flask. The stirring rate was fixed at 500 rpm. An inverted and water-filled gas burette in a water-filled vessel was used to monitor the volume of the evolved H2. The H2 generation specific rates were calculated using the information in the initiating and stabilizing stages (80 mL of hydrogen generated) according the following formula: (2) r B = 80 ( mL ) [ t 140 − t 60 ] ( min ) · w c ( g ) here, r B is denoted the hydrogen generation specific rate, t 140 represents the time for 140 mL of hydrogen generation, and t 60 for 60 mL, w c is the Co weight in catalyst.The designed composite material is synthesized by high temperature calcination using sol–gel method, as shown in (Fig. 1 a). In synthesis process, the resorcinol is completely blended into the ethanol, then ethylenediamine and titanium butoxide are added dropwise to form an intermediate, which resulted in the solution change from clarification to yellow turbid liquid. The intermediate is polymerized with formaldehyde to form a phenolic resin with the addition of cobalt salt to above solution. Finally, the titanic acid and Co ions are converted by calcining in nitrogen atmosphere to prepare the N-doped carbon from phenolic resin. The growth of titanium dioxide and the aggregation of cobalt nanoparticles are effectively limited by carbon [40]. It resulted in the preparation of CTC catalysts then follow its activation in air at 250 °C for 22 h to obtain COTC. Fig. 1b revealed the XRD pattern of the prepared samples. All peaks in the XRD pattern are consistent with the data reported in the literature. It can be observed that TiO2@C-N without Co NPs elaborate a clear diffraction peak at 2theta values of 25.3° and 27.4°, indicating two phases of TiO2 (rutile and anatase). The diffraction peak of CTC-II indicates the accelerated transformation from anatase to rutile due to Co NPs, and only the diffraction peak of TiO2 rutile phase is appeared [41]. X-ray diffraction patterns utter that titanium dioxide nanoparticles at 2θ = 27.4°, 36.0°, 41.2°, 44.0°, 54.3° and 65.4° with typical peak rutile phases, corresponding to crystal surfaces of (110), (101), (111), (210), (211) and (221) (JCPDS Card No. 21-1276). The XRD patterns of Co@N-C clearly confirm the metal Co phase, with the diffraction peak at 44.2°, 51.5° and 75.8° respectively corresponding to Co (111), (200) and (220) (JCPDS No. 15-0806). The diffraction peaks of COTC-II can be observed at 2θ = 19.0°, 31.2°, 36.8°, 38.5°, 55.6°, 59.3°, 65.2°, 77.3° and 78.4° corresponding to (111), (220), (311), (222), (422), (511), (440), (533) and (622) lattice planes of Co3O4 (JCPDS Card No. 42-1467) respectively. More information about the material is obtained by Raman spectral analysis. Raman spectral measurements shown in Fig. 1c elaborate that the information about prepared catalysts. The D-peak and G-peak, the characteristic Raman peaks of C atomic crystal, are obtained around 1361 cm−1 and 1591 cm−1 respectively [42]. The D-peak represents the defects in the lattice of C atom, while the G-peak in-plane stretching vibration of sp2 hybridization of C atom. I D/I G (I = intensity) calculates the intensity ratio between D-peak and G-peak which indicate the state of C atom [43]. Two broad peaks (D and G) of carbon are identified at about 1361 cm−1 and 1591 cm−1. The I D/I G intensity ratio of the COTC-II catalysts is calculated to be 0.86, indicating that the large proportion of graphite carbon in the sample.The microstructure of CTC-II composite materials is examined by TEM images (Fig. 2 a and b), which expresses the granular structure of CTC-II catalyst. The size of single carbon spheres appears about 15–18 nm, and the particle size of Co NPs is observed at about 5–8 nm by HR-TEM image (Fig. 2c). COTC-II obtained by controllable oxidation expresses as spherical particles (Fig. 2d and e). Small particles have a positive effect on hydrolysis of AB [44]. As shown in Fig. 2f, the metal particles clearly exist in the carbon layer which prove that the core–shell structure of COTC-II. The lattice fringe spacing of 0.205 nm is matched with the (111) plane of Co (JCPDS Card No. 15-0806). The lattice fringe spacing matches with Co3O4 (311) crystal face (JCPDS Card No. 42-1467) is 0.24 nm [45]. The spacing of lattice stripes matches with rutile phase TiO2 (110) crystal face (JCPDS Card No. 21-1276) to be 0.32 nm. The element mapping images (Fig. 2g) express the even distribution of Co and Ti in the composite material, and carbon doping with N is also determined.The chemical element composition and chemical valence of typical samples CTC-II and COTC-II surface can be determined by measuring X-ray photoelectron spectroscopy (XPS). Ti, Co, O, C and N elements in the samples are easy to determine. The peaks at 796 eV (780 eV), 531 eV (530 eV), 465 eV (472 eV), 400 eV (399 eV) and 285 eV (285 eV) correspond to Co 2p, O 1s, Ti 2p, N 1s, and C 1s in CTC-II (COTC-II), respectively have been expressed in Fig. S1. The increase of O in COTC-II proves the oxidation of Co. The spectrum of Ti 2p of CTC-II is fitted as 458.3 eV and 464.0 eV. The Ti 2p spectra of COTC-II has a negative shift due to the activation (Fig. 3 a) [46]. The Co 2p spectrum of CTC-II is shown in Fig. 3b. Peak pairs at 778.7 eV (Co 2p3/2) and 793.7 eV (Co 2p1/2) are attributed to Co0 [47]. The peaks at 780.2 eV (Co 2p3/2) and 795.6 eV (Co 2p1/2) are assigned to Co2+, and the formation of Co2+ is due to oxidation of the sample surface. The peaks at 786.0 eV and 802.7 eV are shake-up satellite peaks. We can observe that the peak of Co3+ (781.8 eV) in the Co 2p spectra of COTC-II, which is accompanied by Co2+ (780.2 eV for Co 2p3/2,795.2 eV for Co 2p1/2), and the satellite peak (787.3 eV and 803.9 eV). The extent of shift in peaks express that Co element is converted to cobalt oxide in the activation process at 250 °C. The C 1s of CTC-II can be divided into three peaks of 284.8 eV (CC/C–C), 286.2 eV (C–N) and 289.2 eV (O–CO), respectively (Fig. S2) [48,49]. The C 1s spectrum of COTC-II is similar to that of CTC-II. The O 1s spectrum of CTC-II can be determined into the two peaks as –OH (530.6 eV) and absorbed water (532.1 eV) respectively [50] (Fig. S3). The N 1s peaks of CTC-II and COTC-II are decomposed into pyridine nitrogen (398.1 eV/398.1 eV), pyrrole nitrogen (399.2 eV/398.1 eV) and graphite nitrogen (400.3 eV/398.1 eV) [51,52] (Fig. S4). Table S1 shows the element information from XPS. Due to the existence of Co-NPs, ferromagnetic behavior of catalyst is determined by magnetic testing which indicates its easy recovery by external magnetic field and advantage for the catalytic reaction (Fig. S5).The nitrogen adsorption–desorption isotherms provide us more detailed sample structure information. As shown in Fig. 3c, COTC-II samples show type-IV isotherm and type-H1 hysteresis loop, and S BET values with 54 m2 g−1. COTC-II expresses mesoporous structure (Fig. 3d) for smooth hydrogen production.The hydrogen production equipment for AB hydrolysis is shown in Fig. 4 a. Firstly, hydrogen production is first carried out with different catalysts in the absence of magneton at 500 rpm and 298 K (Fig. 4b). Under the same reaction conditions, the experimental results show that TiO2@N-C has no catalytic activity for the hydrogen production of AB hydrolysis, and there is still no hydrogen production after 30 min of AB hydrolysis reaction (Fig. S6). COTC-II obtained after oxidation treatment and show relatively good catalytic activity. Among the catalysts with different proportions of Co and Ti, COTC-II expresses the highest catalytic activity (Fig. 4c). The results show that TiO2 effectively improves the performance of the catalyst (Fig. S7), samples with TiO2 show higher performance compared to those with samples without TiO2. Based on the synergistic effect of Co and Co oxide as well as TiO2 as carrier, the catalytic activity of COTC-II appeared to be better than other samples. The catalysts are obtained after calcination at different temperatures are further studied (Fig. S8). At the temperature below 700 °C, the hydrogen production performance is limited due to the low crystallinity of the catalyst. The rise in temperature of calcination causes the increase in crystallinity and catalytic activity of the prepared catalyst. The highest hydrogen production resulted at 800 °C. In order to study the catalytic performance of the sample, the CTC-II is treated in air atmosphere at 200 °C with no significant improvement in hydrogen production from AB. The performance of the catalyst is greatly improved at 250 °C after 22 h (Fig. S9). The formation of hydrogen gas at different temperatures, 25 °C–45 °C, is studied (Fig. 4d), with increases in the hydrogen production rate of COTC-II from 5905 to 15,957 mL min−1 g−1, as the temperature increases, ions and water molecules become more active which lead to a rise in the catalytic activity of catalyst. At the temperature range of 298–328 K, the Arrhenius plot of lnk versus the reciprocal absolute temperature (1/T) is obtained as straight line. The apparent activation energy (E a) of catalytic reaction is calculated by following Arrhenius equation: (3) ln k = ln A − E a / R T In the equation, k represents the rate constant, R is the ideal gas constant, the exponential factor is denoted as A, and T is the reaction temperature. According to the slope of Arrhenius, the calculated activation energy of COTC-II is 38.5 kJ mol−1 substantially the same or lower than that of other non-noble metal catalysts (Fig. 4e). The detailed comparison is shown in Table S2. In our study, the prepared catalysts retain 85% of its initial catalytic activity after its use in five cycles of hydrogen production from AB (Fig. 4f). The results prove the stability and good ability of catalyst to recycle it with high catalytic activity for many times in hydrogen production reaction from AB. The decrease in the catalytic activity to some extent after many cycles may be due to the deformation, aggregation and other surface changes occur during catalytic reaction on the surface of a catalyst.In order to study the catalytic effects under different conditions, the effects of different concentrations of AB on the catalyst are further studied (Fig. 5 a). The experimental results confirm that the hydrogen production rate is positively correlated with the reactant concentration (0.136 M, 0.272 M, 0.544 M), as the concentration of AB increases, the rate of hydrolytic dehydrogenation increases. Further analysis of factors affecting hydrolysis kinetics, hydrogen production from AB at different NaOH concentrations has been studied. The hydrogen production rate increases with the rise of NaOH concentration (Fig. 5b). Therefore, NaOH is considered as a co-catalyst and has a positive effect on hydrogen production. In order to verify the positive co-relation of hydrogen production rate with catalyst concentration, AB hydrogen production is performed at different catalyst concentrations (10 mg, 20 mg, 40 mg), as shown in Fig. 5c. As the catalyst concentration increases, the active component also increases which causes improvement in rate of hydrogen production by hydrolysis of AB. Under the same conditions, hydrogen production of NaBH4 is carried out a r B of 2667 mL min−1 gCo −1 at 298 K (Fig. 5d). The catalyst appeared to be negative for the hydrogen production of NaBH4 compared to AB.Based on the above evaluation and analysis of the catalyst, we propose a simple synergistic mechanism based on the catalyst (Fig. 6 ). First, NH3BH3 molecule and H2O molecule are adsorbed and activated on the catalyst surface. Then the B–H bond in NH3BH3 and the O–H bond in H2O are broken into radicals respectively. Finally, two adsorbed H∗ atoms form an H2 molecule and then the H2 desorbs from the catalyst surface. The remaining OH∗ radicals from H2O molecule react with NH3BH2∗ to form NH3BH2OH∗. Secondly, NH3BH2OH∗ and another H2O molecule are activated again on the catalyst surface to release two H atoms to form H2. Similar to the previous step, the remaining OH∗ from the H2O molecule and NH3BHOH∗ form NH3BH(OH)2∗ is adsorbed on the catalyst surface. Because NH3BH(OH)2∗ is unstable, one H2O molecule is released. The remaining NH3BHO∗ is activated to break the B–H bond, the O–H bond in H2O molecule is broken, and the last H2 is released. At the same time NH3BOOH∗ is formed. Due to the attraction of NH3∗ to H, NH3BOOH∗ is converted into NH4 + and BO2 −. Synoptically, during the hydrolysis of NH3BH3, the metallic active components activate both the H2O molecule and the NH3BH3 molecule at the same time, causing the O–H bond and the B–H bond to be dissociated. Two H∗ atoms form one H2 to desorb from the catalyst surface. The synergy effect between Co and Co3O4 enhances the intrinsic catalytic activity. The presence of TiO2 also has a positive effect on the catalytic activity. Rational design strategy greatly improves the catalytic performances of hydrogen production.In conclusion, a core–shell catalyst with good stability and catalytic activity for the hydrolysis of AB are reported. Carbon encapsulation limits the growth of TiO2 and increases the dispersion of Co. Catalysts with different ratios of Co and Ti are studied for hydrogen production. The discussed results express that COTC-II has the best catalytic activity in the hydrogen production of AB. The activity and stability of the catalyst are enhanced by the interaction between the metal and the support, as well as the synergy between the metal and the metal oxide. The synthesized catalyst also delivers its good catalytic activity up to 85% of its initial catalytic performance after 5 cycles of hydrogen production. The separation of catalyst from reaction mixture is quite easy due to its ferromagnetic property. In addition, the good performance of COTC-II catalyst provides a new opportunity for non-precious metal catalysts in the catalytic field for clean energy sources. At the same time, our study has developed the process of hydrogen production for new, economic and ecofriendly source to meet dire need of energy for human beings of recent era.The authors declare no competing financial interests.Financial supports from the National Natural Science Foundation of China (No. 51871090, U1804135, 51671080, 21401168 and 51471065), and Plan for Scientific Innovation Talent of Henan Province (No. 194200510019) are acknowledged.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.2020.03.012.

Ammonia borane (AB) can be catalytically hydrolyzed to provide hydrogen at room temperature due to its high potentaial for hydrogen storage. Non-precious metal heterogeneous catalysts have broad application in the field of energy catalysis. In this article, catalysts precursor is obtained from Co-Ti-resorcinol-formaldehyde resin by sol–gel method. Co/TiO2@N-C (CTC) catalyst is prepared by calcining the precursor under high temperature conditions in nitrogen atmosphere. Co-CoOx/TiO2@N-C (COTC) is generated by the controllable oxidation reaction of CTC. The catalyst can effectively promote the release of hydrogen during the hydrolytic dehydrogenation of AB. High hydrogen generation at a specific rate of 5905 mL min−1 gCo −1 is achieved at room temperature. The catalyst retains its 85% initial catalytic activity even for its fifth time use in AB hydrolysis. The synergistic effect among Co, Co3O4 and TiO2 promotes the rate limiting step with dissociation and activation of water molecules by reducing its activation energy. The applied method in this study promotes the development of non-precious metals in catalysis for utilization in clean energy sources.

The current increasing environmental awareness and the inevitable depletion of fossil fuel reserves have prompted the growth of search for renewable energy sources, with a greater increment in the biodiesel industry [1,2]. Biodiesel has been typically produced by the alkaline transesterification of triglycerides present in the vegetable oils, resulting in the co-production of glycerol in 10 wt% [3]. The actual market is still unable to consume this large surplus of glycerol, mainly because the processes of purification are expensive to be performed at a large scale [4,5]. Therefore, to develop processes to convert crude glycerol to high value-added products arising as an attractive option to consumption the glycerol generated by the biodiesel industry [6].Considering this current trend, several routes to obtain valuable chemicals such as the bio-solvents, surfactants, polymers, and the more highlighted glycerol derivatives have been proposed [7–10]. Among them, glycerol derived bioadditives has attracted very attention, mainly because these compounds can be blended to the diesel or gasoline, reducing the emission of particulate material, and improving the physicochemical properties of these liquid fuels [11–15].The reactions of glycerol with acetic acid or tert-butyl alcohol give mono-, di- and tri-substituted glycerol derivatives, which are highly valuables compounds due to their properties as fuel bioadditives [16–19]. Recently, another glycerol derivative has also attracted attention as a compound able to improve the properties of diesel fossil and gasoline; the solketal (i.e., 2,2-dimethyl-1,3-dioxolane-4-methanol). It can be used to decrease gum formation, increase the octane index, diminish viscosity, improve the flashpoint, and oxidation stability [15,20–22].Solketal, as well as ethers and esters of glycerol, are compounds synthesized through the acid-catalyzed reactions, however, some of these processes involve the use of homogeneous catalysts which are highly corrosive in nature and environmentally dangerous [23]. Some works have demonstrated being possible use homogenous acid catalysts as sulfuric acid to promote the hydrolysis of solketal to generate glycerol with high level of purity, as the requirements of food and pharmaceutical industry [24]. Nonetheless, these soluble catalysts have still serious drawbacks. The use of a homogeneous catalyst involves a tedious workup procedure to be separated from products, which result in a large generation of effluent and neutralization residues, which should be disposed into the environment [25].Acidic solids can circumvent the drawbacks of the homogeneous catalysts, therefore, they have been used in newer cost-effective and selective processes to convert glycerol to solketal; sulfonic acid resins, metal oxides, zeolites, and solid supported catalysts are only some examples [26–28]. The greater challenge of most of the heterogeneous processes is to overcome the leaching of the active phase and the consequent deactivation of solid-supported catalysts. On this sense, several works have assessed the tolerance of acidic catalysts to these challengers; while Amberlyst-36 resin was few effective and quickly deactivated, hydrate aluminum fluoride demonstrated to be a more cost effective and efficient catalyst [29]. Similarly, zeolite Beta was significantly more resistant to the presence of water than Amberlyst-15 resin [30].Keggin heteropolyacids (HPAs) are attractive catalysts and are highly active in oxidative processes or acid-catalyzed reactions [31]. Phosphotungstic acid is the strongest Brønsted acid among the Keggin HPAs; it is soluble in solvent polar and has been successfully used as a homogeneous catalyst in reactions to converting glycerol to bioadditives [32]. Nonetheless, when solid, it has a low surface area, hampering its use in conditions of heterogeneous catalysis. Therefore, they have been used as solid supported catalysts in several acid-catalyzed reactions [33,34].To circumvent the undesirable problem of leaching, the unique chemical-physical properties of HPAs can be easily manipulated with suitable tailoring of their constitution [35]. For instance, the protons exchanging by larger radium metal cations make HPA salts insoluble in polar solvents [36]. Moreover, when other Lewis acid metal cations replace the protons, their catalytic properties can be significantly enhanced toward goal-process [37,38]. Recently, Sn(II)-exchanged phosphotungstic acid salts demonstrated to be an efficient catalyst in glycerol esterification reactions, achieving high ester yielding [39,40]. The same was verified when it was used to synthesize tert-butyl ethers of glycerol [41].In this work, the protons of silicotungstic acid were exchanged by Sn2+ cations, and the salt formed was used as a catalyst in reactions of glycerol condensation with acetone to produce solketal. The catalytic activity of Sn2SiW12O40 was compared to the Lewis and Brønsted acid catalysts including other Keggin HPAs. The impacts of the main reaction variables were investigated. Insights on the reaction mechanism were performed. The reusability of the catalyst was successfully evaluated.All chemicals were acquired from commercial sources and utilized without prior handling as received. Glycerol (99.5 wt%), acetone (99 wt%), and dodecane (99 wt%) were purchased from Vetec. SnCl2 and H4SiW12O40∙n H2O, all 99.9 wt%) were purchased from Sigma Aldrich.Sn2SiW12O40 salt was prepared through a procedure adapted from literature [41]. Usually, a SnCl2 solution at a stoichiometric amount was slowly added to an aqueous solution containing the solved H4SiW12O40. The mixture obtained was magnetically stirred by 3 h at 333 K, followed by the evaporation to dryness to releasing the gaseous HCl. The solid was dried in an oven at 423 K/ 3 h.Aiming a comparison, all the characterization data of pristine silicotungstic acid were also analyzed. The infrared spectroscopy analyses were performed in a Varian 660-IR spectrometer coupled to the attenuated total reflectance accessory (FT-IR/ATR). The patterns of X-rays diffraction (XRD) of the silicotungstic catalysts were recorded in an XRD-rays diffraction system model D8-Discover Bruker using Ni filtered Cu-kα radiation (λ = 1.5418 Å), at 40 kV and 40 mA, with time counting 1.0 s, with diffraction angle (2θ) varying from 5 to 80°.The H2 adsorption/ desorption isotherms were obtained in a NOVA 1200e High Speed, automated surface area and pore size analyzer Quantachrome instrument. Prior to the analysis, the sample was degassed 5 h. The surface area was calculated by the Brunauer-Emmett-Teller equation applied to the isotherms.The strength of acidity of the silicotungstic catalysts was estimated measuring the initial electrode potential (i.e. Bel, model W3B) of a CH3CN solution containing the soluble or suspended sample. The acidic sites number was determined by potentiometric titration as follows; typically, the sample (ca. 50 mg) was magnetically stirred in CH3CN by 3 h and then titrated with an n-butylamine solution in toluene (ca. 0.025 mol L-1) [42].Thermal analyses (TG) were performed in a Perkin Elmer Simultaneous Thermal Analyzer (STA) 6000. Typically, a sample (ca. 10 mg) was heated at a rate of 10 Kmin−1 under nitrogen flow. The temperature of the TG curves varied from 303 to 973 K.The elemental composition of silicotungstate salt was confirmed in an energy dispersive spectrometry system (EDS). The scanning electron microscopy (SEM) images were taken in a JEOL JSM-6010/LA microscope. A working distance of 10 mm and 20 KV acceleration voltage were used to acquire SEM images and EDS spectra.Catalytic runs were carried out using glycerol, acetone, and an adequate catalyst. In a typical reaction, glycerol and acetone at an adequate proportion were magnetically stirred until complete solubilization at room temperature at an adequate molar ratio. After adding the catalyst, the reaction was started and carried out 4 h.The reaction progress was periodically monitored collecting aliquots at regular intervals of time, and analyzing them by gas chromatography (Shimadzu 2014, FID, Carbowax 20 M capillary column). Prior to the analysis, the samples were diluted in methanol. The temperature profile was as follows: 80 °C/ 3 min; 10 °C / min; final temperature 210 °C/ 3 min. Injector and detector were both kept at 250 °C. The glycerol conversion was calculated according to the Equation (1): (1) C o n v e r s i o n % = A s o l k c o r r e c t e d A g l y c e r o l + A s o l k c o r r e c t e d x 100 where Aglycerol, is the unreacted glycerol chromatographic peak area and Asolketal corrected is the corrected chromatographic peak area of solketal, obtained by the ratio of glycerol chromatographic peak area/solketal chromatographic peak area, injected with same concentration.The reaction products were identified by mass spectrometry analysis, performed in a Shimadzu GC 2010 gas chromatograph coupled with a MS-QP 2010 Ultra, with a carbowax capillary column (0.25 μm × 0.25 mm × 30 m), and He as the carrier gas at 2mLmin−1. The temperature program was equal to the GC analyses. The GC injector and MS ions source temperatures were 523 and 533 K, respectively. The MS detector operated in the EI mode at 70 eV, with a scanning range of m/z 50–400.The Sn2SiW12O40 catalyst was reused after a simple procedure. At the end of the reaction, the solution was vapored under vacuum to remove the excess acetone, which was recovered to be used in another reuse cycle. The remaining liquid contains solketal, catalyst, and a small amount of unreacted glycerol (i.e., quantified by GC analysis of aliquot when the reaction was stopped). After three steps of liquid–liquid extraction with ethyl acetate, the solketal was removed. A simple distillation provided the ethyl acetate, which was recycled, and the pure solketal. To the reactor containing catalyst and the unreacted glycerol, recovered acetone and fresh glycerol were added, to start another cycle of reuse. This procedure was four times repeated.The characterization of the Sn(II) silicotungstate catalyst was previously discussed in another work, where they were used in an one-pot synthesis of alkyl levulinates from biomass derivative carbohydrates [43]. Notwithstanding, all the important data obtained in characterization (i.e., infrared spectra, powder XRD patterns, porosity properties, BET surface area, EDS analyses, and measurements of the strength of acidic sites are shown in the supplemental material (Fig. 1 SM-6SM). The main characterization data and the respective conclusions are summarized as follow: • The integrity of the primary structure of HPAs catalysts were confirmed by infrared spectroscopy analysis, after to check the fingerprint region that should contain the characteristic absorption bands of Keggin anion [44]. A comparison of the infrared spectra of tin(II) silicotungstate salt and precursor HPA clearly showed that the primary structure of catalyst (i.e., Keggin heteropolyanion) was kept almost untouched after the synthesis (Fig. 1SM). • Powder XRD patterns analysis can be useful to verify if any changes happened on the secondary structure of HPA when the protons are exchanged by other cations. X-rays diffractograms of Sn(II) silicotungstate and its synthesis precursors (i.e. SnCl2 and H4SiW12O40) evidenced that secondary structure of HPA presented only a little bit changes (Fig. 2 SM); although new diffraction lines were noticed at low angle region (ca. 10°, 2 θ) of diffractogram of the salt, in general, their profile was very similar to the acid. These changes are assigned the difference between the ionic radius of hydrate protons (i.e., H3O+, H2O5 +) and the Sn2+ ions, that may affect the packaging of the heteropolyanions on the secondary structure [41]. • The crystallite size was measured of silicotungstic catalysts was measured applying the Scherrer to the most intense XRD peaks. Values varied from 24 to 42 nm for the salts, while the acid presented values of 26 nm. • The strength of acidic sites belonging to the H4SiW12O40 and Sn2SiW12O40 was estimated measuring the initial electrode potential of their acetonitrile solutions; values of 713 mV and 685 mV, were obtained for the acid and silicotungstic salt, respectively (insert on Fig. 3 SM). It allowed us to classify the acidic sites of both catalysts as very strong [42]. Through the potentiometric titration curves was possible to calculate the acid sites number of silicotungstic catalysts; 1.3 and 1.2 meq g−1 were the values obtained for the acid and slat respectively (Fig. 3SM). Literature has explained the Brønsted acidity is due to hydrolysis undergone by the metal cations coordinated to the water molecules, which result in releasing of H+ cations [39–41]. • Larger metal cation salts with an ionic radius >1.3 Å, like Cs+ ions, make water-insoluble the HPA salts, in addition, they increase its surface area [43]. Nonetheless, the Sn2+ ions included in the H4SiW12O40 have a small radius (≈ 1.3 Å). The Sn(II) silicotungstate salt almost insoluble in acetone. During the reaction, the water formed favor the solubility of salt; however, at the end of the process, the suspension should be centrifugate to give solid at side bottom of the reactor. • The MEV images reveled that silicotungstic acid has like rice grains, while the salt presented particles with greater size (Fig. 4 SM). The EDS analyses confirmed the elemental composition of metal silicotungstate salts (Fig. 5 SM). • The hydration level of silicotungstic catalysts was determined by TG/ DTG curves. While the silicotungstic acid had 14 mol of H2O/ mol of catalyst, their salt presented 7 water mol per mol catalyst (Fig. 6 SM). The integrity of the primary structure of HPAs catalysts were confirmed by infrared spectroscopy analysis, after to check the fingerprint region that should contain the characteristic absorption bands of Keggin anion [44]. A comparison of the infrared spectra of tin(II) silicotungstate salt and precursor HPA clearly showed that the primary structure of catalyst (i.e., Keggin heteropolyanion) was kept almost untouched after the synthesis (Fig. 1SM).Powder XRD patterns analysis can be useful to verify if any changes happened on the secondary structure of HPA when the protons are exchanged by other cations. X-rays diffractograms of Sn(II) silicotungstate and its synthesis precursors (i.e. SnCl2 and H4SiW12O40) evidenced that secondary structure of HPA presented only a little bit changes (Fig. 2 SM); although new diffraction lines were noticed at low angle region (ca. 10°, 2 θ) of diffractogram of the salt, in general, their profile was very similar to the acid. These changes are assigned the difference between the ionic radius of hydrate protons (i.e., H3O+, H2O5 +) and the Sn2+ ions, that may affect the packaging of the heteropolyanions on the secondary structure [41].The crystallite size was measured of silicotungstic catalysts was measured applying the Scherrer to the most intense XRD peaks. Values varied from 24 to 42 nm for the salts, while the acid presented values of 26 nm.The strength of acidic sites belonging to the H4SiW12O40 and Sn2SiW12O40 was estimated measuring the initial electrode potential of their acetonitrile solutions; values of 713 mV and 685 mV, were obtained for the acid and silicotungstic salt, respectively (insert on Fig. 3 SM). It allowed us to classify the acidic sites of both catalysts as very strong [42]. Through the potentiometric titration curves was possible to calculate the acid sites number of silicotungstic catalysts; 1.3 and 1.2 meq g−1 were the values obtained for the acid and slat respectively (Fig. 3SM). Literature has explained the Brønsted acidity is due to hydrolysis undergone by the metal cations coordinated to the water molecules, which result in releasing of H+ cations [39–41].Larger metal cation salts with an ionic radius >1.3 Å, like Cs+ ions, make water-insoluble the HPA salts, in addition, they increase its surface area [43]. Nonetheless, the Sn2+ ions included in the H4SiW12O40 have a small radius (≈ 1.3 Å). The Sn(II) silicotungstate salt almost insoluble in acetone. During the reaction, the water formed favor the solubility of salt; however, at the end of the process, the suspension should be centrifugate to give solid at side bottom of the reactor.The MEV images reveled that silicotungstic acid has like rice grains, while the salt presented particles with greater size (Fig. 4 SM). The EDS analyses confirmed the elemental composition of metal silicotungstate salts (Fig. 5 SM).The hydration level of silicotungstic catalysts was determined by TG/ DTG curves. While the silicotungstic acid had 14 mol of H2O/ mol of catalyst, their salt presented 7 water mol per mol catalyst (Fig. 6 SM).Aiming to investigate the effect of Keggin anion on activity and selectivity of tin salt catalysts we carried out reactions shown in Fig. 1. Additionally, liquid Brønsted acid catalysts (i.e., HCl, H2SO4 and p-toluenesulfonic acid (PTSA)) were also assessed.Regardless of acetone excess (ca. 1: 4 glycerol to acetone), without catalyst, no conversion of glycerol was noticed. Conversely, despite the low catalyst concentration used, in the presence of Lewis or Brønsted acid catalysts the overall selectivity–defined as the percentage ratio of the desired acetalization products (i.e., a total of isomers I and II) with respect to the conversion–was always > 97% for solketal (Scheme 1 ).It is important to note that the reaction conditions were not optimized to achieve the maximum yield. Another important point is that the Brønsted acid catalysts were used at the same H+ ions concentration; similarly, the metal salts were used with the same Sn2+ ions concentration (i.e., 0.01 mol %). Fig. 1 shows that the Sn(II) heteropoly salts had superior performance to that of Brønsted acids. Moreover, amongst them, Sn2SiW12O40 was the most active catalyst.Previously, we have exploited the activity of different catalysts (i.e., SnCl2, SnF2, H3PW12O40) in condensation reactions of glycerol with acetone at the similar reaction conditions used herein (ca. 298 K, 1:4 glycerol to acetone). The main results are summarized in Fig. 2. It is known that in general a catalyst does not affect the thermodynamic equilibrium but only changes the kinetics. Nonetheless, our intention was to compare the conversions achieved within a specific time interval (see footnote of Fig. 2), independent of equilibrium has been reached or not. After comparison, it is possible to realize that even used at the higher loads, the tin halides (i.e. (soluble) SnCl2, (solid) SnF2)), and the (soluble) phosphotungstate acid, achieved lower conversions (Fig. 2) [45–47].When the Brønsted acid catalysts are totally soluble, their activity on glycerol acetalization can be linked to their strength of acidity, which was estimated by pKa measurements in different solvents; H3PW12O40 > H4SiW12O40 > H3PMo12O40 > H2SO4 > HCl [48,49]. On this regard, we had evaluated the activity of soluble Keggin HPAs and verified that a using 1:20 M ratio of glycerol to acetone and 3.0 mol % of catalyst load, the reactions in the presence of H3PW12O40, H4SiW12O40 or H3PMo12O40 achieved conversions of 91, 39 and 41% after 2 h reaction [47]. Comparing the conversions of HPA-catalyzed reactions and those in the presence of their salts we conclude that the presence of Sn(II) ions remarkably improves the performance of acid catalysts.This synergic effect between the heteropolyanion and the tin cation was previously reported in other acid-catalyzed reactions such as glycerol and glycol etherification reactions [48,49]. The higher softness of silicotungstic anion makes him more efficient to stabilize positively charged intermediates, an aspect that besides the high Lewis acid of Sn2+ ions may be useful to explain the highest activity of Sn2SiW12O40 catalyst [41,48,49]. Moreover, the Sn2+ ions can undergo hydrolyzes, reacting with residual water or generated along the process; consequently, the H+ ions produced may itself catalyze the reaction. A possible reaction pathway is depicted in Scheme 2 .The literature has described that to be condensate with glycerol, the acetone should have their carbonyl group activated through the protonation step or polarization by the coordination to the metal cation. In this work, these two mechanisms (i.e., Brønsted and Lewis acid-catalyzed) may be operating. Recently, we demonstrated through pyridine adsorption measurements by infrared spectroscopy that the Sn2SiW12O40 catalyst has these two types of acidic sites [48,49]. Therefore, we suppose that both mechanisms may be operating (step I, Scheme 2); the activation of the carbonyl group may be triggered by protonation and or coordination to the Sn2+ ions of silicotungstic salt catalyst.Another key aspect is the regioselectivity of process; although six-membered ring compounds are thermodynamically more stable than five-membered ones, the solketal (I) was always the most selectively formed product herein. We assigned this preferential formation to an easier attack of the secondary hydroxyl group on the charged positively carbonyl group (i.e., 1a intermediate, Scheme 2) if compared to the attack of the primary hydroxyl group (i.e., 1b intermediate, Scheme 2). Moreover, Mota et al., demonstrated that the methyl group in the axial position dioxane isomer repulsively interacts with two axial hydrogens of six members ring, make him less stable than dioxolane (i.e., solketal) [52]. Regardless of the catalyst, all the reactions in Fig. 2 provided solketal with an average selectivity of 97%.The effect of the variation in the stoichiometry of the reactants was also investigated and the main results are displayed in Fig. 3. It is important to note that the effects of diffusional limitations were also assessed, performing reactions with different molar ratios at different stirring rates (Table 1SM). No significant changes in conversions of reactions were observed using three distinct levels of the stirring speed rate.Since that, the acetalization of glycerol is a reversible reaction, an increase of acetone load shifted the equilibrium toward the products, increasing both initial rate and final conversion of the reactions. Conversely, no significant changes in the selectivity were observed, and solketal was always the main product (ca. 89–97% selectivity), regardless of the excess of acetone.Nanda et al. performed a comparison of impacts of the catalyst one yield of the glycerol ketalization processes developed in both batch and continuous reactors [28]. They conclude that continuous-flow processes are more promising in large scale than batch processes. Those authors verified that since the equilibrium constant of this reaction is low, the best yields are achieved in continuous systems, where an excess of acetone was used, or the water generated is continuously removed. The best performance was achieved in the Amberlyst 36-catalyzed reactions, at a proportion of 4:1 acetone to glycerol, 298 K; nonetheless, it was achieved at continuous-flow conditions (ca. WHSV of 2 h-) and high pressures (ca. 500 psi ≈ 34 atm) [28].Due to slight enhancement obtained on the conversion with reactions at higher proportions, 1:12 was the molar ratio selected to assess the other reaction variables. The effect of catalyst load was then evaluated using this proportion at room temperature and the kinetic curves are displayed in Fig. 4.Different from the observed assessing the effect of the reactant stoichiometry, a variation on the catalyst load had a noticeable impact on the reaction selectivity. Independent of the catalyst load, the conversion and the selectivity of reactions were almost similar after 1 h run (Fig. 4).Regardless of the catalyst load, solketal was always obtained with selectivity equal to or higher than 82%. Nonetheless, during the reactions was verified that the less stable product (i.e., dioxane), even being always the minor product, had its formation more favored in the initial period of reactions, mainly when the catalyst was present in lower load (Fig. 5). It means that the reaction selectivity was under kinetic control, which was impacted by the catalyst load. When a high catalyst load was used, the most stable product was always more selectively formed since the reaction beginning. Whereas, when the catalyst was present at low loads, the reaction becomes slower, and the product kinetically favorable although always the minority, had its formation enhanced, mainly in the initial period of the process.In Fig. 6, a quick survey of literature highlights the main results achieved in HPA salts-catalyzed glycerol acetalization reactions with acetone [53-55]. The physical properties of heteropoly salt depend on the cation nature used to replace the protons of Keggin acids. In special, the solubility of the salt impacts the workup needs to separate the catalyst from the medium of reaction. For these reasons, although the HPA salts containing cations with large ionic radius are almost insoluble in the reaction, sometimes it is difficult to separate or recovery the catalyst. The cesium phosphotungstate is an example, and consequently, it has been used solid-supported [54].Although a better comparison requires to take into account the amount of supported catalyst used in the reactions and its loading in the support, it is possible realize that the Sn2SiW12O40 catalyst has a performance equal or superior to the majority of the solid heteropoly salts or solid supported shown in Fig. 6.The reusability of Sn(II) silicotungstate was also assessed. The suspension formed by the catalyst in the reaction solution requires that it should be centrifugated to be recovered and reused. To circumvent this drawback, we envisaged a simple process to recover the catalyst. In this procedure, the excess acetone is removed and reused in another run. The reactor contains unreacted glycerol, solketal, and catalyst. The addition of ethyl acetate solubilizes the solketal, which is then extracted. To the reactor containing unreacted glycerol and the catalyst, recovered acetone is added and then a new run is carried out (Scheme 3 ).It is noteworthy that even if the catalyst had been recovered by filtration, the steps of recovery of the acetone and purification of the solketal should be also performed, therefore, they are not additional but obligatory steps of the process.Therefore, following the procedure in Scheme 3, we successfully recovered and reused the Sn2SiW12O40 salt in 4 catalytic runs. No decrease in the catalytic performance was noticed. Infrared analysis of catalyst after the last recycle showed that no significant change was observed at the fingerprint region of the Keggin anion spectrum.In this work, the activity of Sn2SiW12O40 salt was assessed on the acetalization of glycerol with acetone. The catalyst was spectroscopically characterized and demonstrated that after the exchange of the protons by Sn(II) cations, no modification in the primary structure (i.e., Keggin anion) was detected. In all the runs, glycerol was majority converted to solketal. The effects of the main reaction variables were investigated. We have found that catalyst concentration affects the reaction selectivity; when lower loads are used, the six-membered ring dioxane has its formation favored, although the solketal remains even as the main product. The Sn(II) cation showed to be a key constituent of catalyst. On the other hand, among the three catalysts with different Keggin anions, that containing the silicotungstic anion was the more efficient. Finally, the catalyst was recovered and reused without loss of activity. Márcio Jose da Silva: Conceptualization, Methodology, Data curation, Investigation, Writing - original draft, Project administration, Funding acquisition. Milena Galdino Teixeira: . Diego Morais Chaves: . Lucas Siqueira: .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 for the financial support from CNPq and FAPEMIG (Brasil). This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001.Supplementary data to this article can be found online at https://doi.org/10.1016/j.fuel.2020.118724.The following are the Supplementary data to this article: Supplementary data 1

In this work, Sn(II)-exchanged silicotungstic acid salt (i.e., Sn2SiW12O40) was synthesized and evaluated as the catalyst on the acetalization of glycerol with acetone to produce solketal, a versatile bioadditive of fuel. The Sn2SiW12O40 salt was compared to the other solid Keggin heteropoly salts (i.e., Sn3/4PMo12O40, Sn3/4PW12O40), and liquid (i.e., HCl, H2SO4 and p-toluenesulfonic acid) catalysts. Amongst the catalysts assessed, it was the most active, achieving a high conversion (ca. > 99%, after 1 h reaction at room temperature) and selective (ca. 97%) toward the formation of solketal. Moreover, the Sn2SiW12O40 salt demonstrated to be more active than acid and precursor Tin (II) chloride salt, as well as other heteropoly salts and solid supported catalysts. The effects of the main reaction variables were assessed. The Sn(II) cation, as well as the silicotungstate anion, showed being essential to convert glycerol to solketal. Insights on the reaction mechanism were performed. In a simple recycle procedure, the solketal was purified, the acetone excess recycled, and the catalyst was reused without loss activity.

Selective hydrogenations of aromatic ketones into the corresponding alcohols are commonly performed by homogeneous catalysis. However, these processes have economic and environmental disadvantages. Taking into account the key role of this type of alcohols in the fine chemical industries (and especially pharmaceuticals), great efforts are performed in order to replace homogeneous by heterogeneous catalysis, achieving an adequate activity and selectivity. In order to produce the hydrogenation of aromatic ketones, noble metals as Pd, Pt, Rh and Ru are commonly used as catalysts. Several items, such as metal and support selection, metal precursor, catalyst preparation and activation methods, particle size, etc., have strong influence on activity and selectivity in these types of reactions [1, 2, 3, 4]. However, these metals have several disadvantages: high cost, low abundance and low selectivity produced by different kind of side reactions such as the aromatic ring hydrogenation, as well as hydrogenolysis of our valuable product, the intermediate aromatic alcohol. With the purpose of increasing selectivity, a second metal is added. These promoters can modify the noble metal either electronically and/or geometrically [5]. Specifically, a decrease in their hydrogenating capacity is required to achieve an increase in selectivity towards the desired product. The promoters used such as Ni, Cr and Sn, are more electropositive than the noble metal [6, 7, 8, 9]. They produce a double effect: the noble metal gets a negative charge density, Meδ-, and the active sites are diluted, consequently the hydrogenating capacity is decreased. Besides, these electronic and geometric rearrangements in the active site could produce changes in the adsorption mode of the substrate increasing the selectivity towards the desired product. However, this methodology normally leads only to an enhancement, but not a complete change in selectivity.More recently, with the advancement of nanotechnology, the chemoselective hydrogenations with nanomaterials of transition metals, such as Ni and Fe (cheaper, abundant and less toxic metals), have begun to be explored [10]. Primarily, these reports have analyzed the hydrogenation of α,β-unsaturated compounds such as acrolein and cinnamaldehyde but not of aromatic ketones. Besides, the use of pure nickel as catalyst in hydrogenation of cinnamaldehyde, both in gas and liquid phases, leads to production of hydrocinnamaldehyde [11, 12, 13]. It is well known that the hydrogenation of C=C bond is both thermodynamically and kinetically favored over the C=O bond, due to the lower C=C dissociation enthalpy (611 kJ/mol) than for C=O bond (737 kJ/mol) [14]. Therefore, these results do not represent examples of chemoselective hydrogenations in order to obtain the more interesting product: the cinnamyl alcohol. Subsequently, Malobela et al. studied the effect of nickel dispersion and they found that the turnover frequency and the selectivity to unsaturated alcohol increased when the nickel crystal size decreased in the following order: 14.5 nm < 7.8 nm < 2.8 nm [15]. In agreement with these results, Viswanathan et al. reported the cinnamyl alcohol production using Ni/TiO2 catalysts prepared by four different methods. The catalysts with the smallest nickel particle size, showed the higher cinnamaldehyde conversion and selectivity to cinnamyl alcohol [16]. On the other hand, there are only reports of hydrogenation of aromatic aldehydes and imides, which were transformed into the corresponding alcohols or amines, using iron nanoparticles supported on polymers and molecular H2 as hydrogenating agent [17].With the aim to prepare catalysts based on nickel, some compounds could be more appropriate to perform chemoselective hydrogenation of aromatic ketones, than the pure nickel. These substances should preserve nickel metallic characteristics, have moderate hydrogenation capacity and catalytic sites with particular geometries. Therefore, the spatial configuration of the adsorbed molecules and the structural characteristics of the catalytic sites would allow to “tune” the proper arrangement to reach the desired hydrogenation. Taking into account these concepts and considering that in last years, nickel phosphides catalysts have emerged as excellent hydrotreating catalysts [18, 19, 20, 21], they can have good activity in hydrogen transfer reactions, such as chemoselective hydrogenations. Nickel phosphides have metallic properties, the phosphorus presence produces a diluting effect on the Ni atoms, and they have a wide range of stoichiometries, from Ni3P to NiP3. Because of these different compositions, they have a great diversity of crystallographic structures, which produce surface sites with very diverse geometries. Therefore, there could be catalytic sites, with geometries that could be able to hydrogenate different aromatic ketones to obtain the desired products.The use of nickel phosphides as chemoselective hydrogenation catalysts in the fine chemical area is scarce and it has not been applied for hydrogenation of aromatic ketones. As an example of chemoselective hydrogenation of other type of molecules using nickel phosphide as catalyst, Carenco et al. [22] reported good conversion of terminal and internal alkynes to cis-alkenes with high selectivity using nanoparticles of Ni2P.From a complete analysis of the previous topics present work explores the possible promoter effect produced by the presence of phosphorus atoms in nickel phosphides when they are used in chemoselective hydrogenation of an aromatic ketone. We decide to study this effect on the activity and selectivity in the chemoselective hydrogenation of acetophenone (AP) to obtain 1-phenylethanol. This is a very important intermediate aromatic alcohol in the fine chemical industry and is conventionally produced by this reaction [23]. Is interesting to remark that the promoting effect on the noble metals is achieved adding a second more electropositive metal (Ni, Cr, Sn). In this case, if nickel phosphides are used as catalysts, phosphorus atoms are more electronegative than Ni. Therefore, a positive charge density on Ni atoms (Niδ+) should be expected. As a consequence, a contrary effect with respect to previous studies would be awaited. Recently, we have published results indicating that nickel phosphides nanoparticles of 9 nm are active as chemoselective hydrogenation catalysts of an aromatic ketone [24]. To our knowledge, this is the first report on this application until now. In the present work, the hydrogenation results obtained with nanoparticles of pure metallic nickel and nickel phosphides of very similar crystal size (≅ 20 nm) are compared.Considering that these type of reactions are structure sensitive [16, 17] both catalysts were prepared with monodisperse nanoparticles (NPs) pre-synthesized with the same average diameter. In this way, the catalytic results will show the specific effect of the electronic and structural differences between the metallic nickel and the nickel phosphides, without the influence of different sizes of active NPs. After the obtaining and purification of the NPs, they are supported on mesoporous silica nanospheres of about 500 nm and are used as catalysts in the hydrogenation of AP in liquid phase.In a one-pot synthesis to obtain Ni0 NPs, determined amounts of nickel (II) acetylacetonate (Ni(acac)2, 1 mmol, Sigma-Aldrich, 98%), trioctylphosphine as ligand (TOP, 0.8 mmol, Sigma-Aldrich, 97%), oleylamine as solvent and reducing agent (OA, 10 mmol, Sigma-Aldrich, 70%) were directly added into a three-neck round bottom flask fitted with a condenser and magnetic stirring. The two remaining necks were used to introduce a thermocouple with a glass sheath and Ar flow, respectively. The mixture was heated at 220 °C for 2 h. Finally, the NPs were purified and isolated precipitating the suspension with acetone and re-dispersed in n-hexane.In order to obtain nickel phosphide NPs, the same procedure was used but TOP was replaced by triphenylphosphine, acting as ligand and phosphorus source (Ph3P, 0.4 mmol, Sigma Aldrich, 99%).The NPs were characterized by X-ray diffraction (XRD), diffuse light scattering (DLS), transmission electron microcopy (TEM), selected area electron diffraction (SAED) and Fourier transformer infrared spectroscopy (FT-IR). Atomic absorption spectroscopy (AA) was used to determine the Ni content in the suspension and in catalysts. In this work only the characterization of Ni0 NPs will be reported, because nickel phosphides characterizations were previously described [24].Ordered mesoporous silica nanospheres (MSNS) were prepared following the methodology proposed by Grün et al. [25] using tetraethyl orthosilicate (TEOS ≥99 %, Aldrich) as silica source, n-hexadecyltrimethylamonnium bromide (CTMABr ≥98 %, Sigma) as template agent, NH4OH (28 % p/p, Merck) to generate an alkaline medium, absolute ethanol (Cicarelli, 99.5 %) and distilled water. All reactants were mixed under vigorous magnetic stirring using the following molar composition: 1TEOS: 0.3CTMABr: 11NH4OH: 58 EtOH: 144H2O. The reaction mixture was kept under stirring at 30 °C for 2 h. The precipitate was collected by vacuum filtration and washed with distilled water. The sample was calcined up to 550 °C in air atmosphere for 2 h, with a heating rate of 10 ºC/min in order to remove the CTMABr. The solid was characterized by N2 adsorption at -196 °C, scanning electron microscopy (SEM) and TEM.The Ni0 and nickel phosphides NPs catalysts were prepared by impregnation of silica nanospheres with the corresponding pre-synthetized NPs suspensions and dried at 60 °C in air during 2 h. They were called Ni-MSNS and NiP-MSNS, respectively. The volume suspensions was fixed in order to obtain a nominal Ni loading of 5 % wt/wt.With the purpose to eliminate the surfactants (TOP or Ph3P) from the surface of the NPs, both solids were washed three times with CHCl3. The catalysts were characterized by TEM. Before their use in the reaction, they were reduced in H2 flow, heating at 10 ºC/min up to 500 °C and maintaining at this temperature during 2h. Afterwards, TEM micrographs were obtained in order to verify that the reduction treatment did not produce sintering of the NPs.The number of the catalytic surface sites of Ni-MSNS and NiP-MSNS was titrated by H2 and CO chemisorption, respectively. Besides, the volumetric oxidation technique was used to evaluate the oxygen total quantity needed to re-oxidize the reduced Ni-MSNS catalyst.XRD patterns were recorded using a standard automated powder X-ray diffraction system (Philips PW170, the Netherlands) with diffracted-beam graphite monochromator, using Cu Kα radiation (λ = 0.15406 nm) in the range 2θ = 30-80° with steps of 0.05° and counting time of 6 s/step. Besides, a diffractogram of MCM-41 nanospheres at low angles (2θ = 1.0–10.0°) was obtained with laboratory beamline Xenocs (Model Xeuss 2.0, France). This equipment has the capability to run simultaneous small and wide-angle X-ray scattering measurements (SAXS-WAXS). The present test was made in WAXS mode.The size distribution of the Ni NPs in suspension was obtained with a Zetasizer Nano (Nano ZSizer-ZEN3600, Malvern, U.K.) commercial equipment at room temperature. The light source was a helium/neon laser (λ = 632.8 nm) and the light scattering was measured at scattering angle of θ = 173°.A Philips CM 200 UT microscope (the Netherlands) equipped with an ultra-twin objective lens was used to obtain the TEM and HRTEM (high resolution transmission electron microscopy) images. A LaB6 filament operated at 200 keV was the electron source. In the high-resolution mode, the nominal resolution was of 0.2 nm. The micrographics were acquired with a CCD digital camera. A commercial program for image treatment was used to adjust linearly the illumination and contrast. Besides, electron diffraction of selected area (SAED) was obtained. Statistics on particles were done with the program Image J 1.43U. Particle size is given as the geometric average size ±standard deviation of the largest particle dimension. In all cases, more than 100 measurements were averaged, sampling in different regions of the sampler holder.A Philips 505 (the Netherlands) microscope was used for SEM analysis.A FT/IR Jasco spectrometer (model 4200, Japan) equipped with a PIKE diffuse reflectance IR cell with a resolution of 1 cm−1 was used. From 200 to 400 scans were accumulated in each case.Textural properties as specific surface area (Sg), specific pore volume (Vp) and pore diameter (Dp) of MSNS were measured with a Micromeritics ASAP 2020 V1.02 E device (U.S.A.).Measurements of the surface Ni atoms were made on the catalysts in conventional static volumetric handmade equipment. Both isotherms were measured at 50 °C, with H2 as titration reactant for Ni-MSNS and CO for NiP-MSNS. Before acquisition of isotherms, the catalysts were reduced in situ as it was previously described. After H2 chemisorption test, Ni-MSNS catalyst was degassed and completely oxidized with a known amount of pure O2.AP hydrogenation reaction was carried out in a stirred autoclave reactor at 1MPa H2 pressure and 80 °C, using 0.25 g catalyst and n-heptane as solvent. Before the catalytic test, the solids were activated in pure H2 following the procedure already described. The operative conditions for the catalytic tests were specifically chosen to avoid mass transfer control. The reaction evolution was followed by gas chromatography in a GC Varian 3400 chromatograph (the Netherlands) equipped with a capillary column of 30 m CP wax 52 CB and FID. The identification of reaction products was accomplished by GC/MS using Shimadzu QP5050 equipment (Japan).The characterization of nickel phosphide NPs was previously reported [24]. Briefly, a mixture of Ni12P5 and Ni2P NPs was obtained, where each NP is monophasic. The major phase is Ni12P5. About of 87% molar of the mixture corresponds to this compound. In this article the geometric average size of the NPs, determined by TEM, was of 9.6 ± 0.2 nm. In the present work the molar ratio of reactants: Ni(acac)2: OA: Ph3P was modified, decreasing the Ph3P quantity from 0.8 to 0.4 mmol, while quantities of Ni(acac)2 and OA were maintained equal. This change was performed in order to obtain NPs with an average diameter similar to that of the Ni0 NPs. In this way, an average value of 15.1 ± 0.6 nm was obtained by TEM. As it will be explained below, this size is nearly equal to the diameter of the Ni0 NPs. Therefore, possible effects of the NPs sizes on activity and selectivity of the catalysts should be avoided [2].In Fig. 1 XRD diffractogram of the Ni NPs, with peaks at 2θ = 44.3, 51.3 and 76.4° is shown. The position of the three peaks are in good agreement with (1 1 1) (2 0 0) and (2 2 0) crystallographic planes of a face-centered cubic (f.c.c.) unit cell, typical of metallic Ni [PDF 88–2326]. The broadening of the diffraction peaks is characteristic of very small NPs. At first sight, the presence of nickel oxide phase cannot be ruled out completely because the strongest peak of NiO appears at 2θ = 43.3° [PDF 89–7390]. Due to the considerably broadening of the Ni0 peaks, could be possible that the left side of 44.3° signal hidden a very small peak of NiO. However, the absence of any distinguishable peak at other characteristic positions of nickel oxide, such as 2θ = 37°, indicates that the presence of this phase is negligible. Therefore, we can conclude that the synthesis procedure used led to the obtaining of Ni0 NPs.In order to obtain information about the average size diameter and the monodisperse character of the Ni0 NPs distribution, DLS measurements were performed. They are easily and quickly performed and provide significant statistical information. However, this technique measures the hydrodynamic diameters, this mean, the real NPs diameters plus the thickness of the NPs coverage. Depending on the difference of the refractive indexes between the solvent used in the suspension and the NPs surfactant, a significant or a negligible disagreement, in comparison with the real diameter, can be observed [26]. The results obtained by this technique show that the Ni0 NPs suspension is monodisperse (polydispersity index <0.06) and they have an average diameter of 19 ± 6 nm.The Ni0 NPs were characterized using TEM and SAED techniques (Fig. 2 A and B). Fig. 3 show the histogram obtained counting 115 NPs. It was fitted using a log-normal distribution considering that particles lower than 20 nm present this class of size distribution [27]. The statistical parameters obtained from the fitting showeda geometric average diameter of 16.0 ± 0.2 nm. Comparing this value with that obtained by DLS, and considering the standard error, there is a good agreement between both techniques. In order to confirm that the NPs obtained are of Ni0, the SAED was acquired and analyzed. The lattice spacings measured from the rings of the diffraction pattern (Fig. 2B) were: 0.200, 0.167 and 0.118 nm. They are in very good agreement with the known lattice spacings for Ni0 bulk: 0.199, 0.173 and 0.122 nm. Therefore, this technique confirms the assignment performed by XRD.To determine if the OA and TOP remain on the NPs surface, the FT-IR spectrum was obtained (Fig. 4 ). The NPs suspension was mixed with KBr mechanically and then dried. This procedure was performed in order to avoid the overlapping of the support bands with the ones belonging to OA and TOP. In Table 1 are shown the detected peaks and its assignations. The bands corresponding to symmetric and asymmetric stretching vibrations of (C–H) and bending vibration of (CH3) can be assigned to the alkyl groups of OA and TOP [28, 29, 30]. It is no possible to distinguish between both compounds from these signals. However, the bending vibrations of (=C–H) (-C=C) and (-N-H) are exclusive of OA, and their presence on the NPs surface could be confirmed [28, 29, 30, 31]. On the other hand, the presence of other bending bands (CH3) and various stretching (C–P) bands in the range of 1159–1023 cm−1 indicates that TOP is also adsorbed on the surface of Ni NPs [29, 30]. The stretching peak of the carbonyl group at 1724 cm−1 appears as a consequence of the reaction between OA and acetylacetonate groups [32]. Finally, the band at 1080 cm−1 could be assigned to the stretching -P=O bonded at a surface nickel atom [33]. These species would be produced by TOP oxidation during the purification process of the NPs and its handling to prepare the sample to obtain the FTIR spectrum, because these steps were performed in air atmosphere.A similar result was found with the nickel phosphide NPs, but instead of TOP, Ph3P was detected on the surface [24].The silica support obtained is made of nanometric spheres of an average diameter of 530 ± 8 nm with interparticular channels. In Fig. 5 A a SEM micrograph is shown. In order to estimate the average value of these channels, the textural properties of MSNS before calcination were measured. The values obtained were: BET specific surface area: 17 m2/g, average pore diameter, from BJH method: 10 nm and pore volume: 0.02 cm3/g. After calcination these values were: BET specific surface area: 1067 m2/g, average pore diameter: 2 nm and pore volume: 0.6 cm3/g. The textural values of the MSNS after calcination were typical of a mesoporous ordered silica (MCM-41). The hexagonal ordering was checked by XRD at low angles (Fig. 5B). Besides, in TEM image (Fig. 5C) this arrangement can be observed. It must be highlighted that the presence of the mesopores typical of MCM-41 (about of 2 nm of diameter) are not useful to locate Ni0 and nickel phosphides NPs inside of them because of steric hindrance. However, the CTMABr must be added within the gel synthesis in order to obtain the SiO2 nanospheres. On the other hand, as it can be seen in the SEM micrograph, there are interparticular macropores with different sizes depending on the nanospheres packing. The sizes of these macropores change from 90 × 80 nm to 450 × 400 nm. The nanospheres and the macropores between them afford the adequate support to anchorage the NPs, as we will describe below.The silica support was impregnated with Ni0 and nickel phosphides pre-synthetized NPs suspensions, respectively. As it can be seen in TEM micrographs neither agglomeration nor changes in NPs size were detected (Fig. 6 A and reference [24]). As it can be seen in Fig. 6A, Ni0 NPs are located preferentially on the surface of the SiO2 nanospheres, but some of them are placed inside the interparticular pores (one of them is highlighted with a red circle). The Ni loadings of both systems, determined by AA, are shown in Table 2 .The HRTEM analysis of the supported Ni0 NPs shows that they have a “core-shell” structure (Fig. 6B). A “shell” thickness of about 3.5 nm was measured in the micrograph. On the other hand, using the inverse Fourier transform of HRTEM images of this “shell”, an average lattice spacing of 0.27 nm was obtained (Fig. 6C). As a consequence of the small thickness of this “shell”, few diffraction points can be selected to produce the inverse Fourier transform. Therefore, not many crystalline distances can be measured in order to obtain an average value. In spite of these constraints, clearly the value of 0.27 nm cannot be assigned to Ni0. Instead, it has a good coincidence with (1 1 1) diffraction plane of NiO with f. c.c. crystalline structure. This is an interesting result because we must remember that the surface of the NPs is covered with OA and TOP. Therefore, during solvent elimination (hexane) at 60 °C in air, this layer of organic molecules cannot inhibit the approach of atmospheric O2, leading to the NiO “shell” production.Considering the previous results, before using the supported NPs as catalysts, two processes were necessaries. In both catalysts the organic layer (phosphorous oxidized species -P=O) was eliminated by washing and reducing with hydrogen flow. This last treatment also eliminated the NiO “shell” in the catalyst with Ni0 NPs.To perform the first step, three washes of the supported catalysts with CHCl3 were done, following the method proposed by Senevirathne et al [33]. In order to check the efficiency of the procedure, the same mixture of (Ni0 NPs + KBr) used to obtain the FT-IR spectrum of Fig. 4, was washed with CHCl3 and a new FT-IR spectrum was obtained. In Fig. 4 it can be seen that, this treatment partially eliminates the OA. In this way, the δ(-C=C) band have disappear completely, but the other bands are visible yet. On the other hand, all bands assigned to TOP were clearly detected. This result proves that the adsorption of TOP is stronger than that of OA on the Ni0 NPs surface.Considering that we used the same sample, the peak intensities between non-washed and washed sample can be qualitatively compared. As consequence, it can be concluded that the band assigned to the stretching -P=O, bonded at a surface nickel atom, increase its intensity in a significant way after washing with CHCl3. Therefore, handling the sample in air atmosphere, increase TOP oxidation.After the partial elimination of the organic layer both systems: Ni-MSNS and NiP-MSNS were reduced as it was previously described. In Fig. 7 A, the micrograph reveals that Ni0 NPs sintering did not occur during this treatment. On the other hand, in Fig. 7B and C HRTEM image of one Ni0 NPs and their corresponding inverse Fourier transform are shown. Because of the reduction treatment is evident that the “core-shell” structure disappeared and the crystalline spacing determined by inverse Fourier transform (Fig. 7C) is of 0.23 nm. Clearly, this value is lower than that assigned to (1 1 1) diffraction plane of NiO with f. c.c. crystalline structure (Fig. 6C).On the other hand, it is higher than Ni0 spacing assigned to (1 1 1) planes. There are several reports in which an increasing of the interplanar distances in NPs, in comparison with the bulk value, has been detected. Thus, Winnischofer et al. [34] found a shifting to higher value crystalline spacing for the (1 1 1) plane in Ni0 NPs with f. c.c. crystalline structure. These authors considered that, most of the metals with f. c.c. crystalline structure and nanometric sizes, exhibit axes with five-fold symmetry. This kind of structure is forbidden in bulk crystals and led to NPs with icosahedral or decahedral shapes. These types of particles are known as “multiply-twinned particles”, and this structural distortion would produce the increasing of the interplanar distances.On the other hand, the same mixture of (Ni0 NPs + KBr), used to obtain the FT-IR spectrum after three washes with CHCl3, was reduced using identical conditions to Ni-MSNS. In Fig. 4, the FT-IR spectrum is shown. Weak bands at 1500 and 1460 cm−1 -corresponding to bending vibrations of (CH3) groups of TOP- and a wide signal in the range of 1159–1023 cm−1 -assignable to stretching of (C–P) bonds of TOP- were detected.Also, the band at 1080 cm−1, corresponding to stretching of -P=O, bonded to a superficial nickel atom, was observed. Strikingly, we can conclude that, after the treatment in pure H2 flow during 2 h at 500 °C, there are remains of TOP and oxidized TOP on the surface of the Ni0 NPs. This fact is undesired because both species will block a certain number of active sites. Table 2 lists the H2 and CO chemisorption results for the Ni-MSNS and NiP-MSNS, respectively. Besides, the metallic dispersions and the crystallite sizes obtained from the corresponding chemisorption measurements are reported. Assuming that the Ni0 NPs have a spherical geometry, the particle size can be estimated from the equation d AV = 101/D, where d AV represents surface-weighted average crystallite diameter in nm and D the metal dispersion, in % [35]. Using this equation and the metallic dispersion calculated from the H2 chemisorption value by assuming a stoichiometry of one H atom per surface metallic atom, the Ni NPs size was calculated (Table 2). As it can be seen, this value is approximately five times larger than the size determined by TEM (101 nm vs 19 ± 6 nm, respectively). Clearly, this discrepancy has its origin in the very low H2 chemisorption value obtained. This experimental fact is coherent with the presence of TOP fragments that remain on the Ni0 NPs surface, as was detected from FT-IR spectrum. A similar procedure was followed to evaluate the average size of the nickel phosphides NPs from the CO chemisorption result, assuming spherical geometry and applying the equation d AV = 6nf/ρL. Here, f is the weight fraction of the nickel phosphides in the catalyst, n is the average surface metal atom density (atoms/cm2), ρ is the nickel phosphide density (g/cm3) and L is the metal site concentration obtained from CO chemisorption by assuming one CO chemisorbed molecule per surface metal atom (atoms/gcatalyst) [36]. Bearing in mind that the NPs suspension used to impregnate the support is a mixture of Ni12P5 and Ni2P, where each NPs is monophasic, we used weighted averages ρ and L values considering the molar composition previously mentioned. The density values used were: ρNi12P5 = 7.53 g/cm3 and ρNi2P = 7.35 g/cm3 [37] and the L values were: LNi12P5 = 1.21 × 1015 atoms/cm2 and LNi2P = 1.01 × 1015 atoms/cm2 [38]. As it can be seen in Table 2 there is an excellent agreement between the calculated d value from the CO chemisorption test and that obtained by TEM. From this result, we can conclude that the surface of the nickel phosphides NPs was properly cleaned during the reduction step and the estimation of the surface metallic atoms obtained from CO chemisorption would produce a reliable TOF number for the catalytic reaction.On the other hand, we can determine from volumetric oxidation test that a very high percentage of Ni reduction was reached in Ni-MSNS catalyst (Table 2). Probably, the small non-reduced quantity remains as Ni2+ diffused inside the walls of the SiO2 support. This process could occur as consequence of the strong interaction between the NiO “shell” of the NPs and the SiO2 support. During the reduction step two parallel and competitive processes could take place: the NiO reduction and the diffusion of Ni2+ ions inside the SiO2 lattice. The first step would be predominant and a 91 % of the total Ni loading is reduced to metallic state. Table 2 shows that, Ni-MSNS catalyst reached the higher AP conversion at 300 min of reaction time with a value of 31 %. At this time, NiP-MSNS only reach 17 % of conversion. Notwithstanding, after 420 min the conversion value of this catalyst is 27 %. Therefore, the AP hydrogenation process takes place more slowly in the catalyst with nickel phosphide NPs. The phosphorus atoms that surround the nickel atoms would produce a diluting effect on the nickel assembly, decreasing the hydrogenation velocity. The corresponding TOF numbers evaluated at 300 min (Table 2) reflects this experimental fact. Is important to remark two aspects about the TOF of Ni-MSNS: - the presence of TOP fragments on the surface of the Ni0 NPs could block active sites. Therefore, it is possible that higher conversion could be obtained if a complete elimination of the surfactant could be achieved, - by the same reason, the quantity of the H2 chemisorbed would be underestimated. As consequence, TOF number would be overestimated. the presence of TOP fragments on the surface of the Ni0 NPs could block active sites. Therefore, it is possible that higher conversion could be obtained if a complete elimination of the surfactant could be achieved,by the same reason, the quantity of the H2 chemisorbed would be underestimated. As consequence, TOF number would be overestimated.In order to determine if these catalysts are chemoselective to hydrogenate the carbonyl group of the AP to produce 1-phenylethanol, the selectivities at the same level conversions (about 30 %) were evaluate (Fig. 8 ). Both catalysts have a very high selectivity to this product. We will analyze these results taking into account the two possible adsorption modes of carbonyl groups on the surface of metal transitions: η1(O) and η2(C,O). For AP hydrogenation with a Pt/SiO2 catalyst, Chen et al. [39] proposed that in η1(O) mode the coordination happens between the oxygen of the carbonyl group and one metallic site and the aromatic ring remains parallel to the metal surface. Instead, in η2(C,O) mode, the coordination takes place between π-electrons of C=O and two neighbors surface metallic sites [39]. Considering that the carbon atom of the carbonyl group has sp2 hybridization, the aromatic ring is tilted with respect to the metallic surface. This configuration would inhibit the phenyl group hydrogenation and high 1-phenylethanol selectivity could be obtained. A similar process would occur with the Ni-MSNS catalyst. Comparing the interatomic Ni–Ni distance (0.249 nm) with the length of the double bond C=O (0.120 nm), in order to get bridge adsorption, the carbonyl bond would be weakened and could be hydrogenated easily.When nickel phosphides are used as a catalyst to hydrogenate AP some important differences respect to pure metal must be considered. Thus, in these phases, P atoms have higher electronegativity than Ni atoms. As consequence, they can be represented as: Pδ− and Niδ+, respectively. The Niδ+ surface atoms behave as Lewis acid sites, attracting the atoms with negative charge density of the AP. Besides they work as metallic sites for hydrogenation [40]. Other difference between these compounds and Ni0, is the presence of charge accumulations along several bonds within the Ni coordination polyhedron surrounding the P atoms as it was demonstrated using the density functional theory [37]. Therefore, it should be unlikely that the AP can be adsorbed in η2(C,O) mode because charge accumulations would repel the π-electrons of C=O. In bibliography has been proposed that the only intermediate of adsorption to produce the hydrogenation of carbonyl molecules, when transition metals are used as catalysts, is η2(C,O) mode [39]. However, considering our experimental results, we assume that when nickel phosphides are used as hydrogenation catalysts, the Niδ+ surface atoms attract the oxygen of the C=O group and, at the same time, a strong repulsion is produced between the aromatic ring and the negative charge density, accumulated on the surface by the P atoms. As consequence, the bond C=O is weakened, and the hydrogenation is possible but through η1(O)-like mode as intermediate. Following this train of thought, nickel phosphides could change the mechanism of the chemoselective hydrogenation of the carbonyl group in AP and the intermediate similar to η1(O) would be reactive to produce 1-phenylethanol. However, this adsorption mode would be less reactive in comparison with η2(C,O) mode and this would be a second reason (besides the diluting effect produced by phosphorus atoms presence) that would explain the slower hydrogenation with nickel phosphides catalysts with respect to metallic nickel.Monodisperse pre-synthetized NPs of Ni0 and nickel phosphides with the same average diameter (16.0 ± 0.2 nm vs. 15.1 ± 0.6 nm respectively) were used to prepare two “quasi-model” catalysts. Both NPs species were deposited on nano-spheres of MCM-41 with an average diameter of 530 ± 8 nm. The textural properties of this support were adequate to inhibit agglomeration and sintering processes during impregnating, washing and reduction steps. In this way, we have obtained two supported and activated catalysts with the same average NPs diameter. This structural characteristic allowed performing the comparison of the catalytic results without misleading produced by crystal size effects of the active species.Both catalysts were tested in hydrogenation of AP and they showed a very similar final conversion of this compound (≅ 30 %) but nickel phosphides present a lower reaction velocity than Ni0. On the other hand, when the selectivities were compared at similar conversion levels (≅ 30 %), in order to avoid some influence of this parameter, both catalysts showed a very high selectivity to 1-phenylethanol (the desired product) of about 95 %. Therefore, we can conclude that the only catalytic difference between both systems would be the hydrogenation reaction velocity. It is necessary to remark that an optimization of the reaction operative conditions was not performed. Thus, it could be possible to get similar reaction velocities if higher reaction temperature is used.On the other hand, if geometric and electronic surface properties of Ni0 and nickel phosphides are compared, important differences appear. Thus, the situation with metallic nickel would be similar to other transition metals: AP could be adsorbed through η2(C,O) mode and the chemoselective hydrogenation occurs successfully. Instead, in nickel phosphides surface there are zones with great negative charge accumulations along some Ni–P bonds. Besides, the electronegativity differences between Ni and P produce charge densities on both atoms: Pδ− and Niδ+. This complex electronic distribution would produce a strong electrostatic repulsion between some areas of the surface of nickel phosphides and the phenyl group and π-electrons of C=O of the AP. As consequence, we propose that AP only could be adsorbed on top Niδ+ atoms through the oxygen atom of the carbonyl group. That means, the AP would be adsorbed with a mode similar to η1(O) as intermediate. Previous results have shown that through this intermediate, the hydrogenation of the carbonyl group cannot occur if transition metals are used as catalysts. Instead, if nickel phosphides are used, we suppose that the AP would be adsorbed through its oxygen to a Niδ+ atom but, at the same time, the molecule would be repelled far away from the surface due to the strong electrostatic repulsion generated between them, as it was previously described. In this situation, the C=O bond would be weakened (by a different reason that in the transition metals case) and it could be hydrogenated.Finally, is interesting to emphasize that nickel phosphides have a very wide range of compositions from Ni3P to NiP3. Among them there are great structural and electronic differences which will produce very diverse catalytic sites. Therefore, we could assume that there would be many different organic substrates, with more than one functional group, on which a chemoselective hydrogenation could take place if nickel phosphides, with different stoichiometries, are used. As consequence of these results, we can infer that due to the great versatility of these phases, they appear to be new potential chemoselective hydrogenation catalysts and new attempts to study different compositions and substrates are justified.The raw/processed data required to reproduce these findings cannot be shared at this time as the data also form part of an ongoing study.Virginia Vetere: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.Dolly Costa: Performed the experiments.Analia Soldati: Performed the experiments; Contributed reagents, materials, analysis tools or data.Jose Fernando Bengoa: Performed the experiments; Analyzed and interpreted the data.Sergio Marchetti: Analyzed and interpreted the data; Wrote the paper.This work was supported by ANPCyT (PICT N° 00549 and 0148) and Universidad Nacional de La Plata (Projects X757 and X710).The authors declare no conflict of interest.No additional information is available for this paper.

Two catalysts were prepared using monodisperse pre-synthetized nanoparticles of metallic nickel and nickel phosphides with the same average diameter. Both nanoparticles species were deposited on the same support: mesoporous silica nano-spheres of MCM-41. This support is suitable to inhibit agglomeration and sintering processes during preparation steps. Therefore, two supported and activated catalysts with the same average nanoparticles diameter were obtained. They differ only in the nature of the active species: metallic nickel and nickel phosphides. The effect of the presence of a second element (phosphorus), more electronegative than nickel, on the activity and selectivity in the chemoselective hydrogenation of acetophenone was studied. The reaction conditions were: H2 pressure of 1 MPa, 80 °C using n-heptane as solvent. With the aim to understand the catalytic results, nanoparticles, support and catalysts were carefully characterized by X-ray diffraction, diffuse light scattering, transmission electron microcopy, high resolution transmission electron microcopy, selected area electron diffraction, scanning electron microcopy, Fourier transformer infrared spectroscopy, N2 adsorption at -196 °C, atomic absorption, H2 and CO chemisorption and volumetric oxidation. Considering these results and geometric and electronic characteristics of the surface of both active species, a change in the adsorption intermediate state of acetophenone in presence of phosphorus is proposed to explain the hydrogenation chemoselectivity of nickel phospides.

With the increasing concern on the emission control of carbon dioxide, the catalytic CO2 conversion has attracted worldwide attentions. Among various options for CO2 conversion, the hydrogenation of CO2 to methanol is very promising with the rapid development of renewable energy. It can convert carbon dioxide in a large scale. The product, methanol, has broad applications as chemical intermediate and fuel. The catalyst with high activity and selectivity for CO2 hydrogenation becomes the key for the further applications. From the reported works, the copper-based catalysts are the mostly investigated ones for CO2 hydrogenation to methanol [1–4]. Noble metals like palladium, platinum, gold, rhodium, and iridium have been frequently used either as the catalyst (including the bimetallic catalyst) or as the promoter. However, a few works can be found in the literature with silver as the principal catalyst or as the promoter for CO2 hydrogenation to methanol. Silver has a relatively low price. Silver based catalysts have indeed attracted significant attentions. However, most of the reported silver catalysts are for oxidation reactions. According to the Web of Science, 16,986 papers was found with topics of Ag and catalyst since 2000 (by January 25, 2021). Among them, 6254 were for oxidation and 1255 for hydrogenation. A well-known example is the epoxyethane production from oxidation of ethylene over Ag catalyst. 33 papers were found with topics of ‘Ag’, ‘catalyst’, ‘hydrogenation’, ‘CO2’ and ‘methanol’. However, only a few papers are directly related with the supported Ag catalysts for hydrogenation of carbon dioxide to methanol [5–12] with some other papers using silver as the promoter [13,14]. Ag/ZrO2 presents a higher methanol selectivity from CO2 hydrogenation than Cu/ZrO2 [6–8], although its activity is lower than Cu/ZrO2. An increase of the t-ZrO2 phase and Ag+ content causes an increase in the rate of methanol formation [11]. A high dispersion or a small size of the silver catalyst is also favored for the selective hydrogenation of CO2 to methanol [10]. In general, the activity of the reported silver catalysts is not high for CO2 hydrogenation to methanol. Different from palladium and platinum, silver is lack of affinity toward hydrogen because of the filled d-band [15]. The further improvement in the catalyst preparation is needed in order to improve the activity of silver catalysts for CO2 hydrogenation to methanol [15].Very recently, we found the In2O3 has an intense interaction with palladium [16–19], nickel [20], rhodium [21], platinum [22,23] gold [24,25] and iridium [26,27]. This intense interaction leads a high metal dispersion and high activity towards selective hydrogenation of carbon dioxide to methanol, which has been confirmed as well by other groups [16,19,23,28]. This is especially unusual for Ni, Rh, Pt, Au and Ir catalysts, which have normally poor activity for CO2 hydrogenation to methanol. In this work, we attempt to load the silver catalyst onto In2O3. We confirm the In2O3 supported Ag catalyst is active towards the selective hydrogenation of CO2 to methanol.All the theoretical calculations were carried out using the Vienna ab initio simulation package (VASP) using a plane-wave basis set [29,30]. The projector augmented wave (PAW) method is used to describe the interaction between the valence electrons and the atomic cores [31]. The exchange and correlation energies were calculated using the Perdew-Burke-Ernzerhof (PBE) functionals [32]. Based on the results of XRD, the (111) facet was chosen because it is the most thermodynamically stable facet of this phase [33]. The In2O3(111) surface was modeled as a periodically repeated slab consisting of 72 O atoms and 48 In atoms distributed in three atomic layers and is separated by a vacuum layer of 12 Å. The oxygen vacancy is created by removing one oxygen atom from the perfect In2O3(111) surface. The supercell has a dimension of 14.56 Å × 12.61 Å × 20.04 Å. A plane-wave basis set with a cutoff energy of 400 eV and a (3 × 3 × 1) k-point grid generated with the Gamma-Centered scheme was found to give the converged results. The geometry optimization and self-consistent field convergence criterion were set to 0.03 eV Å−1 and 10−5 eV, respectively. The top two tri-layers were allowed to fully relax while the bottom tri-layer was fixed at the equilibrium position. The Ag4 cluster in the gas phase was calculated and optimized in an a = b = c = 20 Å lattice. The Ag/In2O3 model is established by placing the optimized Ag4 cluster on the defective In2O3(111) surface.The adsorption energies of intermediates M were calculated as: Δ E a d ( M ) = E M / ( A g 4 / I n 2 O 3 ) − E ( A g 4 / I n 2 O 3 ) – E ( M ) where E M / ( A g 4 / I n 2 O 3 ) , E ( A g 4 / I n 2 O 3 ) and E ( M ) represent the total energies of the Ag4/In2O3 model with the adsorbate and the clean Ag4/In2O3, the free molecule, respectively. As defined above, the negative values of the adsorption energy indicate that the process is exothermic whereas the positive values mean that the process is endothermic. The climbing image nudged elastic (CI-NEB) band method with 4–8 images was used to locate the likely transition state firstly. Then, the likely transition state was relaxed via the Dimer method. The relaxed transition state was confirmed through frequency analysis.The In2O3 support was prepared via the precipitation method. The desired amount of In(NO3)3·4H2O (2.41 g, HWRK Chem, 99.99%) was dissolved into the deionized water as the precursor solution (0.15 mol L−1). 3.50 g of sodium carbonate hydrate (Tianjin Kermel Chemical Reagent, 99%) was also dissolved into the deionized water as the precipitant solution (0.2 mol L−1). Firstly, the precipitant solution was dropped into the precursor solution with vigorous stirring under 80 °C until the pH value reached 7. The mixture was aged for another 3 h under the same condition. The precipitate was then washed with deionized water several times. After dried at 80 °C overnight, the as-prepared solid was calcined in static air at 450 °C for 3 h.The Ag/In2O3 catalyst was prepared via the deposition-precipitation method. Silver nitrate (Aladdin Industrial Corporation, Shanghai, 99.99%, metal basis) was dissolved into 50 mL deionized water. Then the as-prepared In2O3 support was added into the solution, followed by vigorous stirring for 1 h at room temperature. An excessive amount of sodium carbonate solution (1.0 g in 10 mL of the deionized water) was then added into the mixture and the formed precipitate was continuously stirred at 80 °C for 3 h. After washed and filtered with 1 L deionized water, the mixture was dried in a vacuum at 60 °C overnight. The actual loading weight of Ag species was determined by Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES) measurements.77 K N2 adsorption/desorption isotherms of the samples were measured on an Autosorb-1-C instrument (Quantachrome). The specific surface area (SBET) was calculated using the Brunauer–Emmett–Teller (BET) model. The bulk analytical composition of the samples was determined by inductively coupled plasma optical emission spectrometer (ICP-OES) measurements, using a Perkin Elmer Optima 5300DV system. Powder X-ray diffraction (PXRD) was conducted to characterize the textural structures of the samples, using a Rigaku D/max 2500v/pc diffractometer with Cu Kα radiation (40 kV, 200 mA). The scanning rate was set to 8° min−1 within the 2θ range of 10°–80°. The phase identification was made by comparison with the Joint Committee on Powder Diffraction Standards (JCPDSs). Transmission electron microscopy (TEM) measurements were performed on a JEOL JEM-2100F system equipped with an energy-dispersive X-ray spectrometer (EDX) operated at 200 kV. The sample powder was firstly suspended into ethanol and then dispersed ultrasonically for 5 min. A drop of the suspension was deposited on a copper grid coated with carbon. Temperature programmed desorption of carbon dioxide (CO2-TPD) experiments were conducted on a Micromeritics Autochem II 2920 chemisorption analyzer equipped with a Hiden HPR-20 EGA mass spectrometer (MS). 100 mg of the sample was loaded into a U-shaped quartz tube and reduced with a gaseous mixture of 10% H2 in Ar for 1 h at 200 °C and then cool down to 50 °C under flowing helium, followed by the CO2 adsorption at the same temperature for 1 h. After purged by flowing helium for 1 h to remove the physically absorbed CO2, the sample was then heated to 700 °C at a rate of 10 °C min−1. The signals of m/z = 44 and 28 were collected by the mass spectrometer. Electron paramagnetic resonance (EPR) spectra of the samples were collected at room temperature using a Bruker A300 EPR spectrometer operated at the X-band frequency. Raman spectra of the samples were collected using an inVia Reflex Renishaw Raman Spectroscopy System. The scan range is 200–800 nm with 532 nm laser as the excitation source. The laser power was set at 5 mW and the integration time is 5 s. The UV–vis absorption spectra of the samples were recorded at room temperature, using a UV-2600 UV–vis spectrophotometer (Shimadzu Corporation).The catalytic tests for methanol synthesis from CO2 hydrogenation over In2O3 and the Ag/In2O3 catalysts were performed in a vertical bed reactor. 0.2 g of the sample was diluted with 1.0 g SiC before being loaded into the reactor. Firstly, the reactor with the sample was purged by flowing nitrogen for 0.5 h at room temperature. The Ag/In2O3 catalyst was then pre-reduced by a gaseous mixture of 10% H2 in N2 at 200 °C for 1 h. The reactant mixture was introduced into the reactor until the pressure reached 5 MPa at the same temperature. The reaction was performed with a gas hourly space velocity (GHSV) of 21,000 cm3 gcat −1 h−1 and a temperature range from 200 °C to 300 °C. The products were analyzed using an online gas chromatograph (Agilent 7890A), equipped with a thermal conductivity detector (TCD) and a flame ionized detector (FID). To prevent the condensation of methanol, all the valves and lines between the reactor outlet and the GC inlet were maintained at 140 °C.The CO2 conversion ( X CO 2 ), methanol selectivity (Smethanol), and space-time yield (STY) of methanol were calculated according to the following equations: X C O 2 = F C O 2 , i n − F C O 2 , o u t F C O 2 , i n × 100 % S m e t h a n o l = F m e t h a n o l , o u t F C O 2 , i n − F C O 2 , o u t × 100 % M e t h a n o l S T Y = F C O 2 , i n × X C O 2 × S m e t h a n o l W × M where F is the molar flow rate, M is the molar mass of methanol and W is the weight of the catalyst.To investigate the electronic interaction between the Ag species and the In2O3 support, DFT calculations were performed with a model catalyst of the Ag4 cluster on the In2O3(111) with an oxygen vacancy. The loading of Ag species on the In2O3 support was firstly investigated. As shown in Fig. 1 a, the Ag4 cluster is a typical structure of the Ag cluster. The average length of the Ag–Ag bonds is 2.71 Å, which is well consistent with the literature [34,35]. Fig. 1b and c show the optimized structures of the perfect and the defective In2O3(111), respectively. The optimized structure of the Ag4 cluster supported on the defective In2O3(111) model is named “Ag4/In2O3_D” as shown in Fig. 1d. The adsorption energy of the Ag4 cluster on the Ag4/In2O3_D is −1.42 eV, which is closed to that of the Ag4 cluster on the m-ZrO2(111) surface (−1.49 eV) [35]. The average length of Ag–Ag bonds is increased to 2.75 Å due to the interaction between the Ag4 cluster and the defective In2O3 surface. To gain the insight into the electronic structure of the supported Ag4 cluster, Bader charge analysis was conducted. As shown in Table 1 , the total charge of the free Ag4 cluster is zero, indicating electrical neutrality. Due to the presence of the strong metal-support interaction (SMSI) between the Ag4 cluster and the defective In2O3(111) surface, the total charge of the Ag4 cluster is increased to +0.34 |e| [25,36,37]. This result confirms the presence of the strong electron transfer from the Ag4 cluster to the In2O3 support.The oxygen vacancies over In2O3 play a very important role in CO2 adsorption as well as CO2 activation [25]. Therefore, the CO2 adsorption on the interfacial site between the positively charged Ag4 cluster and the surface oxygen vacancy of the Ag4/In2O3_D model is studied. As shown in Fig. 2 a, the CO2 molecule can be activated on such interfacial site of the Ag4/In2O3_D model through R1 (CO2(g) + ∗→CO2∗), with the adsorption energy of −0.50 eV. The bond length of C-Oa and C-Ob are increased to 1.34 Å and 1.24 Å, respectively. It is 1.17 Å in the free CO2 molecule. In addition, the O–C–O angle becomes 123.1°. The Bader charge analysis shows that the CO2 molecule is negatively charged due to the interaction with the interfacial site of the Ag4/In2O3_D model as shown in Table 2 . The total charge of the adsorbed CO2 is −0.95 |e|. This is also confirmed by deformation charge density, shown in Fig. 2c, where the yellow iso-surface around the adsorbed CO2 is clearly seen and is assigned to the accumulation of electrons. As shown in Fig. 2b, hydrogen can be activated by the positive charged Ag species through R2 (H2(g) + ∗ + ∗→2H∗), with the adsorption energy of −0.71 eV. The average H–Ag bond length is 1.76 Å. The active H adatoms can spill over and react with the activated CO2 intermediates on the interfacial site of the Ag4/In2O3 model. All the results indicate that the electron transfer induced by the interaction between the Ag species and the In2O3 support promotes the formation of the positively charged Ag species, which can facilitate the activation of CO2 and hydrogen.Wang et al. [38] and Ye et al. [27] reported that the oxygen vacancies over In2O3 can facilitate the CO2 dissociation via the experimental and the theoretical studies, respectively. Herein, the dissociation of CO2 over the interfacial site of the Ag4/In2O3_D model through R3 (CO2∗ + ∗→CO∗ + O∗) was also calculated. The potential energy surface is shown in Fig. 3 . The distance of C atom and Ag atom is 3.08 Å in the structure of CO2∗, while it is shortened to 2.55 Å due to the formation of C–Ag bond in the transition state (TS1). Meanwhile, the C-Oa bond is slightly increased to 1.38 Å whereas the length of the C-Ob bond is decreased to 1.21 Å. The reaction is slightly endothermic by +0.10 eV with an activation barrier of 0.41 eV. The low barrier indicates that the CO2 dissociation on the interfacial site of the Ag/In2O3 catalyst is kinetically favorable, which also implies that the methanol synthesis on the Ag/In2O3 catalyst can undergo the CO hydrogenation route. Therefore, the methanol synthesis via the CO hydrogenation route on the interfacial site of the Ag4/In2O3_D model is investigated. Fig. 4 shows the potential energy surface of methanol synthesis from CO2 hydrogenation through the CO hydrogenation route on the interfacial site of the Ag4/In2O3_D model, with comparison of the RWGS reaction. All the structural parameters and adsorption energies of the reaction intermediates involved are summarized in Table S1. The details of these structures are shown in Fig. S1. The structural details of initial states (IS), transition states (TS) and finial states (FS) are also shown in Fig. S2. The reaction energies and the activation barriers of all elementary steps involved methanol synthesis on the Ag4/In2O3_D model are listed in Table 3 .After the CO2 dissociation through R3, the produced O∗ can react with the H adatom activated on the Ag4 cluster and form the surface OH∗ through R4 (CO∗ + O∗ + H∗ →CO∗ + OH∗ + ∗). The transition state is TS2. This reaction is exothermic by −0.23 eV with an activation barrier of 0.99 eV, indicating that this process is both thermodynamically and kinetically favorable. In particular, the direct hydrogenation of CO2 to HCOO∗ and COOH∗ are also examined to investigated the feasibility of the formate route and the RWGS route, respectively. As shown in Fig. S3, the hydrogenation of CO2 to HCOO∗ (R14: CO2∗ + H∗→HCOO∗ + ∗) is exothermic by −0.19 eV with a huge barrier of 1.65 eV. The hydrogenation of CO2 to COOH∗ (R15: CO2∗ + H∗→COOH∗ + ∗) is endothermic by +0.85 eV with an activation barrier of 1.35 eV. These results indicate that the direct hydrogenation of CO2 on the interfacial site of the Ag4/In2O3_D model is not feasible, compared to R4 of the CO hydrogenation route.The H adatom activated on the Ag4 cluster can react with the C atom of CO∗ through R5 (CO∗ + OH∗ + H∗→HCO∗ + OH∗ + ∗). The transition state is TS3. This reaction is endothermic by +0.32 eV with an activation barrier of 0.81 eV. Moreover, the activated H adatom can also react with OH∗ and produce H2O∗ through R6 (CO∗ + OH∗ + H∗→CO∗ + H2O∗ + ∗). The transition state is TS4. This reaction is endothermic by +0.72 eV with an activation barrier of 1.27 eV. The CO∗ and H2O∗ proceed to desorb from the interfacial site with an overall energy cost of 1.66 eV to complete the RWGS reaction. These confirm that the activated H adatom prefers to react with the C atom of CO∗ due to the lower activation barrier.Subsequently, the production of H2O∗ from the hydrogenation of OH∗ undergoes R7 (HCO∗ + OH∗ + H∗→HCO∗ + H2O∗ + ∗). The transition state is TS5. The reaction is endothermic by +0.39 eV with the activation barrier of 0.98 eV. Based on the adsorption energies of HCO∗ (−2.26 eV) and H2O∗ (−0.97 eV), H2O∗ can desorb from the interfacial site more easily than HCO∗.H2CO∗ is one of the important intermediates in methanol synthesis from CO2 hydrogenation. H2CO∗ can be produced from the hydrogenation of HCO∗ through R8 (HCO∗ + H2O(g) + H∗→H2CO∗ + H2O(g) + ∗). The transition state is TS6. The reaction is exothermic by −0.71 eV with an activation barrier of 0.98 eV, which indicates that this process is both thermodynamically and kinetically favorable.The hydrogenation of H2CO∗ produces either H3CO∗ or H2COH∗. The production of H3CO∗ (R9: H2CO∗ + H2O(g) + H∗→H3CO∗ + H2O(g) + ∗, ΔE = −0.66 eV, Ea = 1.07 eV) is much more kinetically and thermodynamically favorable than the production of H2COH∗ (R10: H2CO∗ + H2O(g) + H∗→H2COH∗ + H2O(g) + ∗, ΔE = +1.18 eV, Ea = 2.52 eV) due to the lower activation barrier and exothermic nature. The transition states of R9 and R10 are TS7 and TS8, respectively.Finally, H3COH∗ can be produced via H3CO∗ protonation (R11: H3CO∗ + H2O(g) + H∗→H3COH∗ + H2O(g) + ∗, ΔE = +0.45 eV, Ea = 1.01 eV) or the hydrogenation of H2COH∗ (R12: H2COH∗ + H2O(g) + H∗→H3COH∗ + H2O(g) + ∗, ΔE = −0.48 eV, Ea = 1.30 eV). The hydrogenation of H2COH∗ is more thermodynamically but less kinetically favorable than the protonation of H3CO∗. The transition states of R11 and R12 are TS9 and TS10, respectively. CH3OH∗ proceeds to desorb from the interfacial site with an energy cost of 0.79 eV and 0.67 eV for the CH3O pathway and the H2COH pathways, respectively. Furthermore, the potential energy surface of the CH3O pathway is always below the H2COH pathway. Therefore, methanol will mainly be produced through the CH3O pathway in the CO hydrogenation route. Fig. 5 shows the changes of CO2 conversion and methanol selectivity on Ag/In2O3 with the reaction temperature. The catalytic activity of In2O3 is shown as well for the comparative purpose. The activity data in Fig. 5a–c were collected when the reaction was carried out for 30 min. The carbon balances over both In2O3 and Ag/In2O3 catalyst are better than 98%. Only trace methane can be detected when the reaction temperature is beyond 275 °C. As shown in Fig. 5a and b, enhanced activity is achieved for CO2 hydrogenation to methanol by the loading of Ag. The CO2 conversion and the space-time yield (STY) of methanol on Ag/In2O3 are higher than those on In2O3. The STY reaches the highest value of 0.453 gmethanol gcat −1 h−1 on Ag/In2O3 at 300 °C and 5 MPa, whereas it is 0.335 gmethanol gcat −1 h−1 on In2O3 at the same condition. The STY of Ag/In2O3 is around 4 times higher than the reported STY on Ag/ZrO2 at 300 °C [11]. The product distribution of CO2 hydrogenation to methanol over the Ag/In2O3 catalyst at 300 °C and 5 MPa is shown in Table S2. Apparent activation energies for In2O3 and Ag/In2O3 were calculated based on the Arrhenius equation. As shown in Fig. 5c, the apparent activation energy of CO2 conversion of In2O3 is 101.98 kJ mol−1, which is in line with the literature [16,22,39]. The Ag loading significantly reduces the apparent activation energy of CO2 conversion to 86.44 kJ mol−1. This confirms that the Ag loading is favorable for the CO2 hydrogenation. Fig. 5d presents the stability test results of In2O3 and Ag/In2O3. The methanol STY on Ag/In2O3 maintains over 90% of its initial value after the 10-h reaction at 300 °C and 5 MPa whereas the pure In2O3 catalyst loses near 20% of its initial methanol STY after 10-h in the reaction stream. This result indicates that the addition of Ag species significantly improves the stability of In2O3. Table S3 summarizes the comparison of the catalytic activities of Ag/In2O3 with some typical catalysts.Based on the results of N2 adsorption, the specific surface area of In2O3 and Ag/In2O3 was 82.81 m2 g−1 and 89.34 m2 g−1, respectively. The Ag loading has not a significant effect on the specific surface area. Fig. 6 shows the X-ray diffraction (XRD) patterns of In2O3 and Ag/In2O3. In the following discussions, the pristine Ag/In2O3 catalyst is named as ‘Ag/In2O3–P’. The samples after hydrogen reduction are named as ‘Ag/In2O3’ and ‘In2O3-R’. The samples after the reaction at 300 °C and 5 MPa are assigned to ‘Ag/In2O3-AR’ and ‘In2O3-AR’. According to PDF#06-0416, the diffraction peaks centered at 21.5°, 30.7°, 35.5°, 45.7°, 51.0° and 60.7° are assigned to the diffractions from (211), (222), (400), (431), (440) and (622) facets of In2O3. This is in accordance with the results of Pd/In2O3 [16,17], Pt/In2O3 [22,23], Rh/In2O3 [21], Ni/In2O3 [20], Au/In2O3 [24] and Ir/In2O3 [26] catalysts. No diffraction peaks of metallic Ag or Ag2O can be observed due to the low loading weight of Ag species (0.33 wt% according to the analysis of ICP-OES) with the high dispersion over the In2O3 support. To further clarify the dispersion of Ag species, TEM analyses with the corresponding EDX elemental mapping were conducted. As shown in Fig. 7 , the EDX elemental maps show the extremely high dispersion of Ag species on the In2O3 support. The Ag/In2O3 catalyst after H2 reduction (H2/N2 = 1/9, molar ratio, at 200 °C for 1 h) and after reaction (300 °C, 5 MPa) remains the high dispersion of Ag species as shown in Fig. 7b and c, respectively. This result confirms that the interaction between the Ag species and the In2O3 support facilitates the dispersion of Ag species, which further provides much more active sites for the reaction.The previous studies have confirmed that CO2 adsorption occurs on the oxygen vacancies of In2O3. The CO2 molecule can be considered as a probe to characterize the oxygen vacancies of In2O3-based catalysts as well [22,40]. However, CO2 can be dissociated to CO on oxygen vacancies as reported recently by Wang et al. [38]. Therefore, we use a mass spectrometer (MS) to analyze the signals of temperature programmed desorption of carbon dioxide (CO2-TPD). The signals of m/z = 44 were recorded as the intensities of CO2 desorption whereas the signals of m/z = 28 were recorded as the intensities of CO desorption. As shown in Fig. 8 , the CO2-TPD-MS profiles can be distinguished into four regions. In the region (I), the CO2 desorption peaks located at the temperature below 100 °C can be assigned to the desorption of physically absorbed CO2 and CO [17,20,22,38,41]. The CO signals can be attributed to the CO2 dissociation under the process of CO2 adsorption. The CO2 desorption peaks at ca. 240 °C in the region (II) can be assigned to the surface oxygen vacancies of In2O3 induced by hydrogen reduction. Also, the intensity of the CO2 desorption of In2O3 is stronger than that of Ag/In2O3. Furthermore, the CO2 desorption peak of In2O3 at ca. 430 °C in the region (III) is attributed to the thermally induced oxygen vacancies [41], whereas the peak of Ag/In2O3 can be attributed to the CO2 desorption on the interfacial oxygen vacancy site created by the Ag–In2O3 interaction under hydrogen reduction. The CO2 desorption peak of Ag/In2O3 at ca. 410 °C exhibits a much stronger intensity. This indicates a much stronger CO2 adsorption on the interfacial site of Ag/In2O3 than In2O3, which is well consistent with the DFT calculations above. A broad CO2 desorption peak of the Ag/In2O3 sample centered at ca. 560 °C can be attributed to the thermally induced oxygen vacancies in the region (IV), which confirms the presence of the interaction between Ag species and the In2O3 support. The total amount of oxygen vacancies of Ag/In2O3 is higher than that of In2O3 according to the much stronger CO2 desorption on the interfacial oxygen vacancy site created by the addition of Ag species. More importantly, the CO desorption peak of the Ag/In2O3 sample at ca. 410 °C can be observed in the region (III). The onset temperature of CO desorption is ca. 350 °C. This result indicates that the new oxygen vacancy site created by the addition of Ag species facilitates the activation of CO2. This is also consistent with the result of DFT calculation discussed above.To gain the insight into the oxygen vacancies influenced by the addition of Ag species, the electron paramagnetic resonance (EPR) analysis was performed. As shown in Fig. 9 , all samples exhibit a signal near 3510 G with a g factor of 2.003. According to the literature, singly ionized oxygen vacancies (F-centers) are paramagnetic and are expected to yield a single EPR signal centered near the free electron g-value (2.0023) [42–44]. The trapped electron on or near an oxygen vacancy (F-center) can determine the visible light activity because the F-center provides a unique energy level [42,43], which can be characterized by the UV-vis absorption spectra below. Therefore, the stronger intensity of Ag/In2O3 affirms that the increasing oxygen vacancy sites are created by the addition of Ag species, which is consistent with the results of CO2-TPD-MS.Typically, the oxygen vacancies of In2O3 can be characterized by Raman spectra [45,46]. As shown in Fig. 10 , all the scattering features are attributed to the vibrations of the InO6 structural unit [46–48]. The peaks at around 306 cm−1 (I1) and 367 cm−1 (I2) are assigned to the bending vibration and the stretching vibration of In–O–In in InO6 octahedra, respectively. To quantify the oxygen vacancies, the integrated areas of the I1 peak and the I2 peak were calculated. The ratio of I2/I1 can be used to characterize the amount of oxygen vacancies on In2O3 [22,49,50]. For the pristine samples, Ag/In2O3 shows a higher ratio of I2/I1. This implies that the addition of Ag species can promote the formation of the oxygen vacancies of In2O3. Furthermore, the increasing I2/I1 ratios of both In2O3 and Ag/In2O3 after hydrogen reduction indicates that the oxygen vacancies can be created via hydrogen reduction. It can be seen that the difference of the I2/I1 ratio between In2O3 and Ag/In2O3 decreases after hydrogen reduction. This result indicates that Ag/In2O3 exhibits the higher stability under hydrogen reduction. Combined with the results of CO2-TPD-MS, the increasing oxygen vacancies of Ag/In2O3 can be attributed to the interfacial oxygen vacancy site created by the Ag–In2O3 interaction under hydrogen reduction. Furthermore, the amount of oxygen vacancies on In2O3 exceeds that on Ag/In2O3 after reaction at 300 °C and 5 MPa. As reported by Tsoukalou et al. [51], In2O3 itself is not very stable under the reaction conditions of CO2 hydrogenation to methanol. They claimed that the formation of metallic indium as a result of the partial reduction of In2O3 results in the deactivation, which typically occurs at elevated temperatures. Therefore, the addition of the Ag species stabilizes the structure of the catalyst under the reaction conditions.To further understand the oxygen vacancies of Ag/In2O3 induced by the addition of Ag species and hydrogen reduction, UV-vis absorption spectra was analyzed. As shown in Fig. 11 a, a strong UV absorption is observed in all samples due to the semiconductor nature of In2O3 [22]. The band gap can be determined by the absorption using the Tauc relationship. As the results shown in Fig. 11b, the band gap is 2.79 eV, 2.56 eV and 2.47 eV for In2O3, Ag/In2O3–P and Ag/In2O3, respectively. The narrowed band gap for the Ag/In2O3–P can be attributed to the oxygen vacancies created by the addition of Ag species [45,46], which is consistent with the results of EPR above. In addition, the band gap of Ag/In2O3 is further narrowed, corresponding to the formation of interfacial oxygen vacancy site created by the Ag–In2O3 interaction under hydrogen reduction [45,46]. This confirms the results of CO2-TPD-MS discussed above.In conclusion, the DFT and experimental studies confirmed the methanol synthesis from CO2 hydrogenation over the Ag/In2O3 catalyst is feasible. The intense interaction between the Ag4 cluster and the defective In2O3(111) surface makes the Ag4 cluster positively charged. The interfacial site between the positively charged Ag4 cluster and surface oxygen vacancy of the Ag4/In2O3_D model facilitates the activation and the dissociation of CO2, resulting in methanol synthesis via the CO hydrogenation route. The methanol selectivity of Ag/In2O3 reaches 100.0% at reaction temperature of 200 °C. It remains more than 70.0% between 200 and 275 °C. The CO2 conversion reaches 13.6% with the methanol selectivity of 58.2% at 300 °C and 5 MPa. The methanol STY is 0.453 gmethanol gcat −1 h−1 under the same condition. This is the highest methanol STY ever reported on Ag catalysts for CO2 hydrogenation to methanol. The catalyst characterization indicates a high Ag dispersion on In2O3. The intense Ag–In2O3 interaction promotes the formation of interfacial oxygen vacancy site with increasing amount of oxygen vacancies. The enhanced activity is thereby achieved.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 Key Research and Development Program of China (2016YFB0600902).The following is the Supplementary data to this article: Multimedia component 1 Multimedia component 1 Multimedia component 2 Multimedia component 2 Multimedia component 3 Multimedia component 3 Multimedia component 4 Multimedia component 4 Multimedia component 5 Multimedia component 5 Multimedia component 6 Multimedia component 6 Supplementary data to this article can be found online at https://doi.org/10.1016/j.gee.2021.05.004.

Silver catalyst has been extensively investigated for photocatalytic and electrochemical CO2 reduction. However, its high activity for selective hydrogenation of CO2 to methanol has not been confirmed. Here, the feasibility of the indium oxide supported silver catalyst was investigated for CO2 hydrogenation to methanol by the density functional theoretical (DFT) study and then by the experimental investigation. The DFT study shows there exists an intense Ag–In2O3 interaction, which causes silver to be positively charged. The positively charged Ag species changes the electronic structure of the metal, facilitates the formation of the Ag–In2O3 interfacial site for activation and dissociation of carbon dioxide. The promoted CO2 dissociation leads to the enhanced methanol synthesis via the CO hydrogenation route as CO2∗→CO∗→HCO∗→H2CO∗→H3CO∗→H3COH∗. The Ag/In2O3 catalyst was then prepared using the deposition-precipitation method. The experimental study confirms the theoretical prediction. The methanol selectivity of CO2 hydrogenation on Ag/In2O3 reaches 100.0% at reaction temperature of 200 °C. It remains more than 70.0% between 200 and 275 °C. At 300 °C and 5 MPa, the methanol selectivity still keeps 58.2% with a CO2 conversion of 13.6% and a space-time yield (STY) of methanol of 0.453 gmethanol gcat −1 h−1, which is the highest methanol STY ever reported for silver catalyst. The catalyst characterization confirms the intense Ag–In2O3 interaction as well, which causes high Ag dispersion, increases and stabilizes the oxygen vacancies and creates the active Ag–In2O3 interfacial site for the enhanced CO2 hydrogenation to methanol.

Biomass tar contains a variety of organic compounds (mainly aromatic and oxygen containing compounds) which have many disadvantages because of their toxicity, therefore, its removal is highly desirable [1,2]. Among the many methods for tar removing, catalytic steam reforming can convert tar into valuable products at low temperatures. In order to study the reaction mechanism, due to the complexity of real tar, most researchers are using a model compound of tar such as benzene, toluene or naphthalene. In the steam reforming of tar, dehydrogenation of hydrocarbon components and carbon formation take place on the same active sites. Subsequently, for the catalyst activity maintaining, the carbon deposited on the site is necessary to react with steam to generate CO [3].For this reason, the development of catalysts with high activity, long-time stability and reusability, with a low-cost still remains a challenge. Many catalysts were studied and each of them presents advantages and disadvantages.Ni-based catalysts are studied extensively for tar conversion because of their good activity and stability. However, the rapid deactivation by coking is a major limitation. A series of oxides such as MgO, α-Al2O3, γ- Al2O3, SiO2, ZrO2 were studied as support for Ni in the steam reforming of toluene and the activity of catalyst greatly depended on Ni particles size and on the type of interaction between Ni and support [4]. The best catalytic performance was observed on Ni/MgO due to the strong interaction between NiO and MgO. For Ni-Fe alloy supported over iron-alumina catalysts it was observed that the presence of iron plays a role of cocatalyst by increasing the oxygen species [5].Ni/olivine catalysts were studied for steam reforming using toluene as a model compound for tar and exhibited high activity, selectivity to H2 and CO and furthermore, a good resistance to deactivation due the strong metal-support interaction [6]. For example, for improving the stability of Ni/olivine catalyst a second metal such as CeO2 was added as additive, which due to its redox properties could adsorb and dissociate water and the resulting groups react with carbon deposed on Ni sites generating CO and CO2 [7]. Other promoters studied for Ni/olivine catalysts were Ca, K, Mn and it was observed that Mn exhibited the highest catalytic activity (63% conversion) but unfortunately at a high reaction temperature of 800 °C [8].The catalyst based on NiMn/Al2O3 doped with Ru to improve the redox properties was active at low temperature, 600 °C, achieving 100% conversion, but the use of noble metal makes catalyst more expensive [9]. Among others, perovskite-type oxides (Ni/LaSrAl) were investigated as support for Ni, and it was found that larger specific surface area of Ni/LaSrAl produces higher lattice oxygen release rate, high catalytic activity, and low amount of coke deposition on the surface [10]. Another studied metal was Co on the same perovskite-type oxide support emphasizing that lattice oxygen suppresses coke formation activating toluene for a redox mechanism, this catalyst showed higher redox properties than Ni and consequently higher conversion of toluene (61% compared with 50% for Ni) [11].Hydrotalcite type materials represent a novel class of supports which, due to their structure could be excellent precursors of very well dispersed metallic oxide catalysts. La-promoted Ni-hydrotalcite-derived catalysts were studied in the dry reforming of methane [12], doped with Cu were studied for steam reforming of 1-methylnaphthalene [13], and doped with noble metals for Cedar wood steam reforming [14].The identification of highly active heterogeneous catalysts for steam reforming is an open challenge. Mixed oxides MgAlMo derived from hydrotalcites are very complex materials with an important role in heterogeneous catalysis and the recent achievements recommended them as suitable for many catalytic reactions. Because there is no literature reports in terms of use Mo-HT catalysts for toluene steam reforming, the aim of this work was to investigate the influence of molybdenum content on the structure and to study the catalytic performance of these samples in the steam reforming of toluene.The MgAl LDH with molar ratio Mg/Al = 3/1 was prepared by coprecipitation at constant pH = 10 (controlled with a TitraLab TIM 854 apparatus), at a temperature of 60 °C, using Mg(NO3)2·6H2O (from Merck), Al(NO3)3·9H2O (from Tunic), Na2CO3 and NaOH (from Lach-Ner). Two solutions were prepared: a solution of NaOH (1 M) and a second solution containing Mg and Al nitrates (0.4 mol of Mg and 0.13 mol of Al) dissolved in distilled water. The two solutions were simultaneously added dropwise, at a steady rate with vigorous stirring, to an aqueous solution of Na2CO3·10H2O 0.1 M (Mg/Na = 3.33). The resulting gel was aged for 24 h, then separated by filtration, washed up with abundant distilled water until reaching a neutral pH, dried at 100 °C for 10 h and calcined at 200 °C for 2 h, 400 °C for 2 h and finally at 550 °C for 2 h.Molybdenum was introduced in the LDH structure by impregnation, so that the molar ratio Mg:Al:Mo ranged between 3:1:0.04, 3:1:0.08 and 3:1:0.12. The corresponding amount of (NH4)6Mo7O24·4H2O had been calculated so that Mo was 1.5, 3 and 4.5% (wt) respectively, and the catalysts’ abbreviation was Mo1.5MgAl, Mo3MgAl and Mo4.5MgAl. The slurry was maintained under stirring at constant temperature (80 °C) until water was evaporated. Then, the procedures for drying and calcination were followed as described above.Powder X-ray diffraction (PXRD) patterns for prepared catalysts were recorded on a Siemens D5000 diffractometer using Cu Kα radiation (λ = 1.54 Å), operating at 50 kV and 40 mA. They were recorded over the 5–70°, with a step size of 0.0403° and a scan time of 0.75sec/step. The average crystallite size was evaluated according to the Scherrer equation with formula: D(hkl) = 0.9λ/β·cosθ, in which (λ) is the wavelength of Cu Kα, (β) is the full width at half maximum intensity peak, and (θ) is Bragg’s diffraction angle. For the hydrotalcite type samples cell parameter c of the rhombohedral structure was determined from the positions of the (0 0 3) and (0 0 6) diffraction lines. In this case “c” was calculated from two diffraction lines using equation c = 3/2 (d003 + 2d006) [15]. The “a” parameter was calculated from the direction (1 1 0) with equation a = 2·d110 and represents the cation–cation distance in the brucite-like sheets [16].The N2 adsorption–desorption isotherms at 77 K were measured by the static method in an automatic volumetric Micromeritics ASAP 2020 Surface Area and Porosity Analyzer at 77 K. Prior to the measurements, samples were degassed under vacuum at 200 °C for 8 h. To calculate the surface area, the Brunauer–Emmett–Teller (BET) model was applied. Desorption branch was analyzed by applying the Barrett–Joyner–Halenda (BJH) model using the Halsey thickness curve. The total quantity of gas adsorbed at the data point closest to P/Po = 0.98 by desorption branch was used to approximate the total pore volume. The pore size distribution was calculated from the desorption branches of the isotherms.Scanning electron microscopy (SEM) studies, including imaging and electron dispersive X-ray spectra (EDX), were performed using an AMRAY 1910 field emission SEM. The analyses were performed using an accelerating voltage of 15 keV, on dry-ground samples.The surface density of molybdenum was expressed as: S u r f a c e d e n s i t y ( M o a t o m s / n m 2 ) = wt % M o l y b d e n u m l o a d i n g ∙ 6.023 ∙ 10 23 Molecular W e i g h t o f M o l y b d e n u m ∙ 100 ∙ S u r f a c e A r e a For transmission electron microscopy (TEM) studies, dry samples were triturated and scattered on to TEM grids. High resolution TEM images were collected on a CM200-FEG at an accelerating voltage of 200 kV.The Fourier transform infrared (FT-IR), Bruker IFS 66 V/S spectrometer equipped with a diamond attenuated total reflectance (ATR) accessory was used to record the spectra from 400 to 2000 cm−1 with a resolution of 4 cm−1.The UV–VIS spectra were recorded in the range 220–600 nm (using a UV3600 UV–vis spectrophotometer) with a wavelength step of 2 nm, having a slit width of 8 nm.Steam reforming of toluene was performed at atmospheric pressure using a conventional fixed bed flow reactor (i.d 9 mm) placed in a furnace in which the heating is monitored by a thermocouple. The 0.1 g of catalyst was charged in a reaction tube made of quartz. Water and toluene (S/C molar ratio 0.5–2) were introduced by syringe pumps into a vaporization furnace (300 °C) and then were carried to the reactor by a flow of nitrogen (carrier gas). The reaction temperature was varied from 400 to 500 °C, and the nitrogen flow rate was 0.3 L/h.Liquid products were collected by an ice water bath located downstream of the reactor, and the gas and liquid compositions were analyzed using gas chromatography with a thermal conductivity detector (TCD) and a flame ionization detector (FID), respectively.The gaseous product was composed of H2, CO, and CO2, and the main reactions are as follows:Steam reforming (1) C 7 H 8 + 14 H 2 O → 7 CO 2 + 18 H 2 Δ H ∘ = 647 kJ / mol (2) C 7 H 8 + 7 H 2 O → 7 CO + 11 H 2 Δ H ∘ = 870 kJ / mol Water–gas shift (3) CO + H 2 O ↔ CO 2 + H 2 Δ H ∘ = - 41 kJ / mol Dry reforming (4) C 7 H 8 + 7 CO 2 → 14 CO + 4 H 2 The XRD diffractogram of MgAl catalyst, Fig. 1 , reveals a typical pattern of a pure layered double hydroxide structure with lines at 2θ ≈ 11.2°, 22.4°, 34.4°, 38.8°; 46.1° and 60.5° corresponding to (0 0 3), (0 0 6), (0 1 2), (0 1 5), (0 1 8) and (1 1 0). Periclase (MgO) structure corresponding to (2 0 0) and (2 2 0) planes is located at 2θ ≈ 42.5° and 62.5°. Three major changes in the structure were observed at molybdenum introduction. The first modification in structure consists in decrease, until disappearance of line (0 0 3) corresponding to layered double hydroxide structure. The second change consists in the apparition of a new phase (MgAl2O4 spinel) [17] associated with the line from 65°. And finally, the interaction between molybdate anions and magnesium leads to the appearance of small tetrahedral crystallites of MgMoO4. Widening of line (0 0 3) with the entry of molybdate indicates a more disorganized structure or a decrease in crystallinity [18].The average crystal size for the direction perpendicular to the plane (0 0 3) and the lattice parameters are shown in Table 1 . The cell parameter a represents cation–cation distance inside the brucite layer, while parameter c represents interlayer distance and thickness of the brucite layer [19].The full width at half maximum (FWHM) of the basal reflection plane (0 0 3) is used for the assessment of crystallinity in the layering direction [20]. It can be noticed that Mo4.5MgAl has the highest FWHM value (1.4) and consequently, lower crystallinity than others.The basal spacing d003 for sample without molybdenum is 7.8 Å and corresponds to hydrotalcites with carbonate interlayer. This basal spacing was not affected very much when low amounts of Mo were introduced, which suggests that the molybdate anions could be adsorbed on the surface with formation of layered double hydroxide. However, when higher amounts of Mo were used, an increase in d spacing was observed. For example, the Mo4.5MgAl sample has a basal spacing of 8 Å suggesting that a part of molybdate ions have been incorporated into the structure of MgAl layered double hydroxides. This is in concordance with the higher cell parameter “c” for this catalyst. The intercalation of MoO4 2− anions into the interlamellar domain induced an increase in the interlayer distances since the molybdate anion have anionic radius higher than CO3 2−. The parameter “a” was not influenced very much.The average size of crystallites decreased from 8.3 on MgAl to 5.7 on Mo4.5MgAl.The specific surface areas and specific total pore volumes are summarized in Table 2 . The MgAl sample exhibited the highest surface area and pore volume. Both surface area and pore volume decreased drastically with incorporation of molybdenum, resulting in dense phase and blocking of pores. The highest surface of Mo4.5MgAl compared with others is attributed to intercalation of a part of molybdate anions into the interlamellar domain and release of the pores. From the pore size distribution it can be seen that, MgAl has a broader distribution with a maximum at 22.8 nm and a shoulder at 15.8 nm. Mo1.5MgAl has a bimodal pore size distribution with two maxima at 6.8 nm and 22 nm, Mo3MgAl has a maximum at 19.2 nm and a small shoulder at 10.5 nm. By contrast, Mo4.5MgAl sample exhibits narrow unimodal pore size distribution centered at 15.5 nm. The addition of molybdenum decreases the pore size, shifting the pore size distribution to lower values, the pore dimension being higher for HT without molybdenum as expected.The results obtained from EDX analysis are presented in Table 2 and it is noticeable that, with the exception of Mo1.5MgAl for all others samples the Mg to Al ratio is slightly less than the calculated value of 3:1, and for the catalysts with molybdenum this ratio decreases with increasing Mo loading.The SEM micrograph of Mo hydrotalcites presented in Fig. 2 reveals a compact angular shape with smooth faces and irregular sizes and some large interparticle cavities. Also, a very homogeneous distribution of crystal aggregates was observed.The TEM micrograph shown in Fig. 3 confirms the characteristics of LDH platelet structure with the platelets placed one above the other and some regions with fibrous morphology. The micrograph of Mo-hydrotalcites shows the tendency of platelets to cluster together in large conglomerates.The FT-IR spectra of Mo-hydrotalcites are shown in Fig. 4 . All samples present a broad absorption band in the range 3400–3500 cm−1 and another band located at 1650–1660 cm−1 that correspond to the O–H stretching vibration and bending vibration of interlayer water molecules [21], respectively. The band at 1390–1400 cm−1 is assigned to the stretching vibration of CO3 2− and the intensity of these peaks decrease with increasing molybdenum content, indicating the intercalation of less carbonate ions inside the layer and introduction of molybdate ions interlayer. The band located at 1500 cm−1 corresponds to carbonate species adsorbed on the surface. The characteristic band of antisymmetric vibration of Mo-O-Mo in MoO4 2− is located in the range 790–810 cm−1 [22] and is more pronounced for Mo3MgAl and Mo4.5MgAl samples. The band located at 590–600 cm−1 was attributed to the presence of aluminum cations in tetrahedral sites [23], and is more pronounced for catalyst without molybdenum. The intensity of this band decreases with addition of molybdenum (for Mo4.5MgAl it almost disappears) which proves that molybdenum not only gets into the layer but also replaces a part of aluminum in the structure.The presence of tetrahedral coordination of molybdenum (MoO4 species) was also proven by the UV–VIS spectra displayed in (Fig. 5 ).The catalytic performance of MoMgAl catalysts in steam reforming of toluene as a biomass tar model was carried out at 400–500 °C with a ratio of S/C 0.5–2. The major components of produced gas are H2, CO and CO2.The conversion of toluene, expressed in terms of carbon conversion, was calculated with the formula: X toluene % = n CO + n C O 2 7 ∗ n C 7 H 8 ∙ 100 The conversion of toluene and product distribution as function of temperature and molybdenum loading are shown in Fig. 6 . Toluene conversion and H2 amount increase with increasing temperature. Also the toluene conversion increases with molybdenum content, while the H2 is directly proportional to molybdenum surface density (Fig. 6d) and to the presence of both MoO4 2− anions and aluminum cations in the tetrahedral sites on the surface (Mo3MgAl). With temperature increasing from 400 °C to 500 °C, CO increases while CO2 decreases and this could be explained by the fact that, lower temperatures favor the water–gas shift reaction [24]. The effects of metal loading on toluene conversion are similar with those reported by Yue [25] over hydrotalcites with Ni, noticing the increase in conversion with metal content. Our previous works [26,27] on glycerol steam reforming over molybdenum and molybdenum-cerium catalysts also showed an increase in glycerol conversion with metal loading.In this study the following conversions of toluene were obtained: 17% on MgAl, 37.1% on Mo1.5MgAl, 53.9% on Mo3MgAl and 64.3% on Mo4.5MgAl catalyst. The hydrogen amount was 41% on MgAl, 52% on Mo1.5MgAl, 66.2% on Mo3MgAl and 60.7% on Mo4.5MgAl at 500 °C, respectively. These results are in concordance with those obtained by Josuinkas [28] over nickel catalysts derived from hydrotalcite-like compounds (28–60% toluene conversion), but the hydrogen amount was lower (10–15%). Łamacz [29] in steam reforming of toluene over ceria zirconia based Ni catalysts obtained a conversion between 38 and 65%, with a hydrogen yield around 70% at 500 °C, S/C ratio 2.4.In Fig. 6c H2/CO ratio decreases with increasing temperature while H2/CO2 ratio increases with the temperature. The water–gas shift exothermic reaction could be favored at low temperatures and consequently low selectivity of CO and high molar ratio H2/CO (3.2) are obtained. At 500 °C the H2/CO2 ratio is higher on Mo3MgAl (6.8) and Mo4.5MgAl (5.3), so, over these catalysts both reaction (2) through which CO is obtained and dry reforming (reaction (4)) are favored. The CO and CO2 concentrations are different from equilibrium values, the only catalyst that has CO2 concentration very close to 2.6 is Mo3MgAl at 450 °C (2.5).The Arrhenius plot of MoMgAl samples are shown in Fig. 7 , and for these catalysts the activation energy, Ea, is 53.2 (MgAl), 40.6 (Mo1.5MgAl), 39.1 (Mo3MgAl) and 36.8 (Mo4.5MgAl) kJ/mol.The steam to carbon ratio plays an important role in toluene reforming. In this study ratio varied from 0.5 to 2 at 500 °C and its effect on the conversion of toluene and gas product components is shown in Fig. 8 . It is noticeable that, the conversion of toluene is directly proportional to S/C ratios for all catalysts. Also, the hydrogen amount increases with the increasing steam to carbon ratio, while CO decreases. In the same time CO2 amount reached a maximum at S/C ratio 2, proving that a steam excess favors reforming and water–gas shift reaction [30]. Fig. 8d shows ratio between hydrogen and carbon monoxide, respectively between hydrogen and carbon dioxide. The H2/CO ratio increases with increasing steam and the values of S/C = 0.5 is very close to stoichiometric ones (1.6). The values obtained are 1.7 on Mo1.5MgAl, 1.8 on Mo3MgAl and 1.3 on Mo4.5MgAl. The H2/CO2 ratio is inversely proportional to S/C ratio and is very large for Mo3MgAl and Mo4.5MgAl compared with Mo1.5MgAl (2.4–2.5) over that H2/CO2 ratio is close to the stoichiometric conditions (2.6).The stability of the catalysts in time was carried out over all samples for 15 h of reaction at 500 °C and S/C ratio of 2. The results are shown in Fig. 9 . The results reveal a good stability for all catalysts. Furthermore, a good thermal stability was proven by the XRD pattern after the reaction (Fig. 1). The catalyst stability could be associated with a good interaction between Mo and Mg that favors the formation of Mo particles with small sizes.Layered double hydroxides (LDH) with intercalated molybdate anions have been prepared by co-precipitation and impregnation routes. The XRD results showed that, the molybdate anions react with magnesium leading to small tetrahedral crystallites of MgMoO4, tetrahedral coordination of molybdenum (MoO4 species) was also evidenced from UV–VIS spectra. The FT-IR spectrum reveals that molybdenum not only gets into the layer but also replaces a part of aluminum in the structure. The toluene steam reforming reaction was carried out on these catalysts and it was observed that the conversion of toluene and the H2 amount are directly proportional to the temperature. Furthermore, the toluene conversion increases with molybdenum content, while the hydrogen depends on two factors: the presence of molybdate species on the surface and the presence of Al in tetrahedral coordination (Mo3MgAl sample).

This study evaluated the catalytic activity of Mo catalysts derived from hydrotalcite-like compounds for steam reforming of toluene as a model compound for tar. The catalysts with 1.5, 3 and 4.5 Mo loadings (wt%), denoted as Mo1.5MgAl, Mo3MgAl and Mo4.5MgAl respectively, were prepared by coprecipitation and characterized by BET, XRD, SEM, TEM, FT-IR and UV–VIS. The results showed that toluene conversion increased with increasing molybdenum content. The hydrogen amount depended on two factors: the presence of molybdate species on the surface and the presence of aluminum cations in tetrahedral sites (Mo3MgAl), with molybdenum influence being more pronounced. The H2/CO ratio decreased at increasing temperature while, the H2/CO2 ratio increased proportionally with temperature. Mo1.5MgAl catalyst was more selective for CO2 and H2, while, Mo3MgAl and Mo4.5MgAl were more selective for CO and H2.

Recently, society has been concerned by the environmental pollution of soils, atmosphere and water. Specifically, about the water pollution, continuous technological advance and consumption growth in today’s society increase worryingly contamination of the aqueous environment [1], and currently, the effect of the contaminants of emerging concern (CECs) has taken a great interest [2–5]. These pollutants are chemicals with high complexity (chemical and physical) that resist photolytic, biological and chemical degradation. Some examples of these contaminants are the pharmaceutical and personal care products, pesticides, etc. They present a high toxicity in the aquatic medium, even at low concentration, and a high diffusion by air and water, which leads to a more harmful effect on the environment [3,5,6]. Thus, the presence of these compounds in water can affect flora and fauna and, consequently, human health. Specifically, they can cause different types of cancer (breast, ovary, prostate, testes, etc.), disorders in endocrine and neurological systems, reproductive capabilities and hormonal control [7–10].This problem increases because of conventional treatments of water and wastewater treatment plants failed to completely remove these pollutants. Specifically, ofloxacin (OFX), a widely used antibiotic today, cannot be completely removed by these treatments and it can be found in rivers and lakes [11–13]. Its presence in the aquatic medium could pose low to medium risks to aquatic organisms, and the occurrence and distribution of antibiotic resistance genes can take place [14–19].Thus, it is necessary to look for alternative treatments that improve the CECs removal. In this context, it is possible to find non destructive and destructive methods. In the first group, adsorption and solvent extraction stand out in research and industry areas, however, the efficiency of these technologies to remove CECs is low [5,19]. Within the second group, incineration, oxidation process, wet oxidation and supercritical oxidation present a good efficiency in the degradation of different organic compounds from water and wastewater, but not for the removal of CECs [20,21]. However, advanced oxidation processes, based on the generation of hydroxyl radicals and other oxidizing species are considered by many authors as a good option for the treatment of water and wastewater for the degradation of CECs, with a high efficiency and environmental compatibility [1,22–26]. Among them, it is worth highlighting photocatalytic processes and, specifically, the heterogeneous photocatalysis with supported TiO2 nanotubes, which allows, on the one hand, to remove a final stage to take the photocatalyst from the medium after the treatment, significantly lowering the cost of operation. On the other hand, knowing that photocatalysis is a process that takes place mainly on the catalyst surface [27,28], the use of TiO2 nanotubes in the anatase phase (photocatalytically active phase) entails a high activity, because of the higher active surface on the nanotubes able to generate hydroxyl radicals and other oxidants of organic compounds.The next reactions can take place during the water treatment by photocatalysis, where ultraviolet photons reach the catalyst surface and form excited radicals, which can improve the treatment efficiency [24,26]. (1) H2O → •OH + H+ + e- (2) OH- → •OH + e- (3) 2 •OH → H2O2 (4) OH + O2 → O3 + H+ (5) MOx(•OH) → MOx(•OH)* In addition, it is possible to find an uniform distribution and optimal size for the nanotubes with a good transmission in the UV region, stability and resistance to degradation and durability [9,28,29]. In this context, anodization process, carried out for the formation of the TiO2 nanotubes, can be studied in order to look for the optimal conditions for the most active photocatalyst.With this background, the goal of this work is to study the anodization process for the formation of TiO2 nanotubes in the photocatalyst, paying special attention to the electrolyte used (H2SO4 in water or NH4F/H2O in ethylene glycol, chosen according to literature [9,27–30]), the maximum potential applied (20–60 V), the potential ramp (2–4 V min−1) and the subsequent heat treatment (maximum temperature of 450 ºC), in order to transform the amorphous phase into crystalline-phase TiO2 nanotubes, to increase their catalytic capacity after crystallizing in their anatase form [31,32]. The optimization of the process was conducted through the application of factorial design of experiments and surface response analysis [33]. The efficiency of the process will be evaluated by the treatment of synthetic wastewater polluted with OFX, as an antibiotic model of CEC. Moreover, wastewater treatment experimental conditions will be also statistically analyzed (UV wavelength, irradiance and initial concentration of OFX), in order to find the optimal operating conditions.Titanium plates (5 ×5 cm; 1 mm thickness; 99.2% purity) were supplied by Alfa Aesar (Ward Hill, Massachusetts, United States of America). Ofloxacin (OFX), nickel plate (used as cathode for the anodization process), ammonium fluoride, ethylene glycol and acetone (for plate cleaning) were supplied by Sigma-Aldrich (Steinheim, Germany) with > 99.0% purity. Sulfuric acid (97%) and acetonitrile, with analytical grade, were supplied by Panreac Química S.A. (Barcelona, Spain).The anodization process and the subsequent heat treatment were conducted according to Martín de Vidales et al. [34], being the electrolyte used not only 250 mL of NH4F 0.1 M in a solvent formed by 20% v/v of water in ethylene glycol, but also H2SO4 0.5% w/w in water, depending on the studied conditions. Therefore, a Ti plate is immersed in an electrolyte in front of a Ni plate, and a potential ramp is applied up to a maximum potential that is maintained for 90 min. Then, a heat treatment (Energon oven) can be applied up to a maximum temperature of 450 ºC (a ramp of 2 ºC min−1 up to 350 ºC, which is maintained for 30 min, another ramp of 2 ºC min−1 up to 450 ºC, maintained for 150 min, and final drop to room temperature) [27]. Before anodization, the titanium plate requires a previous cleaning treatment with clean water and drying it with paper; and after that, acetone and water ultrasonic baths for 15 min.The formation of TiO2 nanotubes in Ti plates was analyzed by scanning electron microscopy (SEM) with a JEOL JSM-820 analyzer, 1000 – 30,000 V. Image resolution at 25 KV: 3.5 nm (at 8 mm working distance), 10.0 nm (at 39 mm working distance).The maximum absorption wavelength of OFX solution is 332 nm, obtained from the scanning plot measured in a UV–VIS spectrophotometer (UVIKON 941 plus). OFX concentration was measured using a HPLC Jasco MD-2010/2015 with a 5 µm C18 analytical column (4.6 mm × 250 mm), using a mobile phase of acetonitrile/water (50/50 v/v %) at the flow rate of 1 mL min−1. The column temperature was 25 ℃ and samples of 5 µL were injected. Before analysis, the samples were filtered by 0.45 µm cellulose filter.Bench-scale photocatalysis treatments were conducted under batch-operation mode ( Fig. 1). Wastewater was stored in a jacketed glass tank (volume = 500 mL) with agitation acting as a reactor, where the photocatalyst (Ti plate with TiO2) was put into it, near the liquid surface, on a metal mesh. An ultraviolet lamp was placed over the Ti plate and the wastewater (the Ti plate is submerged in the wastewater 1 cm from the surface, leaving only the anodized side irradiated), in order to activate the electron leap from the valence to the conduction band, and improve the generation and/or activation of oxidizing agents, which are mainly found on the catalytic surface [9]. UV wavelength was 365 nm (UV-A), 311 nm (UV-B) or 254 nm (UV-C), and the irradiance on the photocatalyst surface was measured with a HD 2102.1 radiometer, with measurement probes LP 471 UVA, LP 471 UVB and LP 471 UVC, supplied by PCE Instruments S.L. (Albacete, Spain). The ultraviolet lamps were supplied by BCB S.L. (Barcelona, Spain).The initial concentration of OFX was of 15–35 mg dm−3 (synthetic wastewater), a typical concentration found in Wastewater Treatment Plants [2]. The jacket was coupled with a controlled thermostatic bath (Digiterm 100, JP Selecta, Barcelona, Spain) pumping water to maintain the temperature at the desired set point (25 ºC).The statistical analysis was performed using Statgraphics v.17.2.00 software (Statgraphics Technologies Inc.). In order to optimize the anodization process, the influence of the electrolyte used (H2SO4 in water or NH4F/H2O in ethylene glycol), the maximum potential applied (20, 40 or 60 V), the potential ramp (2, 3 or 4 V min−1) and the subsequent heat treatment (to apply it or not) were evaluated using the factorial design and response surface methodology.In a first stage of the experimental design, the effect of the electrolyte used and the possibility of applying a heat treatment after the anodization were evaluated using a two-level factorial design. As defined in the methodology of the factorial design of experiments [33], the main effect of a factor is defined as the mean change of the response variable obtained varying a factor among the higher and the lower level. Accordingly, the effect of the electrolyte factor (EEL) on the kinetic constant (k) as response variable is defined in Eq. 6: (6) E EL = ∑ k EL + − ∑ k EL − N / 2 = ∑ k EL + ̅ − ∑ K EL − ̅ The interaction effects are calculated with a similar equation taking into account the value of the response variable in the experiments with both factors at higher and lower level. Thus, the interaction between the electrolyte factor and heat treatment factor EEL-HT is defined in Eq. 7: (7) E EL − HT = ∑ k ( EL and HT ) + − ∑ k ( EL and HT ) − N / 2 = ∑ k ( EL and HT ) + ̅ − ∑ K ( EL and HT ) − ̅ According to the methodology of the experimental design [33], if the value of one effect is out of the confidence interval then this factor has a significant effect of the response variable. The same can be said about the value of the interaction effect.The confidence interval, at 95% of significance level, was calculated using the Eq. 8: (8) CI 95 % = ± t 0 , 05 , 8 · σ t · 1 / n Where “t” es the value of the Student´s test statistics calculated at 95% of significance level and 8 degrees of freedom (2 degrees of freedom for each 3-replicated experiment and 4 experiments); n is the total number of samples (12 samples accounting for 3 replicates in each of the 4 experiments), and σt is the standard deviation of the 12 values of kinetic constant, calculated by the Eq. 9 with the standard deviation of the kinetic constant of each experiment σ: (9) σ t = ∑ σ i 2 Table 1 shows the experimental matrix for a two-level factorial design carried out in order to study these two factors. The experiments were run at random to minimize errors due to possible systematic trends in the variables. Columns 2 and 3 present the two factors on a natural scale while maintaining constant the potential ramp and maximum potential in the central values of 3 V min−1 and 40 V, respectively. Columns 4 and 5 represent the dimensionless coded levels of the factors (+1 and −1 for the higher and lower level, respectively). As can be observed, plate anodization was conducted in triplicate, in order to evaluate the experimental error. The kinetic constant for the degradation of ofloxacin in water ([OFX]0 = 25 mg dm−3. UV-A: λ = 365 nm) was chosen as response variable to assess the efficiency of the different operating conditions of the anodizing process. Columns 6, 7 and 8 present the kinetic constant (calculated as described below) for the three replicates of each experiment, whereas column 9 presents the mean value for the kinetic constant of each experiment.The results in the photocatalytic tests are shown in Fig. 2. Experiments were carried out in triplicate with each plate, under the same operating conditions, in order to minimize the experimental error. The dots show the average values, and the error bars the standard deviation of the three replicates under the conditions 1, 2, 3 and 4. The y-axis show the dimensionless normalized concentrations of OFX for a better comparison.It is important to highlight that photolysis process was also carried out, and a nil degradation of OFX was observed. Thus, as shown in Fig. 2, photocatalytic process improves the pollutant degradation in all cases. When a heat treatment after the anodization process is not applied, the process efficiency is lower. This can be explained taking into account that the heat treatment improves the disposition and arrangement of the nanotubes [9,27–29,31], transforming the amorphous phase into crystalline-phase TiO2 nanotubes, and increasing the catalytic capacity after crystallizing in their anatase form [31,32]. On the other hand, to use an electrolyte of H2SO4 in water only entails the formation of a film of TiO2 on the Ti plate [30], while NH4F/H2O in ethylene glycol allows the formation of TiO2 nanotubes in its anatase form, as Martín de Vidales et al. observed by X-ray diffraction [34], since the formation and growth of TiO2 nanotubes during anodizing is due to the competition that exists between the formation of TiO2 itself on the surface of the plate and the formation/dissociation of complexes formed by Ti-F, according to the following reactions [27–29]: (10) H2O → 2 H+ + O2- (11) Ti + 2 O2- → TiO2 + 4 e- (12) TiO2 + 6 F- + 4 H+ → TiF6 2- + 2 H2O These reactions, starting with the electrolytic dissociation of water due to the passage of current, continuing with the formation of the TiO2 layer and, finally, with the formation of the TiF6 complex allows the formation and growth of nanotubes. Specifically, the fluorinated complex causes the chemical dissolution of the oxide layer, creating small pores that allow the passage of current and the formation of nanotubes in a direction perpendicular to the surface of the titanium sheet, which acts as an electrode during the anodizing process. Thus, the formation of TiO2-anatase nanotubes improves the active surface for a higher process efficiency.In order to check this result, SEM was carried out to analyze the surface of these plates. Fig. 3 shows some examples of plates anodized with both electrolytes and with or without heat treatment.As observed, when H2SO4 is used as electrolyte, TiO2 nanotubes are not formed and only a film of TiO2 is found, with a much smaller photocatalytically active surface. In this context, it is important to highlight that the anodization process conducted with H2SO4 in water as electrolyte, which does not entail the formation of TiO2 nanotubes and only a layer of TiO2 is observed on the catalyst surface, was evaluated because plates with different colors can be obtained and they can be used in exterior walls of buildings to treat polluted atmospheres, or in the hull of ships to remove pollutants from the ocean [30].In addition, when a heat treatment is applied after the anodization process using NH4F/H2O electrolyte, a better disposition of the nanotubes is observed, and it seems that a thin film of TiO2 that could hinder the access of the pollutant to the surface of the nanotubes, disappears. Therefore, results obtained by SEM analysis corroborates what was said before. On the other hand, it is observed that the heat treatment allows the formation of a more uniform TiO2 film after the anodization with H2SO4 as electrolyte, which can improve catalyst-contaminant contact.Kinetic constants were calculated for all experiments conducted, taking into account a pseudo-first order kinetic [35–37], and results are shown in Table 1 (kinetic constant of each plate was calculated with the medium value of the three experiments conducted under the same operating conditions).A statistical study of the results has been conducted to corroborate these conclusions. Table 2 presents the main effects and interactions of the two studied factors on the kinetic constant as response variable. The confidence interval at 95% of significance level was ± 1.16·10−4 min−1.As it can be seen from the table, the main effects of both factors as well as the interaction effect between them are higher than the confidence interval at 95% of significance level being all of them positive. It indicates that the use of NH4F/H2O in ethylene glycol as electrolyte, the heat treatment after the anodizing process and the interaction between both factors present a statistically significant influence on the kinetic constant. Specially, the influence of the heat treatment is very high indicating that its application after the anodizing process notably increase the decontamination rate of OFX. The use of NH4F/H2O in ethylene glycol as electrolyte and the interaction with the heat treatment is also positive for the photocatalytic activity but in a lesser extension.Takin into account these results, the optimal conditions found for the first stage of the experimental design are the use of NH4F/H2O in ethylene glycol as electrolyte, and a heat treatment after the anodization process.In a second stage the influence of the potential ramp and the maximum potential reached in the anodizing process was studied over the treatment of a water polluted with 25 mg dm−3 of ofloxacin, where kinetic constant is chosen as response variable using a face-centered central composite design ( Table 3). Taking into account the results of the first stage of the experimental design, electrolyte and heat treatment factors were maintained in their high values. Table 3 presents the experimental design matrix for this second set of experiments, where columns 2 and 3 present the values of the factors (maximum potential and potential ramp) on a natural scale, maintaining NH4F/H20 in ethylene glycol as electrolyte and heat treatment applied in all experiments. Columns 4 and 5 represent the dimensionless coded value for the maximum potential and potential ramp factors, labeled as + 1 in the upper level of the factor and − 1 in the lower level, being the label 0 for the central value of the factor. Kinetic constant as response variable ([OFX]0 = 25 mg dm−3. UV-A: λ = 365 nm) was calculated for all experiments conducted, also taking into account a pseudo-first order kinetic [34–36], and results are shown in column 7.The experiments 5–8 constitute the four points of the experimental design. The experiment 13 constitutes the central point of the design and allows calculate the error in order to carry out the statistical analysis of the results. The experiments 9–12 constitute the star points (face-centered points), corresponding to anodizing conditions resulting of the combination of points of the factorial design.The ANOVA analysis for these results is presented in Table 4, where the influence of factors and its squares and the interactions between factors is reflected. According to ANOVA analysis, those parameters with F-value higher than 1 are significant. Moreover, the p-value is related to the significance level and a p-value less than 0.05 indicates that this parameter has statistical significance with a 95% significance level.According to these results, the maximum potential is not significant whereas the potential ramp and the square of both factors are significant. Thus, the response variable has a linear relationship with the potential ramp and a quadratic relationship with both studied factors.Among the significant factors, the square of the potential ramp is significant at 93% of significance level being this p-value the minor of those corresponding to significant factors. Therefore, this percentage is selected as a limit, and the effects with p-value less than 0.07 will be considered as significant. According to this, the potential ramp and the square of maximum potential and potential ramp will be taking into account to obtain the mathematical model that relates these parameters with the response variable. Eq. 13 present the statistical fitting model that express the relationship among the response variable (kinetic constant) and the codified values of the maximal potential and potential ramp factors, within the experimental range considered. (13) k min − 1 = 1.11 · 10 − 3 − 1.2 · 10 − 3 · X R − 1.75 · 10 − 4 · X V max 2 + 1.65 · 10 − 4 X R 2 The mathematical model allows to predict the response in the experimental region and to obtain the optimum value for the kinetic constant, that would correspond to the central value for the maximum potential (40 V corresponding to a coded value of XVmax = 0) and the lower value for the potential ramp (2 V min−1 corresponding to a coded value of XR = −1).SEM images of plates corresponding to the second part of the experimental design, i.e. experiments 6–13 (Figs. S1 and S2), show that an array of TiO2 nanotubes has been obtained in all cases. However, the authors have not found a clear relationship between the structural/morphological properties of the nanotubes film and the experimental conditions in the anodizing process, i.e. the maximum potential and potential ramp.According to the literature [28,38], the growth and dimensions of the nanotubes are affected by several parameters of the anodizing process such as maximum potential, profile potential-time or number of steps with different maximum potential. The results obtained in the statistical analysis suggest that a low potential ramp allows the growth of an uniform array of nanotubes that provides higher surface and, consequently, a better photocatalytic performace. A high potential ramp could favour a less uniform nanotubes array, as well as the breaking of the nanotubes or/and prevent their growing. In addition, acording to the results of the factorial design, the influence of the maximum potential applied during the anodizing process, at least in the experimental interval tested in this work, would not be a critical parameter for controlling the structural morphology of the nanotubes array and, hence, the photocatalytic activity. Fig. 4 shows the surface response of kinetic constant based on coded factors of maximal potential and potential ramp.The obtained surface has the shape of a saddle, and the maximum is obtained for the central point of the maximum potential and for the minimum value of the potential ramp. The surface is projected on the lower plane of the figure and in the side bar, coloring the optimal zone of the response variable in red and the zones where the minimum is reached in blue.Finally, a two factors three-level factorial design has been applied for the third stage of the experiments design, where wastewater treatment conditions were studied, in order to find the optimal operating conditions. The studied factors were irradiance and initial concentration of OFX at different UV wavelength. To do this, different experiments were carried out applying UV-A, UV-B or UV-C in the photocatalytic process, with an irradiance of 5, 10 or 15 W m−2, to treat a synthetic wastewater polluted with 15, 25 or 35 mg dm−3 of OFX. All experiments were carried out using a new Ti plate anodized with the optimal conditions obtained in the previous section (NH4F/H2O as electrolyte, 2 V min−1, 40 V and heat treatment). Table 5 shows the experimental matrix design with the natural value of the factors in columns 2 and 3, and the coded value in columns 4 and 5. The first-order kinetic constant of each experiment for the different UV lamps are in columns 6, 7 and 8.The ANOVA analysis of these results is shown in Table 6 for UVA-B, and Tables S1 and S2 (see supplementary information) for the UV-A and UV-C radiation, respectively. Eqs. 14, 15 and 16 present the statistical fitting model that express the relationship among the response variable (kinetic constant) and the codified values of irradiance and initial concentration of OFX, within the experimental range considered, for UV-B, UV-A and UV-C, respectively. (14) k min − 1 = 7.88653 · 10 − 4 + 4.25113 · 10 − 4 · X I − 3.29613 · 10 − 4 · X OFX 0 (15) k min − 1 = 6.98895 · 10 − 4 + 3.39167 · 10 − 4 · X I − 1.93667 · 10 − 4 · X OFX 0 (16) k min − 1 = 6.54895 · 10 − 4 + 3.03333 · 10 − 4 · X I − 3.48167 · 10 − 4 · X OFX 0 Table 6 shows that the main effects of both factors, irradiance and the initial concentration of OFX, are significant at 95%. The Eq. 14 shows that kinetic constant increases with the applied irradiance (positive coefficient in the fitting equation), because more radiation is supplied to the photocatalyst and more oxidants can be formed [32]. On the other hand, the constant decreases when the initial concentration of the pollutant increases in the reaction medium (negative coefficient in the fitting equation). This can be explained taking into account that a higher concentration of the pollutant can negatively affect to the radiation that can reach the catalytic surface [39]. In this context, it is important to highlight that photocatalysis is a surface process where the contact between pollutant, radiation and photocatalyst is necessary [40].For UV-A and UV-C experiments the same conclusions are obtained (Tables S1 and S2, Eqs. 15 and 16), and therefore the irradiance and initial concentration of OFX are significant factors at 95%, with a positive effect of the irradiance and negative effect of the initial concentration of OFX on the kinetic constant.The coefficients of the mathematical fit for both factors, irradiance and initial concentration of OFX (Eqs. 10, 11 and 12), are higher for the experiments with UV-B lamp. This means that the effect of both factors is higher with this wavelength and therefore, UV-B seems be the most efficient radiation, which agrees with what was observed by McMurray et al. [41], because of this is the optimal radiation when TiO2-anatase is used as photocatalyst.It is important to highlight that this is an initial study and next works are necessary in order to find more efficient processes with a higher degradation of the contaminants. Nevertheless, this study opens the door to a promising technology for the treatment of water and wastewater polluted with CECs. In addition, it is possible to use sunlight as radiation source, obtaining a low cost technology, with easy implantation, high efficiency and environmental compatibility.From this work, the following conclusions can be drawn: - Heterogeneous photocatalysis with TiO2-anatase has been applied for the removal of contaminants of emerging concern from wastewater, specifically, ofloxacin degradation has been tested. - Synthesis of the photocatalysts has been carried out by anodization of Ti plates. Anodization conditions and the possibility of to apply a heat treatment after the anodization process have been studied by experimental design, evaluating the electrolyte used (NH4F/H2O in ethylene glycol or H2SO4 in water), the potential ramp applied (2–4 V min−1), the maximal potential (20–60 V) and the heat treatment after anodizing (maximal temperature of 450 ºC). The influence of these variables has been evaluated using the kinetic constant of the OFX degradation experiments as response variable, and results show that NH4F/H2O as electrolyte, 2 V min−1, 40 V and a heat treatment are the optimal parameters for the formation of TiO2-anatase nanotubes, with a good size and distribution. However, a statistical study shows that the maximum potential is not a determinant factor. - Characterization of the prepared photocatalysts is carried out by SEM, and results show that, when NH4F/H2O is used as electrolyte and a heat treatment is applied, TiO2 nanotubes are formed with better disposition and arrangement, in the anatase form (active phase for the photocatalysis process). - The operating conditions for the wastewater treatment (UV radiation, irradiance and initial concentration of the pollutant) also have been evaluated. Results of the experimental design show that UV-B, the maximal irradiance and the minimal initial concentration of the pollutant are the optimal conditions, because of a higher formation of oxidants of organic matter. Heterogeneous photocatalysis with TiO2-anatase has been applied for the removal of contaminants of emerging concern from wastewater, specifically, ofloxacin degradation has been tested.Synthesis of the photocatalysts has been carried out by anodization of Ti plates. Anodization conditions and the possibility of to apply a heat treatment after the anodization process have been studied by experimental design, evaluating the electrolyte used (NH4F/H2O in ethylene glycol or H2SO4 in water), the potential ramp applied (2–4 V min−1), the maximal potential (20–60 V) and the heat treatment after anodizing (maximal temperature of 450 ºC). The influence of these variables has been evaluated using the kinetic constant of the OFX degradation experiments as response variable, and results show that NH4F/H2O as electrolyte, 2 V min−1, 40 V and a heat treatment are the optimal parameters for the formation of TiO2-anatase nanotubes, with a good size and distribution. However, a statistical study shows that the maximum potential is not a determinant factor.Characterization of the prepared photocatalysts is carried out by SEM, and results show that, when NH4F/H2O is used as electrolyte and a heat treatment is applied, TiO2 nanotubes are formed with better disposition and arrangement, in the anatase form (active phase for the photocatalysis process).The operating conditions for the wastewater treatment (UV radiation, irradiance and initial concentration of the pollutant) also have been evaluated. Results of the experimental design show that UV-B, the maximal irradiance and the minimal initial concentration of the pollutant are the optimal conditions, because of a higher formation of oxidants of organic matter.The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Maria Jose Martin de Vidales reports financial support was provided by Polytechnic University of Madrid.The authors acknowledge the acceptation of this paper to be published in Catalysis Today - Special Issue Selected Contributions of the 11th European Meeting on Solar Chemistry and Photocatalysis: Environmental Applications (SPEA11).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.cattod.2023.01.002. Supplementary material .

In this work, the treatment of wastewater polluted with contaminants of emerging concern is evaluated by heterogeneous photocatalysis with TiO2. Titanium plates are anodized by different experimental conditions, and the influence of these conditions on the catalytic performance is evaluated by factorial design of experiments. The formation of TiO2 nanotubes in the anatase form is sought for photocatalysts with high active surface. Thus, electrolyte used in the anodization process (NH4F/H2O in ethylene glycol or H2SO4 in water), potential ramp (2 – 4 V min−1), maximal applied potential (20 – 60 V) and heat treatment are evaluated as influencing factors of the experimental design. The activity of these catalysts is tested by ofloxacin (pharmaceutical compound used as model of contaminant of emerging concern) degradation in wastewater treatment, obtaining the kinetic constant of the process, parameter chosen as response variable of the experimental design. Results show that NH4F/H2O as electrolyte, 2 V min−1, 40 V and a heat treatment are the optimum conditions for the formation of TiO2–anatase nanotubes with a good disposition and arrangement, observing the higher efficiency in the wastewater treatment process. In addition, wastewater treatment conditions are evaluated (UV wavelength, irradiance and initial concentration of ofloxacin), and it is found that UV-B, maximal irradiance and minimal initial concentration of the pollutant are the optimal factors for a higher process efficiency. This is a work that opens the door to the removal of contaminants of emerging concern from wastewater with a technology of low cost, with easy implantation and environmental compatibility, making possible to use only sunlight as a reagent.

Ammonia is regarded as a safe and sustainable energy carrier due to its high hydrogen content and narrow flammable range [1,2] enabling the long term (days to months) energy storage in chemical bonds versus the short-term storage (seconds to hours) offered by electrochemical storage (i.e. batteries). In this way, the use of ammonia as an energy vector could facilitate the balance of seasonal energy demands and intermittent renewable energy production (e.g. solar, tidal and wind) in a carbon-free society [3–5]. Established safety protocols and existing transportation and distribution networks applicable for ammonia [6] make it also an effective possible solution in comparison to hydrogen due to the current lack of viable methods to store hydrogen in a compact, safe and cost-effective manner [7]. Despite its potential, the implementation of ammonia in the energy landscape relies on the capability of releasing hydrogen on-demand, preferably at temperatures aligned to those of fuel cells [8]. A considerable scientific effort is currently focused on the design of catalysts for the low temperature activation of ammonia for the production of hydrogen [2,9,10]. The most active catalysts reported in the literature are ruthenium-based [11–16]. The optimum properties of ruthenium actives sites are associated with optimum N-adsorption energy [17] which enables activation of the ammonia molecule while avoiding poisoning by N-adatoms at low temperature (known to be the limiting step at such conditions). Enhancement of the ruthenium activity can be achieved by the use of electron donating promoters [11,13,18] and highly conductive supports, such as graphitised carbon nanotubes [19]. However, there is much interest in the identification of alternative catalysts which rival, if not exceed, the performance of ruthenium. Our recent review on the subject identifies cobalt as an attractive alternative, however studies show it to possess poor activity compared to ruthenium-based systems, especially at low temperatures [2,20–23]. The work reported within this manuscript demonstrates a systematic approach to replicate or improve upon the ammonia decomposition activity of ruthenium-based catalysts. The strategy employed by us has been the development of bi-metallic systems combining metals possessing different N-adatom adsorption energies following the DFT simulations by Hangsen et al.[17] to achieve an optimum binding energy for catalytic performance. Within our studies, cobalt-rhenium systems present activity at conditions comparable to Ru/CNT catalysts [11] with very high stability under consecutive runs and no observed formation of nitrides (Co-N and Re-N) under the ammonia atmosphere. Even though we recognise the scarcity and cost of Re, the knowledge provided in this study is useful for the development of catalysts of enhanced activity. The low temperature activity is directly related to the intimate Co-Re interaction with the activity onset related to the contraction of the Re-Co bond distance.Cobalt rhenium materials were prepared to yield different Co/Re ratios, by mixing varying amounts of ammonium perrhenate (NH4ReO4, Sigma Aldrich, >99%) in deionized water with cobalt nitrate (Co(NO3)2.6H2O, Sigma Aldrich, >98%). The solutions were stirred for 1 hour then dried in an oven at 125 °C for 12 hours. After drying, the materials were ground by hand and calcined in air at 700 °C (using a 10 °C min-1 ramp rate) for 3 hours. Ruthenium supported on carbon nanotube (Ru/CNT) catalysts were prepared by incipient wetness impregnation using Ru(NO)(NO3)3 (Alfa Aesar). Multi-walled carbon nanotubes (Sigma Aldrich, OD 6-9 nm, length 5 μm, SBET 253 m2 g-1) were used as support. After impregnation of the aqueous solutions, the catalysts were dried at 100 °C under vacuum for 3 hours and then reduced under hydrogen at 230 °C for 1.5 hours.Following degassing the materials, nitrogen physisorption isotherms were measured at −196 °C using a Micromeritics ASAP 2020 instrument. The surface area was calculated using the Brunauer, Emmett and Teller (BET) method. Temperature programmed reduction (TPR) experiments were carried out in a Micromeritics Autochem 2920 instrument equipped with a thermal conductivity detector (TCD). The samples characterised from room temperature to 900 °C using a temperature ramp rate of 10 °C min-1 under 50 mL min-1 flow of 5 % H2/Ar. CO pulse chemisorption analyses at 35 °C were carried out using the Micromeritics Autochem 2920 instrument equipped with a TCD. Samples were pre-treated at 250 °C under a helium flow for 1 hour to ensure desorption of water.Cobalt K-edge and rhenium LIII-edge XAS data were collected at the Swiss-Norwegian Beamline (SNBL, BM1B) at the European Synchrotron Radiation Facility (ESRF) in transmission mode. The data was collected in the 16-bunch filling mode, providing a maximum current of 90 mA. A bending magnet collected the white beam from the storage ring to the beamline. The SNBL is equipped with a Si(111) double crystal monochromator for EXAFS data collection. The incident and transmitted intensities (I0 and It + I2) were detected with ion chambers filled with, I0 (17 cm) 50 % N2 + 50 % He, and It and I2 (30 cm) with 85 % N2 + 15 % Ar at the cobalt edge. Cobalt references (Co-foil, CoO, Co3O4) and rhenium references (Re-foil, NH4ReO4) were also collected. The cobalt K-edge XAS data were measured in continuous step scan mode from 7600 eV to 8300 eV with a step size of 0.5 eV and counting time of 300 ms per step. The rhenium L-III data were collected in transmission mode, using ion chambers fillings of 100 % N2 (I0, 17 cm), 50 % N2 + 50 % Ar (It, 30 cm). Step scans were collected between 10350 eV to 11800 eV, with a step size of 0.5 eV and counting time of 200 ms per step.For all in situ measurements, great care was taken to ensure similar conditions were applied for both edges, hence sample weight, cell thickness and gas flow were kept constant. The CoRe catalysts were mixed with boron nitride, pressed to wafers and sieved fractions (above 375 μm) were then placed inside 0.9 mm quartz capillaries with quartz wool on either side. The capillary was heated by a blower placed directly under the sample, and the exhaust was continuously sampled using a Pfeiffer Omnistar Mass Spectrometer. The protocol for the ammonia decomposition includes pre-treatment in 75% H2 in argon at 600 °C for one hour using a 5 °C min-1 ramp rate using a total flow of 10 mL min-1. EXAFS step scans were collected continuously, with XRD patterns being collected at end points. After the pre-treatment, samples were cooled to 200 °C before switching to 5% NH3 in helium and heating to 700 °C using a ramp rate of 2 °C min-1. EXAFS step scans were collected continuously, and the exhaust was continuously analysed using the mass spectrometer.The XAS data were binned (edge region −30 eV to 50 eV; pre-edge grid 10 eV; XANES grid 0.5 eV; EXAFS grid 0.05 Å-1) and background subtracted, and the EXAFS part of the spectrum extracted to yield the χi exp(k) using Athena software from the IFFEFITT package. [24] The XANES spectra were normalised from 30 to 150 eV above the edge, while the EXAFS spectra were normalised from 150 eV to the end point. The data were carefully deglitched and truncated when needed. For cobalt the threshold energy (E0) was set to be at the mid-point (0.5) of the normalised absorption edge step ensuring it was chosen after any pre-edge or shoulder features. For rhenium samples E0 was determined to be the first inflection point in the first derivative spectra, as there are no pre-edges or shoulder features. All XANES spectra were energy corrected against the corresponding reference foil (Co = 7709 eV, Re = 10535 eV).Due to the bimetallic nature of the CoRe-catalyst, as reported previously [25], obtaining accurate comparable results from linear combination of XANES using known references was difficult. For this reason, reduction and reaction profiles were obtained using multivariate curve resolution (MCR). MCR using the alternating least-square (ALS) mathematical algorithm is a chemometric method which is well-known for its ability to provide the pure response profile of the chemical constituent (species) of an unresolved mixture. Nowadays, MCR is heavily used as a blind source separation method (no reference spectra) to process large data-sets generated in labs and synchrotron facilities all over the world. For a detailed description of the method employed, software and usage, the reader is directed to literature from Jaumot et al. [26,27] and Ruckebush et al. [28] MCR-ALS was used to analyse the operando time-resolved (TR) XANES data-sets for Co-Re bimetallic catalysts during the pre-reduction step. For the assessment of the minimum number of principal components that describe the system, i.e. rank analysis, a built-in method based on the singular value decomposition (SVD) was used [29]. The SVD results display the calculated eigenvalues of the data versus the component number (a so-called scree plot), which allows understanding how much variance each component can explain. A break in the slope of such a plot is generally associated to the minimum number of components able to simulate the initial mixture. The MCR-ALS graphical user interface (GUI) for Matlab® used in the present manuscript (http://www.mcrals.info/) was applied on the XANES data-sets for both Co and Re edges (Co: 7600-8000 eV and Re: 10450-10800 eV). Positive constraints were utilised for both concentration and spectra profiles and closure constraints for the concentration (i.e. no mass transfer; constant concentration of the absorber throughout the experiment).The peak fitting feature in the Athena software was used to calculate the area for the white line intensity at the Co K-edge for CoRe1.6 during NH3-treatment. A Gaussian curve was used to calculate the area. Difference spectra were made with the Athena software.EXAFS least-squares refinements were carried out using DL-EXCURV [30], which conducts the curve fitting of the theoretical χcalc(k) to the experimental χexp(k) using the curved wave theory. The fit parameter reported for each refinement procedure is given by the statistical R-factor, defined as: R = ∑ i χ i e x p - χ i c a l c k W T 2 / ∑ i [ χ i e x p k W T ] 2 x 100 % kWT is the weight factor and a k3 -weighting was used for the analysed data. Ab initio phase shifts were also calculated within DL-EXCURV and verified using reference compounds. The least-squares refinements were carried out in typical wave number k range 2-8.5 Å-1 for cobalt and k range 3.5-9.5 Å-1 for rhenium using a k3 weighting scheme.Bimetallic fractions were calculated from the multiplicities of the Re LIII-edge from the EXAFS analysis after the method of Shibata et al. [31] The coordination number of the bimetallic contribution (NRe-Co) was used to determine the ratio of bimetallic phase compared to the total coordination number (NRe-Re + NRe-Co).Ammonia decomposition reactions were carried out in a continuous differential packed bed reactor using 25 mg of catalyst diluted in a silicon carbide bed. The reactor system was equipped with mass flow and temperature controllers. The reaction tubing was heated to 60 °C to avoid ammonia condensation and consequent corrosion. Prior to each catalytic reaction study, the catalysts were pre-reduced in situ under a H2 flow at 600 °C for 1 hour (unless otherwise stated). After pre-reduction, the temperature was returned to ambient under the H2 flow. Following this, the reaction temperature was ramped from room temperature to 600 °C using a Carbolite tube furnace equipped with a PID controller. A 2.6 °C min-1 ramp under 2.5 mL min-1 NH3 and 6 mL min-1 He (GHSV of 6000 mLNH3·gcat -1 h-1) was applied. The reactor exit gas was analysed using an on-line gas chromatograph fitted with a Porapak column and employing a thermal conductivity detector. Mass balance calculations were carried out to account for the molar expansion occurring as a result of the reaction. Mass balance was achieved within a ± 10 % error.A number of unsupported bimetallic combinations including CoRe1.6, Ni2Mo3N, Co3Mo3N were tested for hydrogen production from ammonia decomposition (Fig. 1 a) with the aim of achieving similar activities than the state-of-the-art ruthenium-based catalysts by combining transition metal catalysts with respectively higher and lower N-adatom adsorption energy than ruthenium. In addition to metal based catalysts, nitrides have also been investigated for ammonia decomposition, for example. [32,33] While Co-Mo alloy has been previously predicted as an optimum bimetallic combination [17,34,35], a considerably higher activity and, most importantly, an onset of activity at lower temperatures was achieved by the Co-Re alloy. Indeed, CoRe1.6 has an activity comparable to 7 wt.% Ru/CNT [11]. 7 wt.% Ru/CNT is considered to be one of the optimum catalysts for this reaction due to the high concentration of B5 sites (an arrangement of three Ru atoms in one layer and two further Ru atoms in the layer directly above) expressed in the 3.5-5 nm Ru nanoparticles present which are promoted by the CNT support [36]. While 3-5 nm supported Ru nanoparticles (7 wt.% Ru/CNT) present a considerably higher activity than the unsupported CoRe1.6 per mol of metal (Fig. 1b), a rate a few orders of magnitude higher is shown by the CoRe1.6 when activity is normalised by metallic surface area. It is important to highlight that the surface area of the unsupported CoRe1.6 catalyst is only 0.2 m2 g-1 as determined from the BET measurement (although the limitations in this regard with such a low surface area must be recognised) while the metallic surface area of the 7% Ru/CNT catalyst is 10 m² g-1 as measured by CO chemisorption. Pre-reduction in a H2 flow was undertaken prior to surface area determination in both cases. Whilst the limitations of the BET method applied to CoRe1.6 in view of its very low surface area are acknowledged, triplicate analyses confirmed the absence of micro- and meso-porosity. Considering the surface catalysed nature of ammonia decomposition, these results imply that the active sites in the Co-Re system might be considerably more active than those in their ruthenium counterparts. Both catalysts show a similar activation energy with values of 91 kJ mol-1 and 85 kJ mol-1 for 7 wt.% Ru/CNT and unsupported CoRe1.6 respectively (the Arrhenius plots are available in the SI, Figure S1). These values are in agreement with those previously reported for Ru-based catalysts [11,19] and may possibly suggest similarities in the nature of the active sites in both systems, thereby confirming the potential of the design principle adopted, although this is a matter for further exploration.As shown in Fig. 2 , cobalt only and rhenium only counterpart catalysts show limited activity for ammonia decomposition demonstrating that the high activity of the CoRe1.6 material is due to the synergetic effect achieved by the alloy formation [25].The activity of the CoRe1.6 catalyst was significantly increased as the pre-reduction temperature of the catalyst was increased (Fig. 3 ) from 400 to 600 °C. As the catalysts have been calcined at 700 °C under air, the role of pre-reduction is most likely not related to the promotion of the thermal interaction between the Co and the Re atoms but rather the oxidation state of the active species. However, temperature programmed reduction (TPR) of the CoRe1.6 catalyst (Figure S2 in the SI) reveals full reduction of both Co and Re components at 400 °C under hydrogen indicating that the apparent increase in catalytic activity at higher pre-reduction temperatures is likely to be associated with thermally-induced modifications in the bi-metallic material as evidenced below. For comparison, Figure S2 also shows the TPR of the calcined cobalt and rhenium precursors where full reduction requires temperatures of ∼ 450-500 °C.To gain a better understanding of the promotion of the interaction between the Co and Re during the pre-reduction of the CoRe1.6 material, in situ X-ray absorption spectroscopy was applied. It is particularly applicable to the current study where there could be concerns that the ex situ nature of the material may differ from that under reaction conditions. Attention was particularly directed to the near edge (XANES) region as it is highly sensitive to the local environment and oxidation state. Fig. 4 reveals that a cobalt intermediate state is formed during reduction at temperatures above 100 °C comprising an apparent mixture of Co2+ and Co0 species. Full reduction to Co0 occurs over a narrow temperature window starting at 350 °C with the final reduced state being complete at 400 °C. Reduction of rhenium occurs abruptly and in one step initiating at 300 °C. Fig. 5 presents spectra determined at different stages of the activation and reaction process.During heating of the CoRe1.6 catalyst in 5% NH3, partial oxidation of cobalt is evident by the increased white line intensity observed between 200-400 °C [37,38](Fig. 5c), as in agreement with supported CoRe in a silica aerogel. [39] Even if changes were observed in the more sensitive XANES for cobalt, EXAFS analysis (Table 1 ) revealed no light atom scattering pairs (i.e. Co-N) formed during the low-temperature ammonia treatment. Hence, any oxidation must be limited to the surface which would be difficult to distinguish in the EXAFS analysis of CoRe1.6. Starting at 400 °C during ammonia treatment, the white line intensity (Fig. 5d) gradually decreased reaching a similar intensity at 600 °C as observed for CoRe1.6 during pre-treatment. This corresponds to a partial reduction of cobalt coinciding with CoRe1.6 becoming active for ammonia decomposition. Interestingly, no changes were observed in the XANES region (Figure S3 in the SI) for rhenium during ammonia decomposition. These observations are in agreement with the widespread use of Re as a promoter in cobalt-based catalysts for the Fischer-Tropsch reaction where it facilitates not only Co reduction but also retards its sintering [40].Results from EXAFS analysis (Table 1) and average coordination numbers for the unsupported CoRe1.6 catalysts from pre-reduction between 400-600 °C show an average multiplicity of the Co-Co pair between 3.5 and 4 at around 2.45 Å, while the first Re-Re shell at 2.64 -2.69 Å contains on average 6-7 neighbours. While no Co-Re contribution could be fitted, a small Re-Co contribution could refined between 2.55-2.57 Å with an average multiplicity between 1.2-1.4. Attempts were made to refine the corresponding Co-Re shell, however they were not successful. This can be explained by Co-Co and Co-Re backscattering pairs having similar bond distances which were not resolved by our limited Δk-window. Attempts were also made to include mixed site option in the EXCURV software, essentially forcing Co-Co and Co-Re shells to have the same bond length, however, this did not improve the fit-factor and therefore the Co-Re contribution was removed. It is believed that the refinement of the Co-Re contribution is limited by the experimental constraints, i.e. filling mode (16-bunch), high rhenium content and elevated temperatures. EXAFS refinements indicated a complete reduction of CoRe1.6 from 400 °C in 75% H2 as no metal-O pairs could be fitted, in agreement with the TPR results described above. The addition of a metal-O pair in the fitting model for Fig. 6 a) and Fig. 7 a) produced no valid results and did not improve the fit. This reveals that the major contributing species formed during pre-treatment of CoRe1.6 are Co-Co and Re-Re with only ∼20% of the bimetallic Re-Co pair.After cooling and switching to 5% NH3, partial oxidation of Co was observed in the XANES, however this was not reflected in the EXAFS (at 300 °C) where four Co-Co pairs at 2.45 Å were found similarly to the pre-treatment. A better fit was obtained at the Re LIII-edge at lower temperatures yielding an average of 3.3 Re-Re backscattering pairs at 2.65 Å. We believe the lowered multiplicity originates from better accuracy and is not a reflection of reduction in particle size. The Re-Co bimetallic pair remained but at an elongated bond distance of 2.61 Å with a slightly increased average multiplicity to 1.6. Hence, there were minor structural changes observed in the EXAFS for CoRe1.6 when switching from a hydrogen to an ammonia atmosphere (Fig. 6 and Fig. 7)When increasing the temperature from 300 °C to 600 °C under the ammonia atmosphere, the Co-Co backscattering pair was surprisingly stable (Fig. 8 and Fig. 9 ) despite the observed partial oxidation and re-reduction observed in the XANES. These observations align with the excellent stability of the CoRe1.6 catalysts under consecutive reaction runs as shown below. The average Co-Co multiplicity remained at 4 with a bond distance of 2.45 Å throughout the ammonia treatment. This contrasts with the Re-Re backscattering shell where significant elongation of bond distance from 2.65 Å to 2.73 Å was observed (Fig. 9) during ammonia decomposition. Great care must be taken when comparing coordination shell distances from EXAFS at different temperatures, as other effects, such as thermal expansion, may contribute to the observed changes. [41] While multiplicities of Re-Re, Re-Co and Co-Co remained constant after the reduction stage, we observed variations in bond distances for these coordination shells as mentioned above. Fig. 9 illustrates these changes, ΔR, for these shells during pre-treatment and ammonia decomposition. During the pre-treatment step, relative changes of the first metal coordination shells were apparent relative to their initial appearance (400 °C), while relative changes during ammonia decomposition were shown in relation to the pre-treated sample. While the Co-Co bond distance remained unchanged during pre-treatment, an observed elongation of the Re-Re bond (2.65-2.69 Å) concurs with a slight contraction of the Re-Co bond (2.57-2.55 Å). This strongly indicates that local restructuring of both monometallic and bimetallic particles occurs between 400-600 °C, after TPR and MCR data suggest CoRe1.6 is fully reduced. After switching to ammonia at 200 °C the Re-Re bond was significantly shortened to 2.65 Å and the Re-Co distance is elongated (2.61 Å). During ammonia decomposition however these two distances were separated as Re-Re increased to 2.73 Å, while Re-Co was shortened (2.56 Å) coinciding with CoRe1.6 becoming active. While the observed elongation in bond distance approaches that of Re-foil (2.74 Å), it did not seem to be associated with sintering of a pure Re-Re phase in CoRe1.6 as the average multiplicity remained between 6-7 similarly to that observed in the H2 pre-treatment. Similar bond distances (2.56-2.57 Å) were found for rhenium promoted Co/Al2O3 during CO oxidation by in situ XAS. [42]It should be noted that higher order cumulants [43] (C3 and C4) were introduced during EXAFS refinements, however they did not improve the fits for either rhenium or cobalt. Somewhat ordered higher coordination shells are observed at the Co K-edge (Fig. 6) at low temperatures during pre-treatment, but these could not be fitted. However, these shells seem to disintegrate at higher temperatures and during ammonia treatment. Higher coordination shells were not observed at the Re LIII-edge (Fig. 7) at any reaction stage. This indicates a high degree of disorder in the CoRe1.6 during ammonia treatment. This is also the case for the material for ammonia synthesis. [25]It is of interest to monitor the bimetallic Re-Co contribution during the reaction stages and we have included two methods where we utilise information from XANES and EXAFS. Bimetallic fractions calculated from EXAFS (Fig. 10 , bottom) are somewhat stable between 0.1 and 0.3 through the reaction stages. However, due to high uncertainties of the Re-Re multiplicities especially, the XANES has been examined to look for features associated with the Re-Co bimetallic pair. The difference spectra (Fig. 10, top, black line) between Co-foil and the pre-treated CoRe1.6 show several features, one at 7712 eV associated with the pre-edge and a negative peak at 7722 eV together with a peak at 7730 eV related to the white line. Of these, the feature at 7712 eV would be more characteristic for assigning Re-Co bimetallic species, as the latter two can also be affected by oxidation and reduction processes. The feature at 7760 eV will be affected by temperature and particle size and is therefore less reliable for this purpose. The difference spectra for CoRe1.6 during ammonia treatment (Fig. 10, top, green line) monitor possible changes in the XANES compared to the final state at 600 °C in ammonia. It is evident that there are no significant changes in bimetallic Re-Co contribution during the process, rather the observed changes are attributed to partial oxidation and re-reduction.These results are similar to the behaviour of CoRe1.6 during ammonia synthesis where the Ar/H2 pre-treated material shows a stable bimetallic fraction throughout the process. [25] The complete reduction observed from the lack of Co-O and Re-O interactions in the EXAFS confirms that all of the metal is in the metallic state, and while the system is made up of Co-Co, Re-Re and Co-Re/Re-Co coordination pairs, the local structure of each particle is more difficult to predict.Results from XAS reveals that the CoRe1.6 catalyst remains largely unchanged during the ammonia treatment. This also precludes nitride formation as no Co-N/Re-N coordination shells were found during any stage of the process with the interaction with reactants and productsThe stable Co-Re contributions are in agreement with the excellent stability of the unsupported CoRe1.6 catalysts in at least 6 consecutive ammonia decomposition runs up to 500 °C where the temperature is returned to ambient values after each run before being increased again (Fig. 11 ).This paper provides new information for the development of active ammonia decomposition catalysts using bi-metallic combinations. Unsupported CoRe catalysts present an ammonia decomposition activity comparable to the state-of-the-art Ru-based systems with high stability under consecutive runs. The high activity is related to the bimetallic Co-Re contribution with no significant change across the studied temperatures with the exception of partial oxidation and re-reduction of the cobalt species. The re-reduction in the presence of the Re component coincides with the onset of ammonia decomposition activity. In addition, no Co-N or Re-N backscattering shells are found during EXAFS analysis under ammonia atmosphere. In terms of the CoRe catalyst, future attention should be directed towards the application of supports to increase the dispersion of the active phase and also to the application of potential promoters, whilst recognising the relative high cost of the Re component. Karsten G. Kirste: Investigation, Data curation, Writing - original draft, Writing - review & editing. Kate McAulay: Investigation, Writing - review & editing. Tamsin E. Bell: Investigation, Data curation, Writing - original draft, Writing - review & editing. Dragos Stoian: Investigation. Said Laassiri: Investigation, Writing - review & editing. Angela Daisley: Investigation, Writing - review & editing. Justin S.J. Hargreaves: Supervision, Funding acquisition, Writing - review & editing. Karina Mathisen: Supervision, Funding acquisition, Writing - original draft, Writing - review & editing. Laura Torrente-Murciano: Supervision, Funding acquisition, Writing - original draft, Writing - review & editing.The authors report no declarations of interest.The authors would like to acknowledge UK Engineering and Physical Science Research Council (EPSRC grant numbers EP/L020432/2, EP/N013778/1, EP/L02537X/1 and EP/L026317/1) for funding. The Norwegian University of Science and Technology and the Norwegian Resource Council is acknowledged for grants supporting the Swiss-Norwegian Beamlines (SNBL) and K.G. Kirste and K. Mathisen acknowledge the grants from the Anders Jahre fund for promotion of science. The assistance of beamline scientists M. Brunelli and W. van Beek is very much appreciated.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcatb.2020.119405.The following is Supplementary data to this article:

On-demand production of hydrogen from ammonia is a challenge limiting the implementation of ammonia as a long term hydrogen vector to overcome the difficulties associated with hydrogen storage. Herein, we present the development of catalysts for the on-demand production of hydrogen from ammonia by combining metals with high and low N-adatom adsorption energies. In this way, cobalt-rhenium (Co-Re) catalysts show high activity mimicking that of ruthenium. EXAFS/XANES analyses demonstrate that the bimetallic Co-Re contribution is responsible for the activity and the stability of the catalysts in consecutive runs with no observable formation of nitrides (Co-N and Re-N) occurring under the ammonia atmosphere. While cobalt is partially re-oxidised under ammonia, re-reduction in the presence of rhenium is observed at higher temperatures, coinciding with the on-set of catalytic activity which is accompanied by minor structural changes. These results provide insight for the development of highly active alloy based ammonia decomposition catalysts.

Dichromate is an essential additive in the chlorate process (the electrolytic production of NaClO3 from NaCl), to catalyze the chlorate formation as well as to inhibit side reactions. Thus, no chlorate production is possible in the current technology without adding chromium(VI) (Colman and Tilak, 1995). However, chromium(VI) compounds are in the REACH annex XIV list which means that a special authorization is required to use them within the EU after 2017. Chlorate is produced by electrolysis of sodium chloride in an undivided cell, where chlorine is produced at the anode and hydrogen at the cathode (Cornell, 2014a; Cornell, 2014b). Chlorine is immediately hydrolyzed in the electrolyte forming hypochlorite/hypochlorous acid (their ratio is defined by the pH) which disproportionate into chlorate and chloride ions. The latter reaction is catalyzed by chromium(VI) in the electrolyte. The chromium(VI) additive has further functions in the process and large efforts have been spent to find alternative components for replacing it in all its roles in the chlorate process (Endrődi et al., 2017; Endrődi et al., 2019).The uncatalyzed decomposition of hypochlorite has been thoroughly investigated over the years and is known to follow overall third order kinetics (Adam et al., 1992). Furthermore, the maximum rate of the reaction occurs at a pH where the ratio of HOCl and OCl− is 2:1. A deeper understanding of the mechanism of the uncatalyzed reaction was recently achieved in a theoretical study showing that the reaction is initiated by a fast equilibrium between HOCl, OCl−, Cl2O and Cl3O2 − and the subsequent abstraction of Cl− to form Cl2O2 is the rate determining step in chlorate formation (Szabó et al., 2018). The catalytic effect of chromium(VI) in the chlorate process has been considered for a long time, and lately this effect has been explored in detail (Endrődi et al., 2019).Replacing chromium(VI) as a catalyst in the chlorate process is a particularly challenging task due to the extreme operating conditions applied in the corresponding technology (Endrődi et al., 2017; Sandin et al., 2015; Busch et al., 2019). In our previous work, the chromium(VI) catalyzed decomposition of HOCl was thoroughly studied, and it was concluded that the catalytically active species is CrO4 2−. Such an effect was not observed with the structurally analogous phosphate ion, thus it was concluded that the catalytic activity of CrO4 2− is associated with a partial electron transfer process in the transition state. This enhances the conversion of Cl2O into HCl2O2 − (Kalmár et al., 2018). The main objective of this study was to find efficient alternative catalysts for the conversion of hypochlorous acid into chlorate ion. We mainly focused on compounds which were expected to feature the noted partial electron transfer phenomenon. The experiments were performed at elevated temperature (80°C) in order to mimic the process conditions.All chemicals were of analytical reagent grade, purchased from commercial sources and used as received, without further purification. Doubly-deionized and ultrafiltered (ELGA Purelab Classic system) water was used to prepare the stock solutions and samples. Sodium hypochlorite (NaOCl) solution was prepared by bubbling gaseous chlorine into sodium hydroxide solution. The stock solution of NaOCl was standardized by iodometric titration. A Metrohm 785 DMP Titrino automatic titrator equipped with a Metrohm 6.0451.100 combination platinum electrode was used. The excess NaOH concentration was determined by pH-metric titration with standard HClO4. In this case, a Metrohm 6.0262.100 combination glass electrode was used.The decomposition reaction was triggered by simultaneous addition of NaOCl and the catalyst to well stirred aqueous perchloric acid solution. In the case of heterogeneous systems, the progress of the reaction was monitored by taking individual samples from the reaction mixture at different reaction times, and the nominal concentration of the catalyst is given, i.e. the weighted amount of catalyst divided by the volume of the reaction mixture. Obviously, the heterogeneous catalysts were not dissolved, and the catalytic process occurred predominantly on the surface of the catalyst. The decomposition reaction was studied at 80±0.1°C and the samples were stirred with a magnetic stirrer. High ionic strength was not set in these experiments because it would have saturated the ion chromatographic column in the ionchromatographic experiments. Thus, the ionic strength was always defined by the ionic forms of the reactants and the catalyst in the samples. Since some of the species are involved in acid-base equilibria the actual total concentration of the ions was also affected by the pH. It follows, that constant ionic strength could not be used in these studies, it was somewhere between 0.10 and 0.15M (estimated value). Within this range, significant ionic strength effects are unlikely on the kinetics and stoichiometry, and the corresponding results are directly comparable.To measure the pH of the inhomogeneous reaction mixtures, a special, Metrohm Unitrode Pt 1000 (6.0258.010) electrode was used equipped with a temperature sensor unit. Before use, it was confirmed that the electrode is reliable and reproducible in heterogeneous model systems. The electrode was calibrated every day at 80°C using KH-phthalate (c =0.05M, pH=4.159) and borax (c =0.01M, pH=8.910) standard solutions (Covington et al., 1985). In this study, the pH readout was not converted into log[H+] to correct for the ionic strength effect as recommended by Irving et al. (1967). Accordingly, the readout of the pH meter is plotted as pH in the corresponding figures. It needs to be emphasized that the actual correction factor (Irving factor) is electrode specific and may exceed ±1 pH unit at 80°C. In principle, the correction should have been made point by point due to the lack of constant ionic strength even within a kinetic run and the presence of the heterogeneous phase. Obviously, such a procedure is not feasible. In general, the correction of the pH readout could have varied within 0.1–0.2 pH unit only.During a kinetic run, 1mL sample was retracted from the reaction mixture in every 5min. This aliquot of the sample was immediately cooled to 25°C in an ice bath and filtrated with regenerated cellulose membrane-filter (pore size 0.45μm). The filtered solution was diluted using NaOH as quenching agent. The final concentration of NaOH was 0.1M. In the case of YCl3, a small amount of hydroxo precipitate formed which was removed by a second filtration. UV-Vis spectra of these solutions were recorded in the 200–400nm wavelength range on an Agilent-8453 diode array spectrophotometer. It was confirmed that unwanted photochemical side reaction did not occur during the measurements (Fábián and Lente, 2010). The amount of NaOCl was quantified at the absorption band of OCl− (λ max =292nm, ε =339.5M−1 cm−1).The formation of chlorate as a function of time was monitored with a Thermo Scientific Dionex ICS-5000 + ion chromatographic system by using a 25μL injector loop. Isocratic elution was carried out using NaOH solution (0.020M). The method was calibrated by a dilution series of chlorate solutions. In each system, the concentration of the product chlorate ion as well as the concentration change of hypochlorous acid was measured as a function of time.The protonation constants (pK a) of telluric acid were determined by pH-potentiometric titration method using a carbonate free NaOH solution (ca. 1M). The carbonate contamination was determined using the appropriate Gran functions (Gran, 1952). In this titration, 45mL aliquots of telluric acid (ca. 0.018M) were titrated using NaCl (I =1.7M) and NaClO3 (I =4.7M) as background electrolytes. The headspace over the sample was purged with argon to ensure the absence of oxygen and carbon dioxide. The pH measurements were made using a Metrohm Unitrode Pt 1000 (6.0258.010) electrode. In this case, the pH reading was converted to hydrogen ion concentration as described by Irving et al. (1967). The protonation constants were calculated by using the designated computational program, SUPERQUAD (Gans et al., 1985).The rate of the decomposition reaction of hypochlorous acid depends on various parameters, especially on the pH. In the present study, it is assumed that the actual pK a of HOCl is around 6.79 obtained at 80°C, 0.5M NaCl (Wanngård and Wildlock, 2017). No attempt was made to determine the exact pK a for the conditions applied here for the following reasons. The main goal of this study is to establish how various substances catalyze the decomposition of HOCl as a function of pH, but it is not studied which form of a given species is active in these reactions. Because the pH decreases steadily, the [OCl−]/[HOCl] ratio always changes significantly over the course of the reaction. Thus, the interpretation of the general trends and the comparison of the results at different pH values does not require the exact knowledge of the pK a. A more quantitative approach would require a different set of experiments at constant ionic strength which would not make possible the stoichiometric measurements with the method used here.As it was reported in the case of Cr(VI) catalyzed decomposition of HOCl, the catalytic reaction path is also pH dependent (Kalmár et al., 2018). This was interpreted by considering that only one form of Cr(VI) is catalytically active and the speciation of Cr(VI) is controlled by the pH (Szabó et al., 2018). The very same features are expected in any system where the potential catalyst is involved in pH dependent equilibrium processes. The practical consequence of this is that the catalytic activity needs to be tested in a broader pH range. Thus, test experiments were run by varying the initial pH.The kinetic traces (HOCl decay) in the presence of the potential catalyst were compared to those obtained in the Cr(VI) catalyzed reaction. The results of control experiments are also presented under identical initial conditions in either non-buffered and phosphate buffered solutions in the absence of the catalyst. The comparison of these traces revealed significant differences in the pH profiles. This is quite reasonable if we consider that different acid–base reactions and, as a consequence, different buffering effects take place in the compared systems.Earlier it was established that the decomposition of hypochlorous acid may proceed via two distinct reaction paths (Busch et al., 2019), called the chlorate (Eq. (1)) and the oxygen (Eq. (2)) paths. (1) 3 HOCl = Cl O 3 − + 2 Cl − + 3 H + (2) 2 HOCl = 2 Cl − + 2 H + + O 2 It is noteworthy to mention that only the chlorate path is preferable for the industrial process and the oxygen path needs to be avoided as much as possible. We characterize the relative significance of the two decomposition paths by the stoichiometric ratio of the reactant consumed and product formed as follows: R =Δ[OCl−]/[ClO3 −]. When only the chlorate path is operative R =3.0. Higher ratios indicate that the oxygen path also contributes to the overall process. Accordingly, the main goal is to achieve R =3.0 in a catalytic system. At this point, it needs to be emphasized that the presence of chlorate ion impurities in the reactants may have significant effect on the value of R. It was noticed that a small initial amount of chlorate ion was always present as impurity in the reaction mixtures. It could originate either from the hypochlorite stock solution or from other reagents such as perchloric acid. This problem was circumvented by fitting the kinetic traces with a polynomial function using a non-linear least squares routine to estimate the initial concentrations of HOCl and ClO3 − (ORIGIN, 2014). In the case of HOCl, the estimated values were in excellent agreement with the values calculated by considering the dilution of the stock solutions. During the calculation of R, the measured concentration of chlorate ion was corrected by its initial concentration in each point. The uncertainties associated with this procedure are small and coherent results were obtained which are useful to establish the major trends in these reactions.In agreement with the above considerations, the following requirements need to be satisfied by an ideal catalyst. It must accelerate the decay of HOCl and the catalytic process should solely progress via the chlorate path.Preliminary experiments have demonstrated slight catalytic activity with Y2O3, therefore, we explored the role of this species and the related YCl3 in more detail. Typical chromatographic peaks of chlorate ion in the spent reaction mixtures in the absence and presence of the catalysts are shown in Fig. S1. It is quite obvious that approximately the same amounts of chlorate ion are formed in the control experiment and in the presence of Y2O3. The concentration of chlorate ion is considerably smaller with YCl3 which is already a strong indication that this compound promotes O2 formation.Kinetic traces for the decomposition of HOCl at initial pH=7.2 are shown in Fig. 1 . In the presence of YCl3, the reaction is faster than in the control experiment or with Y2O3 (Fig. 1a). In accordance with Eqs. (2) and (3), the decomposition of hypochlorous acid always generates hydrogen ion and the pH profile as a function of time strongly depends on the acid–base side-reactions in the reaction mixture (Fig. 1b). The pH drops suddenly in the control experiment at around 3000 s because the buffer capacity of the system diminishes when sufficient amount of HOCl decomposes. In the presence of Y2O3 or YCl3, the reaction mixtures “self-buffer” themselves to the pH 5.5–6.0 region. In the case of YCl3, the pH decreases below 6.0 much earlier than in the other systems. This is the consequence of the hydrolysis of the catalyst which generates substantial amounts of proton. The kinetic traces in the control experiment and in the presence of Y2O3 are identical for all practical purposes (see also Fig. S2). This confirms the lack of catalytic effects under such conditions. Phosphate ion exerts significant buffer capacity and the pH cannot decrease to the optimum range for the decomposition in the first hour. As a consequence, the reaction proceeds slower than in the control experiment. As expected on the basis of our previous studies, CrO4 2− has a marked catalytic effect (Kalmár et al., 2018).As shown in Fig. 2 , the final R is around 4.0 in the control experiment and in the presence of Y2O3, PO4 3− and CrO4 2−. In the case of YCl3, R is above 40 confirming that the O2-path is dominant in the decomposition of hypochlorous. Very similar results were obtained over the entire studied pH range, i.e. when the initial pH was systematically varied from pH 7.5 to 6.15.The actual form of Y(III) has a significant role on the overall process. When an YCl3 solution is added to the neutral or slightly alkaline reaction mixture, a gelous hydroxo precipitate forms immediately. This precipitate is presumably Y(OH)3 or some sort of an oxo–hydroxo precipitate with unknown stoichiometry. The solution becomes opaque indicating the presence of a colloidal system. Accordingly, the precipitate is expected to have a relatively large specific surface area. The morphology and the surface of this precipitate are very different from that of Y2O3. Apparently, the surface of this precipitate is an excellent catalyst of the O2 path, similarly to the hydroxides of other metal ions such as Ni2+, Co2+, Cu2+, etc. In contrast, Y2O3 is a well-defined compound with well-defined surface structure. This may explain the difference between the catalytic activities of these compoundsWhile Y2O3 is not an active catalyst of O2 formation, it slightly catalyzes the chlorate path under slightly acidic conditions . As shown in Fig. 3 , the decomposition becomes somewhat faster in the presence of Y2O3 when the pH is decreased.In the control experiments, the usual pH profiles were observed (Fig. S3a). As expected, the sudden pH change occurs at shorter reaction times when the initial pH is lower. The final pH is between 3 and 3.5 in all cases. In contrast, the final pH is between 5.5 and 6.0 in the presence of Y2O3 regardless of the starting pH (Fig. S3b). This implies that Y2O3 consumes the proton formed in the decomposition process in an acid-base reaction.The titration of an aqueous suspension of Y2O3 with HClO4 confirms the existence of a protonation process and, as a consequence, the appearance of a buffered region in the titration curve (Fig. S4). It should be added that the shape of the titration curve is highly dependent on the duration of the titration. Slower dosing of HClO4 results in a longer buffered region in the titration curve indicating that the acid consumption process is relatively slow. In any case, these titrations prove the origin of the buffering effect of Y2O3 in the corresponding experiments. It is possible that the noted catalytic effect (Fig. 3) is a buffering effect in reality. It is well known that the rate of decomposition decreases by decreasing the pH below pH ∼7.2. Thus, the reaction proceeds at higher rate in the presence of Y2O3 because it does not allow the pH to drop suddenly.The comparison of Fig. 4a and b confirms that the oxygen path is somewhat more pronounced in the control experiments than in the presence of Y2O3. It is quite apparent that the value of R increases more significantly when the starting pH of the control experiments is decreased. This leads to the conclusion that acidic conditions promote the decomposition of HOCl into O2.In the case of telluric acid (Te(OH)6) remarkable results were obtained under slightly acidic conditions (Fig. S5). The kinetic traces of decomposition are compared in the presence of Y2O3, telluric acid and the control experiments. The results in the presence of phosphate ion are not included in the comparison because the buffer capacity of phosphate ion vanish at this relatively low pH. In the presence of telluric acid, about 80% of the initial HOCl decomposes in about 2h. Furthermore, the stoichiometric ratio (R) is about 3 (Fig. 5 ). This excludes that the oxygen path has a significant role in this system. As shown in Fig. S6, telluric acid has a considerable buffer capacity at the studied pH, because the pH decreases only slightly compared to the control experiment. Furthermore, the acceleration of the decomposition is more pronounced with Te(OH)6 than with Y2O3 although the pH converge to about the same value at about 2h reaction time in both cases. Thus, buffering effect alone cannot be responsible for the faster reaction and it is reasonable to assume that a real catalytic process is observed in the Te(OH)6 system.In order to explore whether decomposition is catalyzed or the oxidation of HOCl by telluric acid occurs, the following experiment was designed. First, the decomposition reaction was monitored for 2h. After this period, an aliquot of HOCl was added to the reaction mixture bringing the HOCl concentration and the pH to the initial values. Then, the decay of HOCl was monitored again. The consecutively kinetic traces were very similar (Fig. 6 ), confirming that a real catalytic process takes place.A systematic pH dependent study reveals that the optimum pH of decomposition is at somewhere around 6.7–6.9 (Fig. S7).The noted pH effect is most likely associated with the acid–base equilibria of telluric acid. In accordance with literature results, the titration curve of acidic telluric acid solution with NaOH is consistent with two acid dissociation steps (Fig. S8). The corresponding pK a-s are strongly dependent on the conditions applied. The speciation diagram as a function of pH was calculated by using pK 1 =7.5±0.1 and pK 2 =9.3±0.1 (Fig. 7 ). The estimated pH values are in good agreement with those available in the literature (Filella and May, 2019; McPhail, 1995). The results confirm that the first deprotonation step coincides with the optimum pH region of the decomposition. This implies that somehow the Te(OH)6 and TeO(OH)5 − forms are active in the catalytic process.The comparison of the results obtained in the presence of chromium(VI) and telluric acid reveals that the former is more active catalyst (Fig. S9), but the stoichiometries of the reaction are very similar with both species (Fig. 8 ).The results presented here reveals that the decomposition of HOCl is clearly accelerated by YCl3, Y2O3 and Te(OH)6. By comparing the stoichiometries and relative rates of the catalytic reactions, only telluric acid seems to be a useful candidate to replace Cr(VI). The comparison of the results obtained in the presence of chromium(VI) and telluric acid reveals that the former is more active catalyst, however, its adversary impact on workers health may justify the introduction of Te(OH)6 as an alternative catalyst in the chlorate process. Further studies should be directed toward understanding the details of telluric acid catalysis and exploring effects of this catalyst on the electrochemistry of the industrial process. It is important to note that the higher price of the latter catalyst is an important issue regarding 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.N.L. is indebted to the New National Excellence Program of the Ministry for Innovation and Technology from the source of the National Research, Development and Innovation Fund (ÚNKP-20-4-II).Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.cherd.2021.03.010.The following are the supplementary data to this article:

By pursuing the aim of identifying new types of catalysts in the electrolytic production of NaClO3 from NaCl (chlorate process), we report a detailed study on the decomposition of hypochlorous acid accelerated by yttrium(III) chloride (YCl3), yttrium(III) oxide (Y2O3) and telluric acid (Te(OH)6) at 80°C. The results were compared with those obtained in the uncatalyzed and chromium(VI) catalyzed reactions. In general, the decomposition of HOCl occurs via two competing paths toward the formation of ClO3 − or O2. In the case of YCl3, the decomposition proceeds via the oxygen path over the entire studied pH range. Y2O3 slightly catalyzes the chlorate path under acidic conditions, however the noted catalytic effect is probably due to the “self-buffering” of the reaction mixture (Y2O3 suspension). Although, a real catalytic process takes place in the presence of Te(OH)6, a significant pH-effect is also observed which is most likely associated with the acid–base equilibria of telluric acid. pH dependent studies demonstrate that the optimum pH of decomposition is at around 6.7–6.9 in this case. The comparison of the results obtained in the presence of chromium(VI) and Te(OH)6 reveals that the former is a more active catalyst. On the basis of kinetic and stoichiometric results, it is reasonable to assume that Te(OH)6 may be utilized as an alternative catalyst in the chlorate process.

The indole structure is one of most important heterocycles appears in many popular drugs, agrochemicals, advanced organic materials as well as bioactive alkaloids [1–7]. Notably, the indole scaffold has been known as an important pharmacophore in medicinal chemistry present in over 3000 natural products and 40 pharmaceutical compounds [1]. Especially, bis(3-indolyl)methane derivatives (BIMs) are highly important molecules due to their presence in the core structure of many bioactive natural alkaloids (arundine, vibrindole A, arsindoline A, barakacin, etc.). They also play an important role in develoment of novel bioactive compounds (anti-inflammatory, anticancer and antimetastatic etc.) [3–10].The construction of indole structure from simple building blocks based on the cyclization reactions in the absence or presence of metal catalysts has been well established [11,12]. However, the construction of large molecules containing more than one indole moieties in the structure from simple building blocks is a challenging issue. Due to the importance of BIM derivatives in the development of novel bioactive molecules, many new synthetic methods to synthesize BIMs by using indole derivatives as starting materials have been disclosed [3,6,12].Most of reports based on the direct alkylation of indoles with aldehydes or ketones in the employment of Lewis or Bronsted acids [6,12]. In 2013, Bhaumik et al. reported the preparation of porous organic polymer bearing built-in-CO2H groups and described its utility an efficient heterogeneous carbocatalyst for the alkylation of indoles with benzaldehyde and secondary benzylic alcohol derivatives to form BIMs under room temperature [13]. As a result of developing green and sustainable processes, several new procedures for the preparation of BIMs by the direct transition metals-catalyzed coupling of indoles with a variety of alcohols (including aliphatic alcohols) have been demonstrated [14–19]. Grigg and coworkers reported the first isolation of BIM as a side product in the Ir-catalyzed alkylation reaction of indoles with alcohols [14]. In 2012, Liu group disclosed the convenient synthesis of BIM derivatives relied on the Ru-catalyzed reaction of indoles with benzylic alcohols [15]. One year later, Ohta et al. developed a practical Ru-catalyzed alkylation of indole with benzylic alcohols for 24 h at 110 °C [16]. In 2020, Srimani group have just disclosed the ruthenium pincer complex catalyzed transformations of indoles with alcohols to give either C3-alkylated indoles or BIMs through modifying reaction conditions [17]. Hikawa and Yokoyama reported a Pd-catalyzed domino process for the preparation of BIM derivatives involving C3H benzylation of indoles and benzylic CH functionalization in water [18]. In the recent development of cheaper and greener catalysts relying on base metals for the synthesis of BIMs, in 2014, Sekar group reported the first FeCl2/BINAM catalyst in the use of dicumyl peroxide as an oxidant for the synthesis of BIMs in moderate yields [19]. Even though these homogeneous metals catalysts often offer higher yield and selectivity, their practical applications in industrial processes are limited by the difficulty in separation and reuse of catalysts after reactions [20,21]. In addition, the transition metal contamination in desired products could be a serious issue in the pharmaceutical and fine chemical industries [20,21]. To overcome these drawbacks, Babazadeh and coworkers demonstrated the preparation of BIMs under air using Ni nanoparticles supported on ionic liquid-functionalized magnetic silica as a recyclable heterogeneous catalyst [22]. Very recently, a useful method for the synthesis of BIMs by the alkylation reaction of alcohols and indoles in the employment of Fe3O4@SiO2@TPP-Cu as the photocatalyst under blue LED light was described [23]. In general, the procedures for the synthesis of BIM derivatives often require utilization of well-designed catalysts or special operating conditions which are inconvenient in practical applications. In addition, these heterogeneous catalysts only can work well with benzylic alcohols, and aliphatic alcohols remains as challenging substrates in the formation of desired BIM products.In recent years, magnetic nanoparticles have been considered as efficient heterogeneous catalysts in importantly organic transformations due to their outstanding properties, for example, easy recyclability, very low amount of metals leaching and high catalytic activities [24]. Copper ferrite nanoparticles have been known as a practical heterogeneous catalyst for several important organic transformations such as hydrogen transfer, CN, CO, CS coupling, Sonogashira and Click reactions [25–29]. An additional advantage of using this catalyst is an easy recyclability by applying of an external magnet [24–29]. Therefore, we put our effort in exploring new interesting properties of this catalyst in alkylation reaction of indoles with alcohols as alkylating reagents. Herein, we are reporting an air stable and highly efficient CuFe2O4 catalyst system for alkylation of indoles with alcohols (including challenging aliphatic alcohols) to give BIM derivatives in high yields under mild condition. In addition, this CuFe2O4 catalyst could be considering as a sustainable catalyst due to the non-toxic, easy preparing and recyclable properties which are promising for finding potential applications in pharmaceutical and fine chemical industries.A mixture of Indole (0.3 mmol), Alcohol (1.2 mmol), CuFe2O4 (7.2 mg, 10 mol%) and LiOtBu (24 mg, 1 equiv.) were charged in a pressure tube. The pressure tube was immersed in a pre-heated oil bath at 80 °C and stirred for 24 h. After cooling, the reaction mixture was filtered over a plug of Celite with hot water to eliminate the excess of alcohol. The organic phase was washed by Ethyl acetate, then dried by sodium sulfate (Na2SO4). The concentrated residue was purified by column chromatography (Hexane/Ethyl acetate).A mixture of indole (0.3 mmol), alcohol (2 mmol), CuFe2O4 (7.2 mg, 10 mol%) and LiOtBu (48 mg, 2 equiv.) were charged in a pressure tube. The pressure tube was immersed in a pre-heated oil bath at 120 °C and stirred for 24 h. After cooling, the reaction mixture was filtered over a plug of Celite with hot water to eliminate the excess of alcohol. The organic phase was washed by Ethyl acetate, then dried by sodium sulfate (Na2SO4). The concentrated residue was purified by column chromatography (Hexane/Ethyl acetate).Our first optimizations for the alkylation of indole with benzyl alcohol were carried out. After screening several conditions using homogeneous copper catalysts (Cu(OAc)2, CuCl2, CuCl), we only obtained a mixture of hydrogen transferred product and bis(3-indolyl)phenylmethane (BIM). Indeed, BIM could be regioselectively formed under lower temperature. In order to maximize the formation of bis(3-indolyl)phenylmethane, other copper sources were examined. Interestingly, in the presence of CuFe2O4 nanocatalyst (10 mol%), we can achieve 95% yield of BIM product under mild condition (80 °C). Then, several nanoparticle oxides (CuO and Fe3O4) were also employed as the catalysts for this reaction, only low yields of the desired product were observed (Table 1 , entries 1, 2). Therefore, a synergistic effect in the cooperation of iron and copper metals in spinel structure would probably be counted for the high catalytic activity of CuFe2O4 nanocatalyst. In order to investigate the role of base, a series of bases have been examined. While KOtBu, KOH, K2CO3 bases did not give a satisfactory yield, LiOtBu was found to be the most suitable base for the formation of BIM product in highest yield (Table 1, entry 3). In the absence of any bases, only trace amount of BIM product could be formed after 24 h reaction (Table 1, entry 8). Interestingly, up to 84% yield of BIM product was formed in the presence of only 5 mol% of CuFe2O4 catalyst under the optimized reaction (Table 1, entry 7). This alkylation reaction was carried out at lower temperature, i.e. 60 °C, only a trace amount of BIM was observed. Interestingly, up to 93% isolated yield of BIM product was formed within only 12 h reaction. In this research, we believe that oxygen should be the oxidant in the oxidation step of alcohols to corresponding aldehydes. A control experiment was carried out in argon which being resulted in a very low yield of BIM product (Table 1, entry 12).With the optimized condition in hand, we like to explore the scope of this reaction. First, a series of benzylic alcohol derivatives were used in the alkylation of indole. In general, BIM derivatives with the tolerance of both electron donating and withdrawing groups were prepared in good to excellent isolated yields (Table 2 , compound 1–9). Interestingly, among these products, turbomycin B alkaloid and an anticancer agent were easily prepared in 95% and 78%, respectively (compounds 1, 4). The reactions of indole with p-trifluoromethylated benzylic alcohol and bulky 2-napthyl alcohol also afforded to corresponding products in 85% and 90% yields, respectively (compounds 10, 11). Notably, we could extend the scope of this reaction with aliphatic alcohols as well. In fact, aliphatic alcohols are challenging alkylating agents compared to benzylic alcohols due to the difficulty in their in-situ oxidation to the corresponding aldehydes. To the best of our knowledge, so far, there are no heterogeneous catalyst systems which can work properly in the alkylation reaction of indoles with aliphatic alcohols. Remarkably, up to 87% yield of the desired products (Table 2, compounds 12–15) containing aliphatic alkyl groups were obtained under harder condition (120 °C, 24 h). In addition, the CuFe2O4 catalyst also showed very good perfomance in the alkylation of indole derivatives with benzylic alcohol. Indoles bearing with methyl, bromine, chlorine and flourine groups were very well tolerated which gave high yields of the desired products (up to 93% yield) (Table 3 ). Attractively, an antileukemic agent was successfully prepared in 72% isolated yield (compound 20).Due to i) control experiments in Table 1 (entries 8, 12) demonstrated the essential role of base and oxygen in the transformation and ii) large amount of aldehyde was formed in the course of reaction, we would like to proposed a plausible mechanism as presented in Fig. 1 . The conversion is initiated by the adsorption of the alcohol on the surface of CuFe2O4 nanoparticles via the lone pair electron on oxygen atom of alcohol (structure A in Fig. 1) [30]. However, due to the weak interaction of this coordination, high activation barriers are usually required to activate the OH bond, as reported for this type of reaction on copper-based catalyst [31]. Therefore, this step is more feasibly facilitated via the proton-transfer reaction with the LiOt-Bu base. After the initial structure A has been activated, the generated alkoxy could be easily converted on metal oxides surfaces (intermediate B in Fig. 1) [32], generating the key aldehyde intermediate (II) and metal hydride species C. The hydride strongly bound to the surface of CuFe2O4, but subsequently is converted to metal hydroxy (M-OH (M = Cu or Fe)) by combining with molecular oxygen [33]. This step is very important in stabilizing the structure of metal oxide during the reaction, since the accumulation of surface hydrogen on the surface of metal oxides usually result in the fast reducing of metal oxides to metallic phases and lost the activity as were reported in our earlier integrated theoretical and isotope-labelling studies [30,32]. Furthermore, the surface hydroxyl group of those metal hydroxy can also facilitate the OH bond activation of the starting alcohol material via the low barrier hydrogen-abstraction reaction (intermediate E in Fig. 1), regenerating the metal alkoxy intermediate and H2O [30,31]. Indeed, the generation of H2O was also proposed by several previous reports. When aldehyde (II) was formed, it easily reacts with indole to give intermediate III which is converted to give intermediate IV via dehydration process. Finally, the most important step for the overall conversion is the 1,4-addition reaction between the intermediate IV and the indole to form BIM product (via intermediate F). Without the catalyst, the barriers of these type of reaction were reported as extremely high of ~46 kcal/mol in our earlier studies [32], hindering those reaction to be processed. The metal oxides surface therefore could stabilize the transition state of this reaction via coordinating with the lone pair electron of N atom and reduce the barrier of this step by 20–30 kcal/mol, making this reaction much more feasible [32]. Then, the formation of intermediate G is possible when the activated intermediate IV reacts with another indole molecule via 1,4-addition reaction. The last step is the desorption of BIM product from the surface of catalyst. Therefore, the crucial role of the bifunctional catalyst of CuFe2O4 is to facilitate the formation of key intermediate aldehyde and then drive the 1,4-addition reaction via the lower barrier pathway.The main advantage of heterogeneous catalysts to homogeneous catalysts is the stability and recyclability. Notably, by applying of an external magnet, the CuFe2O4 catalyst could be recycled at least five times without losing significant catalytic activity for the standard alkylation of indole using benzyl alcohol (Fig. 2 ). However, the catalytic activity of CuFe2O4 was gradually decreased after each run due to the aggreation of nanoparticles, as was evidenced by the TEM analysis. The TEM image of the fresh catalyst showed the presence of spherical CuFe2O4 nanoparticles with an average particle size of 27.56 ± 1.11 nm (Fig. 3A). The inter planar spacing of 0.258 nm was visualized by its HRTEM image which was characterized for the (311) crystal planes of cubic spinel CuFe2O4 structure (Fig. 3A, inset) [34]. However, the TEM image of the reused catalyst indicated that these CuFe2O4 nanoparticles were aggregated to much larger particles of 164.41 ± 4.82 nm (Fig. 3B). Furthermore, beside the presence of (311) crystal plane on the reused catalyst (corresponding to the inter planar spacing of 0.256 nm) [34], there was a new appearance of (220) crystal plane in the HRTEM image of the reused catalyst which was characterized by the inter planar spacing of 0.294 nm [35] (Fig. 3B, inset). This observation might imply that there were structural changes during the aggregation, resulting in the formation of less active (220) surface on the reused catalyst. As reported earlier in the study of Singuru et al. [32], the barriers for both CH and NH bonds activations on the (220) facet of metal oxides are significantly higher than on the (311) facet. Thus, two reasons for the gradual decrease of activity from the TEM analysis: (i) the nanoparticles were aggregated to larger particles during the reaction, resulting in the decrease of active areas and (ii) structural changes occurred during the nanoparticle aggregation, inducing the formation of less active (220) facet and consequently the catalytic activity.The XRD results were in good agreement with the TEM analysis. As shown in Fig. 1S (see supplementary material), the fresh CuFe2O4 catalyst displayed typical, resolved peaks at 2θ = 30.1°, 35.6°, 43.1°, 57.1°, 62.7° which can be attributed to Miller indices (220), (311), (400), (511) and (440) reflections of cubic spinel structure of CuFe2O4 [25–29]. After five cycles, the reused CuFe2O4 catalyst showed identical reflections compared to the fresh one, suggesting that the crystalline spinel structure was well preserved. Moreover, the reflections of the reused catalysts looked more pronounced with respect to those of the fresh one, confirming the formation of larger CuFe2O4 nanoparticles by the aggregation. The SEM and EDX images of the fresh and reused CuFe2O4 samples were also taken (Fig. 2S, supplementary material). The fresh CuFe2O4 sample exhibited fine, regular particles while the reused sample showed bigger and irregular ones, further confirming the occurrence of particle aggregations. However, the EDX images of both fresh and reused CuFe2O4 catalysts showed well-dispersion of inherent elementals, i.e. Cu, Fe and O in the studied catalysts, suggesting that the elemental composition was not changed during the reaction.We have reported the application of an active and stable heterogeneous CuFe2O4 nanoparticles catalyst for the efficient alkylation of indoles with a series of alcohols to form BIM derivatives under mild operating conditions in air. Especially, we successfully figured out a good solution for the utilization of challenging aliphatic alcohols as alkylating reagents in the alkylation reaction of indoles which resulted in high yields of the desired products. O2 in air was confirmed to be the oxidant in the oxidation of alcohol to aldehyde. The CuFe2O4 catalyst is highly stable and can be recycled at least five times without losing the significant catalytic activity. Several applications of the CuFe2O4 catalyst were described for the synthesis of various important products, such as turbomycin B alkaloid, anticancer and antileukemic agents in high yields. This procedure would be considerable interest for exploring synthetic applications in medicinal chemistry. Further studies to understand the mechanistic insights of this reaction and actual role of CuFe2O4 catalyst are in progress.Ngoc Khanh Nguyen, Ha Minh Tuan and Bui Hoang Yen: synthesis and investigation; Quang Thang Trinh, Bich Ngoc Tran: characterization, data analysis and writing; Van Tuyen Nguyen: supervision and reviewing; Tran Quang Hung, Tuan Thanh Dang and Xuan Hoan Vu: conceptualization, supervision, data analysis, writing, reviewing and 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 is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number 104.01-2017.320. The authors would like to thank Dr. Duc-The Ngo, University of Manchester (UK) for his helpful instruction and discussion in performing the HRTEM measurement and analysis. Supplementary material Image 2 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2020.106240.

Bis(3-indolyl)methanes (BIM) are highly valuable and appear in the core structure of many natural products and pharmacologically active compounds (anticancer, anti-inflammatory, antiobesity, antimetastatic, antimicrobial, etc.). Herein, we have disclosed an air stable and highly efficient CuFe2O4 heterogeneous catalyst for alkylation of indoles with alcohols to give bis(3-indolyl)methanes in very good yields. The CuFe2O4 catalyst has been found to be magnetically recycled at least five times without losing significant catalytic activity.

Hydrogen, as an energy carrier, has potential to provide inexhaustible energy because it has massive reserves on Earth [1–3]. In addition, because hydrogen has a series of outstanding advantages such as environmental friendliness and three times higher calorific value than oil, it is considered to be an attractive alternative to fossil fuels in the field of transportation [4–6]. Major industrial countries have made detailed plans for the wide use of hydrogen fuel cell vehicles. It appears that one day our environmental problems will be solved and we will no longer worry about exhausting our limited fossil fuels. However, the low volumetric energy density of gaseous hydrogen is unacceptable [7]. The main technical hurdle in automotive and portable electronic devices is the design of hydrogen storage systems with high efficiency, non-hazardous and low cost [8–12]. There are three storage states of hydrogen, namely gaseous, liquid and solid [13]. At present, to obtain 4.8 wt.% of hydrogen storage capacity, the needed hydrogen pressure is as high as 70 MPa [14]. There are two disadvantages in the storage of gaseous and liquid hydrogen: short storage time and high evaporation loss [15]. Obviously, both of them have safety issues and technical challenges. Solid hydrogen storage is widely studied for its high energy storage density and safety, and is considered to be the ideal solution to the problem of hydrogen storage and transportation [16–18]. A large number of solid hydrogen storage materials have been developed in recent decades. However, the capacity of the currently developed hydrogen materials is too low and the cycle performance is too poor to satisfy the commercial requirements of vehicle fuel cells. MgH2 is seen as a front-runner in the commercialization of hydrogen storage materials, in particular with coming up with hydrogen absorption and desorption capacities, abundant reserves and non-toxicity [19–22]. Unfortunately, the application prospect is severely hampered because of its slow hydrogen absorb/desorb rate, activation problem and high hydrogen releasing temperature [23,24]. In recent years, the decomposition temperature and kinetic properties of MgH2 have been studied extensively by scientists due to many attractive advantages, and some achievements have been obtained [4,25–31].In terms of preparation techniques, high energy ball milling and introducing additives have been testified to be a beneficial method to dramatically improve the hydrogen storage performance of Mg, which is owing to the reduction of particle size, the increase of active sites and crystal defects [32,33]. In the last decades, lots of literatures have reported the strong improvement effects of introducing additives on the kinetics of MgH2. Common additives include transition metals [34,35] and their oxides [36,37], halides [38,39] and hydrides [40], and rare earth elements [41,42] and their compounds [43,44]. Improvements in hydrogen storage properties are found to be associated with the decreased dissociation energy of hydrogen molecule and the weakened Mg-H bond energy [45]. After introducing additives in MgH2, it can be achieved successfully that very fast hydrogen absorption/desorption kinetics behavior (around minutes or seconds) and a significant reduction on dehydrogenation temperature (around 523 K) [46].Ti, Ni and their compounds often are used as the additives. Research showed that adding the additives can improve the hydrogen storage properties of the alloys. For example, Ti [47], TiH2 [48], TiO2 [49] and TiF3 [50] can improve the hydrogenation kinetics significantly. Wronski et al. [51] found that doping 5 wt.% of nano-nickel in MgH2 can reduce the hydrogen absorption temperature of MgH2. The alloy can be hydrogenated at 523 K. Compared with the original alloy, the reaction temperature dropped by 90 K. Jin et al. [52] reported that NbF5 and TiF3 were better additives. MgH2 reacted with the transition metal fluorides to form corresponding hydrides, which catalyzed the subsequent reaction of MgH2. In addition, fluorine has high electronegativity, which is beneficial to weakening the strength of Mg-H bond.Our previous studies have shown that adding rare earth elements can improve the dynamic properties of magnesium alloys [53,54]. Therefore, in this study, the property of magnesium alloys is expected to be improved by partly replacing Mg in the alloy with Ni, Ce and ball milling with TiF3. At present, there are few relevant studies. In addition, it has been reported that temperature has important effect on the performance of the function materials [55–57]. Therefore, the hydrogen storage performances of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys at different temperature also have been studied.The Ce5Mg85Ni10 alloy was prepared through induction melting with cerium, magnesium and nickel (purity ≥ 99.9%) as raw materials. Metals Ce and Mg will volatilize during melting. Therefore, extra 5 wt.% Ce and 8 wt.% Mg were added to compensate for the loss. The composition of the alloys was examined by an inductively-coupled plasma system (ICP) (Agilent ICPOES730). Before ball milling, the ingots were crushed and ground mechanically to powders of 200–400 mesh. The received powders and 3 wt.% TiF3 were milled together in a planetary ball mill (QM-3SP2 type; produced by Nanjing Chi Shun Technology Development Co., Ltd). The weight ratio of ball to powder was 40:1 and the rotating speed was set at 350 rpm. In the milling process, the mill was interrupted for 30 min after running every 30 min for dissipating heat and reducing clustering of powders. After milling for 5 h, the mill was stopped for scraping the powders adhered to the mill chamber walls and grinding balls (material: stainless steel). A glove box (lab2000 type; produced by Etelux Inert Gas System (Beijing) Co., Ltd) full of a protective atmosphere of argon was used for all these operations. The oxygen and moisture in the glow box were less than 1 ppm.The phase compositions of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10 + 3 wt.% TiF3 (named as Ce5Mg85Ni10+3TiF3) alloys before absorbing hydrogen and after releasing hydrogen were examined by X-ray diffraction (XRD) (D/max/2400). The morphology of the alloy particles was observed by scanning electron microscope (SEM) (QUANTA 400). High-resolution transmission electron microscopy (HRTEM) (JEM-2100) was used to characterize the crystal structure.Isothermal kinetics of the hydrogenation and dehydrogenation of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys were tested through an automatic Sieverts apparatus with ±1 K of the temperature accuracy. The mass of each sample was 300 mg. Before the experiment, the samples were activated four times under the condition of 3 MPa H2 for hydrogenation and 1 × 10−4MPa H2 for dehydrogenation at 633 K. The hydrogenation was conducted at 423, 473, 533, 553, 573, 593, 613 and 633 K at 3 MPa H2, and the dehydrogenation was conducted at 533, 553, 573, 593, 613 and 633 K at 1 × 10−4MPa H2, respectively. In general, the heating rate has an effect on the nonisothermal dehydrogenation performance of the samples. Therefore, with the heating rates of 5, 10, 15 and 20 K min−1, the thermogravimetry (TGA) and differential scanning calorimetry (DSC) (SDTQ600) were used to study the dehydrogenation performances of the samples.The constituents and evolutions of the un-hydrogenated, hydrogenated and dehydrogenated alloys (as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys) are subject to identify by XRD, as illustrated in Fig. 1 . The hydrogenated and dehydrogenated samples are prepared at 633 K in 3 and 1 × 10−4MPa H2 respectively. The incisive and narrow diffraction peaks of the as-cast Ce5Mg85Ni10 alloy show it is a typical crystal structure, in which three phases CeMg12, Mg and Mg2Ni can be detected. After mechanical milling, the diffraction peaks become wider caused by the increase of internal strain and decrease of grain size, suggesting the samples belong to a nanocrystalline structure [58]. The diffraction peaks have significant changes after hydrogenation. Three hydrides, viz. MgH2, Mg2NiH4 and CeH2.73, are viewable in the as-cast and milled Ce5Mg85Ni10 alloys. We have carried on a suitable conjecture of the following path after the XRD analysis: (1) CeMg12+Mg+Mg2Ni+H2→MgH2+Mg2NiH4+CeH2.73 The XRD results show that there are five phases in the TiF3-added alloy after hydrogenation: MgH2, Mg2NiH4, CeH2.73, TiH2 and MgF2. The way to produce these substances can be described as: (2) CeMg12+Mg+Mg2Ni+TiF2+MgF2+H2→MgH2+ Mg2NiH4+CeH2.73+TiH2+MgF2 After dehydriding, three phases appear in the as-cast sample, viz. Mg, Mg2Ni and CeH2.73, which may be formed by the following reaction: (3) MgH2+Mg2NiH4+CeH2.73→Mg+Mg2Ni+CeH2.73+H2 Five phases appear in the as-milled Ce5Mg85Ni10+3TiF3 alloy after dehydriding, viz. Mg, Mg2Ni, CeH2.73, TiH2 and MgF2, which can be described by the following reaction: (4) MgH2+Mg2NiH4+CeH2.73+TiH2+MgF2→Mg+Mg2Ni+CeH2.73+TiH2+MgF2+H2 It is obvious that CeH2.73, TiH2 and MgF2 phases keep unchanged. The reason for this phenomenon is their high thermostability [59]. It allows us to conclude that the reversible reactions occurring in the alloys are Mg+H2↔MgH2 and Mg2Ni+H2↔Mg2NiH4.The microstructure evolutions of the un-hydrogenated, hydrogenated and dehydrogenated alloys (as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys) are examined by HRTEM, as illustrated in Fig. 2 . Nanocrystalline and some crystal defects including grain boundary, phase boundary, lattice distortion zone and dislocation can be seen clearly. The addition of TiF3 significantly changes the microstructure of the alloys. The interplanar spacing of Mg grains is measured using Digital Micrograph software, which is 0.167 nm (the average value of five adjacent atomic planes), as marked in Fig. 2(a). By comparing the data in PDF cards, it can be determined that the adjacent atomic planes belong to (110) crystal plane of Mg phase. The other crystalline phases also are identified using the same method. It is found that there are three phases of CeMg12, Mg and Mg2Ni in the as-cast and milled Ce5Mg85Ni10 alloys, and the addition of TiF3 creates two phases of TiF2 and MgF2. After hydrogenation, the density of crystal defects (especially the lattice distortion region) increases significantly, which is assigned to the enlarged lattice caused by absorbing hydrogen atoms. With the aid of the index of ED rings, three hydrides (MgH2, Mg2NiH4 and CeH2.73) can be easily found in the hydrogenated as-cast and milled Ce5Mg85Ni10 alloys, and a new hydride TiH2 can be found in the hydrogenated Ce5Mg85Ni10+3TiF3 alloy. After dehydrogenation, Mg, Mg2Ni and CeH2.73 phases appear in the as-cast and milled Ce5Mg85Ni10 alloys. The result is consistent with that of XRD. The hydrides CeH2.73 and TiH2 in those alloys remain unchanged during the dehydrogenation process. The CeH2.73 (or TiH2)/Mg (or MgH2) interfaces provide special channels for the rapid diffusion of hydrogen atoms, which is beneficial to improving the kinetics. Besides, the containing-Ti additives can reduce the thermodynamic stability of MgH2, which is mainly due to the fact that the bond energy of Mg-H is able to be weakened under the action of the stronger interaction of Ti-H [60]. It is reported that the dehydrogenation enthalpy of the formed MgH2-0.1TiH2 system is 68 kJ mol−1 H2, which is lower than that of pure MgH2 61.It is generally believed that the surface state of the alloy has great influences on the decomposition of hydrogen molecules [62]. A cracked and coarse surface is deemed to be beneficial in imparting faster hydrogen absorption rate [63]. Fig. 3 shows the morphology of the un-hydrogenated, hydrogenated and dehydrogenated alloys (as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys). The alloy particles produced by mechanical crushing have a very smooth surface and the particle size ranges from 20 to 80 µm. Ball milling makes the particle dimension decreased significantly. The addition of TiF3 has an insignificant effect on the particle size, but it engenders an obvious change in the surface state of the as-milled particles. It can be seen that there are much more cracks on the particles with TiF3 additive and their surfaces become more rough.After absorbing hydrogen, the morphology of the particles has undergone tremendous changes, as shown in Fig. 3(b), (e) and (h). Numerous cracks appear on the surface. They are caused by the increase of lattice stress. As discovered by Antisari et al. [64], the cell volume of magnesium is about 33% smaller than that of MgH2. The hydrogen storage materials will inevitably undergo lattice distortion [65,66], expansion and contraction of the unit volume, thereby introducing lattice stress and lattice defects in the alloys. The accumulation of lattice stress rapidly grows with the increase of hydrogen atoms in the space lattice. Alloy pulverization will occur when it exceeds the maximum value that the lattice can withstand. The morphology of the dehydrogenated alloy particles is similar with that of the un-hydrogenated alloy particles except the increased cracks on the surfaces, which indicates that most of the hydrides in the hydrogenated alloys have been decomposed.Usually, the alloy powders prepared by conventional mechanical crushing and grinding have difficulty in absorbing hydrogen at the beginning stage. This happens because the alloys tend to form an oxide film on the sample surface when they are exposed to air. The oxide film hinders the dissociation process of hydrogen molecules. Fortunately, it has been found that the oxide film can be destroyed when the alloys are subject to the proper temperature and hydrogen pressure. During the process, some fresh surface will appear and the alloys obtain the hydrogen absorption ability.The activation curves of the experimental alloys are demonstrated in Fig. 4 . In order to achieve the complete activation state, the alloys are hydrogenated and dehydrogenated four cycles. It is evident that there is no obvious incubation period for the first cycle under the activation conditions, suggesting that the experimental alloys have good activation property. The first hydriding curves of the experimental alloys are presented in Fig. 4(d). It reveals that ball milling and adding TiF3 have apparent effects on the hydrogenation rate of the alloys in the activation process. The needed time for absorbing 4 wt.% hydrogen is 8944, 6163 and 2852 s for the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, respectively.Research showed that it was difficult for pure Mg to react with hydrogen at low temperature, even with clean Mg surface [67]. It is because that the dissociation energy of hydrogen molecules on magnesium surface is very high. For the as-cast Ce5Mg85Ni10 alloy, the distributed Ni and RE elements have good catalytic activity, which can greatly reduce the dissociation energy mentioned above and promote the hydrogenation reactions [68,69]. For the as-milled alloys, the improved activation ability is definitely ascribed to the modification of surface state caused by ball milling and adding TiF3. The microstructure analysis has proved that ball milling can reduce the particle size, produce reactive clean surfaces and form crystal defects. The reduction of particle size means the increase of the reaction interfaces between hydrogen molecules and the alloy particles. In addition, hydrides are easier to nucleate at the positions of crystal defects. Therefore, the surface activity of the alloys becomes better after ball milling and adding TiF3. The formed MgF2, TiH2, and CeH2.73 nanoparticles strengthen the above effect [70].The dehydriding activation curves of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys are given in Fig. 5 . The curves for the last three cycles almost overlap, suggesting that the alloys have been completely activated after the first hydrogen absorb and desorb cycle. The evolution of the first desorption curves of the experimental alloys is given in Fig. 5(d), from which the time required for desorbing 3 wt.% hydrogen can be obtained. It is 90, 60 and 49 s for the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, respectively. It suggests that ball milling and adding TiF3 can evidently improve the hydrogen desorption kinetics of the alloys because of the following factors. First, TiF3 and TiF2 can promote the decomposition of MgH2. XRD analysis has proved that TiF2 is generated in the process of ball milling the as-cast alloy with TiF3. It should be noted that there are probably a small number of TiF3 in the alloy, which cannot be found in the XRD analysis due to the less content. The roles of TiF3 and TiF2 are similar. Lu et al. [61] also mentioned that the Mg-H bond energy in the Mg-Ti-H system was relatively weak. Song et al. [71] published a theoretical prediction about the reaction enthalpy of MgH2-Ti system. This prediction supported that the above-mentioned thermodynamic change resulted from the weakened Mg-H bond energy. Second, the result in Fig. 3 indicates that hydrogen atoms enter the alloy will cause new cracks, resulting in the increase of specific surface area of the particles and prompting the dehydrogenation kinetics of the alloy. Third, the dispersed MgF2 and TiH2 can provide some sites for the recombination of hydrogen atoms [72].The hydrogen absorption kinetics of the activated as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys are tested at temperatures ranging from 423 to 633 K under 3 MPa H2 pressure, as illustrated in Fig. 6 . It is clear that the hydriding rate is quite fast at the initial stage, and hydrogen absorption capacity can reach over 85% of the saturated capacity in less than 200 s except for 423 K. After that, it takes a long time to reach saturation. The hydrogenation curves at 423 K are compared and the needed time for absorbing 4 wt.% hydrogen is 984, 517 and 228 s at 423 K for the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, respectively. It is found that ball milling and adding TiF3 effectively heighten the hydrogen absorption kinetics of the experimental alloys. The phase structure change during the hydrogenation process (3 MPa H2, 633 K) is analyzed through XRD pattern shown in Fig. 6(d). The result shows that the hydrogenated alloy includes MgH2, Mg2NiH4 and CeH2.73 phases, and the relative content is 59.8%, 24.1% and 16.1%, respectively. The results show that the alloy is to be hydrogen-saturated.There are three steps to generate MgH2, viz. (a) hydrogen molecule decomposes into atoms on the surface of the alloy; (b) hydrogen atoms enter the alloy through defects on the surface of the alloy; (c) Mg converts to MgH2 at the interfaces additive/magnesium. The surface of the alloy is one of the main keys to hydrogen absorption kinetics. The schematic diagram in Fig. 7 shows the mechanism for hydrogenation of the alloy with TiF3. On the one hand, hydrogen molecules are inclined to adsorb on the TiF3 and TiF2 sites with high activity, which reduces the dissociation energy of hydrogen molecule and speeds up the splitting process. On the other hand, adding TiF3 during the period of ball milling makes the grain of the alloy considerably refined and the crystal defect density enormously increased, which has a strong promoting effect on the diffusion of hydrogen atoms. As hydrogen molecules decompose and hydrogen atoms diffuse, the hydrogen concentration quickly reaches the critical concentration required to the formation of metal hydrides. The formed hydride layer hinders the diffusion of hydrogen in the alloy [59]. Hydrogen diffusion in the hydride layer will stop when the thickness of hydride layer is more than 30–50 µm [73]. MgH2 phase only grows slowly through the interfaces of Mg and Mg hydride [74].The TGA and DSC curves of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys are used to study the stability of the generated hydrides, as provided in Fig. 8 . The hydrogenated alloys (at 633 K and 3 MPa H2) are tested in a closed space with the heating rate of 5 K min−1. It is found that ball milling with TiF3 engenders a significant impact on the onset dehydrogenation temperature, which is 544.6, 541.3 and 525.6 K for the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, respectively. The initial dehydrogenation temperature is usually used to show the stability of hydrides. The experimental results reveal that ball milling with TiF3 decreases the initial dehydrogenation temperature of the alloys distinctly. It is found by Berube et al. [75] that ball milling can lower the stabilization of the hydrides, the premise of which was that the alloy particle size was small enough. Whereas, the positive contribution of ball milling to the reduction of the stability of hydride is very limited in this experiment since the particle size is much greater than nanometer scale. As for the positive contribution of adding TiF3 to lessen the steadiness of hydrides, it is believed to be relevant to MgF2 and TiH2 phases that originated from TiF3 phase and Mg phase, which can act as the nucleation centers of dehydrogenation phases. Proper additives can facilitate surface reactions, and some researchers believed that the dehydrogenation reaction was limited by surface reactions [76]. The unsaturated d/f electron shells of transition metals can react with valance electrons of H so as to weaken the Mg-H bond energy [77].The hydrogen desorption kinetics of the hydrogenated as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys are tested at temperatures ranging from 533 to 633 K under 1 × 10−4MPa H2, as illustrated in Fig. 9 . The experimental results show that the alloy can achieve complete dehydrogenation state even at 533 K and has good kinetic property. The reasons have been given in the section of activation behaviors (see the description and explanation about Fig. 5). Meanwhile, the experimental results show that temperature has a great influence on the kinetics of hydrogen desorption, and higher temperature is conducive to the process of desorbing hydrogen. The time required for desorbing 3 wt.% hydrogen at 533, 553, 573, 593, 613 and 633 K is 1410, 390, 185, 108, 72 and 48 s for the as-cast Ce5Mg85Ni10 alloy, and 906, 324, 180, 102, 66 and 42 s for the as-milled Ce5Mg85Ni10 alloy and 724, 286, 161, 89, 58 and 36 s for the as-milled Ce5Mg85Ni10+3TiF3 alloy, respectively. The results show that ball milling and the addition of TiF3 can accelerate the hydrogen desorption kinetics of the alloy at the same temperature, but as the temperature increases, this positive effect weakens gradually. The phase structure change during the dehydrogenation process (1 × 10−4MPa H2 and 633 K) of the dehydrogenated as-cast Ce5Mg85Ni10 alloy is analyzed through XRD pattern showed in Fig. 9(d). It can be seen that there are Mg, Mg2Ni and CeH2.73 phases in the dehydrogenation alloy. The mass percentage of each phase is 59.2%, 25.1% and 15.7%, respectively. Obviously, the rare earth hydride CeH2.73 cannot decompose under the experimental temperature and pressure.The dehydrogenation of MgH2 can be divided into three steps as well: (a) Mg nucleates and grows at the defect sites; (b) hydrogen atoms diffuse from the decomposed MgH2 to the particle surface; (c) the combination of two adjacent hydrogen atoms into one hydrogen molecule [78]. As mentioned above, the main reason for the choice of ball milling method is that ball milling can reduce the size of alloy particles, increase specific surface area and form micro nanostructure and many defects. Adding TiF3 can weaken the Mg-H bond energy, thus enhancing the dehydrogenation behavior of MgH2 [79]. The schematic diagram of the action mechanism of TiF3 for dehydrogenation is described in Fig. 10 . Firstly, the electronegativity of Ti is 1.8, which is between that of H (2.2) and Mg (1.3). Therefore, adding TiF3 can weaken the Mg-H bond energy sufficiently. Ti ions gain electrons more easily than Mg2+. Compared with H−1, Ti ions are more likely to lose electrons. Secondly, Ti ions with different valence states are prone to transformation. Hence, the effect of TiF3 on hydrogen desorption of the metal is as follows: (a) The H− on the Mg-H bond passes one electron to Ti3+, and H− loses one electron and goes to the free state H0, then Ti3+ gets one electron and turns into low valency Ti2+. (b) SinceTi ions in different valence states are easy to be transformed, Ti2+ can spontaneously lose one electron and transform into Ti3+. At the same time, Mg2+ obtains two electrons and form Mg; (c) free state H0 diffuses to the surface of the alloy and forms H2 [80]. This catalytic electron transfer process is much easier than the direct transfer of electrons from H− to Mg2+. Therefore, the activation energy of dehydrogenation is diminished.To verify the above conclusions, apparent activation energies of the dehydrogenation reactions of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys are evaluated by the Arrhenius and Kissinger method. Apparent activation energy is usually used to represent the energy hurdle that needs to be overcome in a gas-solid reaction. Therefore, the apparent activation energy can be used to represent the minimum level of the reactions that occur in the system. The main driving mode of dehydrogenation of Mg-based alloys is nucleation and growth [42]. Generally, these solid-state reactions can be simulated by the Johnson-Mehl-Avrami-Kolmogorov (JMAK) theory [81] through the following equation: (5) l n [ − l n ( 1 − α ) ] = h l n k + h l n t In this equation, the proportion of MgH2 transformed into Mg in fixed time t is represented by α, k and η are the kinetic parameter and Avrami exponent respectively. Referring to the data in Fig. 9, the relationships between ln [-ln(1-α)] and lnt at 573, 593, 613 and 633 K are plotted in Fig. 11 . The apparent activation energy E a can be derived from the following equation 42: (6) k = A e x p [ − E a / ( R T ) ] The above formula is the Arrhenius equation, in which A, R and T respectively represent temperature-independent coefficient, gas constant and absolute temperature of the reaction, and k has the same definition as above. The obtained apparent activation energy (E a) of the experimental alloys is provided in Fig. 11(d). It is obvious that ball milling and adding TiF3 markedly decrease the apparent activation energy E a of the alloys. Data analysis shows that E a of MgH2 is much higher than that of the experimental samples, because the addition of Ni in the sample improves the catalytic activity of the alloy surface. Therefore, the in-situ formed Mg2Ni/Mg2NiH4 has a good catalytic effect on Mg-based alloys [82]. There are also sources that Ce has a good effect on MgH2 dehydrogenation. Mustafa [83] reported that the reduction of E a significantly promotes the decomposition of MgH2. Hou et al. [84] found that adding a suitable catalyst is an effective strategy to reduce the E a of MgH2.The DSC curves at different heating rates are also tested for comparison with the JMAK model. The DSC curves are shown in Fig. 12 and the apparent activation energy (E k) of hydride decomposition is determined by Kissinger equation [85]: (7) d [ l n ( β / T p / T p ) ] / d ( 1 / T p ) = − E k / R In Eq. (7), β and T p are heating rate and absolute temperature corresponding to the peak of DSC curves, respectively. R has the same definition as Eq. (6). The graphs of ln(β/T p/T p) vs. 1/T p can be charted through Eq. (7), as illustrated in Fig. 12. E k values of the alloys can be obtained by the slope of the Kissinger diagram, as presented in Fig. 12(d), which is very close to the results in Fig. 11(d). The above results allow us to believe that the heightened dehydrogenation kinetics caused by ball milling and adding TiF3 is assigned to the decrease in dehydrogenation activation energy.The thermodynamic parameters of the hydrides are important indicators to assess their stability. The main problem of practical application is the high thermal stability of MgH2. Here, P-C-T curves of the experimental alloys are gauged at 573, 593, 613 and 633 K in order to inspect the thermodynamics of the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, as presented in Fig. 13 . Each P-C-T curve has two platforms. The high platform corresponds to the formation and decomposition of Mg2NiH4, and the low platform corresponds to the formation and decomposition of MgH2. Moreover, it is found that the hysteresis coefficient (H f=ln(P a/P d)) of the plateau pressures of Mg/MgH2 is very small, while that of Mg2Ni/Mg2NiH4 is quite large, which may be due to the lattice stress caused by hydrogen atoms enter and overflow the alloys [86]. The pressure values corresponding to the longer and lower plateaus are adopted to calculate the hydrogenation and dehydrogenation thermodynamic parameters of Mg/MgH2 in the alloys. For the Mg2Ni/Mg2NiH4 platform, the thermodynamic parameters are not calculated because the equilibrium pressure is too short. The enthalpy change (ΔH) and entropy change (ΔS) of the formation and decomposition of MgH2 can be calculated by Van't Hoff equation. The calculation is based on the pressure value at the center point of the platform at equilibrium [87]: (8) l n ( P H / P 0 ) = Δ H / R T − − Δ S / R In Eq. (8), P H and P 0 are hydrogen equilibrium pressure and standard atmospheric pressure respectively. An explanation of R and T has been given before. By fitting the linear relationship between ln(P H/P 0) and 1/T, the Van't Hoff diagram of Mg/MgH2 can be obtained as shown in Fig. 13. Therefore, the thermodynamic parameters can be calculated by the slope and intercept of the Van't Hoff diagram, as presented in Fig. 13(d). It is evident that ball milling and adding TiF3 result in a slight decrease of the absolute values of the hydrogenation enthalpy change (ΔH ab) and dehydrogenation enthalpy change (ΔH de). The ΔH de value is 77.24, 75.28 and 75.16 kJ mol−1 H2 for the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys, respectively. The change in microstructure caused by introducing TiF3 during ball milling has a positive effect on the reduction of enthalpy change. As has been reported in references [77,88], sample particles of magnesium-based alloys can be reduced to the nanoscale by ball milling, and its thermal stability will be greatly reduced. Since the size of the as-milled particles in this experiment is much bigger than nanometer scale, the improvement of thermodynamic properties of the experimental alloys caused by ball milling is very limited. Agarwal et al. [89] considered that the small change of enthalpy may be related to the defects in the alloys. In addition, the added TiF3 and the reaction products (TiF2, MgF2 and TiH2) are beneficial to promoting the reduction of Mg-H bond energy and the decomposition of MgH2.The thermodynamics and kinetics of storing hydrogen in the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10+3TiF3 alloys have been studied in detail in this paper and some conclusions can be reached: (1) The as-cast and milled Ce5Mg85Ni10 alloys are composed of CeMg12, Mg and Mg2Ni phases. The addition of TiF3 creates new phases of MgF2 and TiF2. The as-milled alloys exhibit a nanocrystalline structure. Ball milling engenders an obvious refinement of the grains and increment of the density of crystal defects. The addition of TiF3 makes the surface of the milled particles rough and irregular. (2) The experimental as-cast alloy exhibits a good activation property, and ball milling and adding TiF3 further facilitate the activation property. The crystal lattice expansion and shrinkage caused by hydrogen entry and release during the activation process results in the severe pulverization of the alloy particles along with the generation of numerous new surfaces and crystal defects, which improve the hydrogen storage kinetics of the alloys. (3) Ball milling and addition of TiF3 has slightly beneficial effects on the thermodynamics of the experimental samples, which is most likely assigned to the reduction of the particle size and the generation of numerous crystal defects and reaction products (TiF2, MgF2 and TiH2). The as-cast and milled Ce5Mg85Ni10 alloys are composed of CeMg12, Mg and Mg2Ni phases. The addition of TiF3 creates new phases of MgF2 and TiF2. The as-milled alloys exhibit a nanocrystalline structure. Ball milling engenders an obvious refinement of the grains and increment of the density of crystal defects. The addition of TiF3 makes the surface of the milled particles rough and irregular.The experimental as-cast alloy exhibits a good activation property, and ball milling and adding TiF3 further facilitate the activation property. The crystal lattice expansion and shrinkage caused by hydrogen entry and release during the activation process results in the severe pulverization of the alloy particles along with the generation of numerous new surfaces and crystal defects, which improve the hydrogen storage kinetics of the alloys.Ball milling and addition of TiF3 has slightly beneficial effects on the thermodynamics of the experimental samples, which is most likely assigned to the reduction of the particle size and the generation of numerous crystal defects and reaction products (TiF2, MgF2 and TiH2).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.It is sincere thanks to the National Natural Science Foundation of China (Nos. 51871125, 51761032, 52001005 and 51731002) and Major Science and Technology Innovation Projects in Shandong Province (No. 2019JZZY010320) for financial support of the work.

Mg-based hydrides are too stable and the kinetics of hydrogen absorption and desorption is not satisfactory. An efficient way to improve these shortcomings is to employ reactive ball milling to synthesize the nanocomposite materials of Mg and additives. In this experiment, TiF3 was selected as an additive, and the mechanical milling method was employed to prepare the experimental alloys. The alloys used in this experiment were the as-cast Ce5Mg85Ni10, as-milled Ce5Mg85Ni10 and Ce5Mg85Ni10 + 3 wt.% TiF3. The phase transformation, structural evolution, isothermal and non-isothermal hydrogenation and dehydrogenation performances of the alloys were inspected by XRD, SEM, TEM, Sievert apparatus, DSC and TGA. It revealed that nanocrystalline appeared in the as-milled samples. Compared with the as-cast alloy, ball milling made the particle dimension and grain size decrease dramatically and the defect density increase significantly. The addition of TiF3 made the surface of ball milling alloy particles markedly coarser and more irregular. Ball milling and adding TiF3 distinctly improved the activation and kinetics of the alloys. Moreover, ball milling along with TiF3 can decrease the onset dehydrogenation temperature of Mg-based hydrides and slightly ameliorate their thermodynamics.

The depletion of fossil resources together with a strong drive to limit greenhouse gas emissions has led to an increasing effort in the development of sustainable and green transportation fuels. Well known examples are ethanol from sugars using fermentative approaches [1] and biodiesel from vegetable oils [2], which have both been commercialized in the last decades. When considering ethanol, some disadvantages have been identified, including a low energy density, high vapor pressure and high water solubility, which cause corrosion issues when using ethanol-rich ethanol-gasoline blends [3]. These disadvantages may be alleviated by using C3+ alcohols, which have superior fuel properties, such as higher energy density, lower volatility and better solubility in hydrocarbons (HC), while at the same time possessing comparable octane numbers as found for gasoline [4].When considering chemo-catalytic routes to higher alcohols, syngas appears an interesting feed [5]. Various catalytic systems have been identified for this purpose [6]. Among them, molybdenum sulfide-based catalysts are of particular interest due to their low cost, high water-gas shift activity and high resistance to sulfur poisoning [7], thus avoiding the need for water separation and deep desulfurization units. MoS2 alone mainly produces CO2 and hydrocarbons (HC) from syngas, while alkali metals, especially potassium (K) modified MoS2 catalysts are commonly used to achieve good selectivity for alcohols [8]. K promotion suppresses hydrogenation of metal-alkyl species to HCs and enhances the rate of CO insertion in the M-alkyl bond to form metal-acyl species, which are subsequently converted to alcohols [9]. It is proposed that KMoS2 phases, formed by the intercalation of K into the MoS2 structure, are responsible for the higher selectivity to alcohols when compared to MoS2 alone [10–13].However, K modified MoS2 catalysts normally suffer from low activity [6], leading to relatively low CO conversion and thus a low yield of alcohols. Efforts have been undertaken on tailoring the structure of the K modified MoS2 catalysts to enhance the selectivity to C3+ alcohols [14,15]. In previous work from our groups, we prepared multilayer K modified MoS2 catalysts with well-contacted MoS2 and KMoS2 phases and showed that these catalysts lead to improved alcohol selectivities [16]. Another approach involves promotion by group VIII metals, such as Co and Ni [7,17–19]. Especially cobalt is known to promote carbon chain growth, leading to higher selectivities to higher alcohols [20,21], though often ethanol is the major product.Co promoted MoS2 catalysts are widely used in hydrodesulfurization (HDS) reactions and the promoting effect of Co is attributed to the formation of a Co-Mo-S phase [22], formed by partial substitution of Mo atoms at the edge of MoS2 slabs by Co atoms [23]. This particular phase has also been observed in K modified, Co promoted MoS2 catalysts for alcohol synthesis [18,20,24–27]. To elucidate the function of cobalt, Mo free, K modified cobalt sulfide catalysts were employed for the reaction. In this case, the amount of higher alcohols was low and C1-C4 alkanes were prevailing [20], indicating that K-CoSx phases are not suitable for higher alcohol synthesis. It also has been shown that, the number of active Co-Mo-S species decreases at high Co loadings due to the formation of Co9S8 phases, which are stable under typical reaction conditions and have a low activity for higher alcohols [28–31].Thus, literature data imply that a Co-Mo-S phase in Co promoted MoS2 catalysts is the active phase, [20,28–33], though the exact mechanism to promote carbon chain growth is still under debate. However, the role of both K and Co in K modified CoMoSx catalysts has not been explored in detail. We therefore performed a systematic investigation on the effect of these promotors on the performance of MoS2 catalysts for higher alcohol synthesis from syngas. For this purpose, a series of K modified Co promoted molybdenum sulfide catalysts with different Co contents and a fixed K content were prepared, characterized in detail and tested for the conversion of syngas to higher alcohols. The results were compared with a Mo free catalyst in the form of K-CoSx and a K-free catalyst (CoMox-0.13). In addition, for the optimized catalyst regarding Co content, the effect of process conditions, such as temperature (T), pressure (P), gas hourly space velocity (GHSV) and H2/CO ratio was explored. The results were quantified using statistical approaches allowing determination of the optimal process conditions for higher alcohol selectivity and yield.The cobalt-molybdenum sulfide was prepared by sulfurization of the cobalt-molybdenum oxide precursor with KSCN according to a method reported in the literature [34] with some modifications. The cobalt-molybdenum oxide precursor was typically synthesized by mixing Co(NO3)2·6H2O and (NH4)6Mo7O24·4H2O (20 g in total, Sigma-Aldrich) in 50 mL of deionized water. The resulting suspension was heated and maintained at 120 °C for 3 h, during which most of the water evaporated. The resulting mixture was calcined in air at 500 °C for 3 h to form the cobalt-molybdenum oxide. The amount of Co(NO3)2·6H2O and (NH4)6Mo7O24·4H2O was varied to adjust the atomic ratio Co/(Co + Mo) between 0 and 0.7.For sulfurization, the cobalt-molybdenum oxide (0.648 g), KSCN (0.875 g, Sigma-Aldrich), and deionized water (35 mL) were mixed in an autoclave, which was kept at 200 °C for 24 h. Then the autoclave was rapidly cooled with ice, and the resulting precipitate was filtered and washed with deionized water (total 500 mL). The product was obtained after drying at ambient conditions overnight. The molybdenum sulfide is labelled as MoSx and the mixed metal sulfide catalysts are labelled as Co-MoSx-R, where R represents the actual Co/(Co + Mo) ratio as obtained from ICP-OES. The elemental composition of the sulfurized catalysts is shown in Table 1 .The K promoted K-Co-MoSx-0.13 catalyst, used for detailed analyses by XRD, HRTEM and STEM with EDS mapping, was prepared by physically mixing Co-MoSx-0.13 with K2CO3 followed by a treatment under hydrogen (1 bar, 8 h, 400 °C) and subsequent passivation (1% O2/N2, 4 h, 25 °C).The K promoted CoS2 catalyst was prepared by physically mixing a CoS2 sample (Sigma Aldrich) with K2CO3 followed by a reduction procedure as described above.The cobalt-molybdenum sulfide samples were characterized with ICP-OES (Spectroblue, Germany) to quantify the elemental composition.The specific surface area and pore parameter were determined using N2 physisorption, which was conducted at 77 K using an ASAP 2420 system (Micromeritics, USA). Prior to analysis, the samples were degassed at 150 °C under vacuum for 12 h. The specific surface area was calculated using the Brunauer-Emmett-Teller (BET) method in the P/P0 range of 0.05–0.25. The total pore volume was estimated from the single point desorption data at P/Po = 0.97. The pore diameter was obtained from the desorption branch according to the Barrett-Joyner-Halenda (BJH) method.X-ray diffraction (XRD) patterns of the sulfurized samples were collected for a 2θ scan range of 5–80° on a D8 Advance powder diffractometer (Bruker, Germany) with CuKα radiation (λ = 1.5418 Å) operated at 40 kV and 40 mA. XRD spectra of the K modified sample (K-Co-MoSx-0.13) were recorded in the same way.H2-TPR measurements were conducted using 10 vol.% H2 in He (30 ml min-1) and the samples were heated from room temperature to 900 °C at a temperature ramp of 10 °C/min using an AutoChem system (Micromeritics, USA) equipped with a thermal conductivity detector (TCD). Raman spectroscopy was measured using a WITec Alpha 300R microscope with a 532 nm excitation laser.The micro-structure of the sulfurized samples was examined with high-resolution transmission electron microscopy (HRTEM, JEOL 2010 FEG, Japan) operating at 200 kV. The samples were first ultrasonically dispersed in ethanol and then deposited on a carbon-coated copper grid. Processing of the HRTEM images was accomplished using DigitalMicrograph software.High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of the K-Co-MoSx-0.13 sample were obtained using a probe and image aberration corrected Themis Z microscope (Thermo Fisher Scientific) operating at 300 kV in STEM mode with a convergence semi-angle of 21 mrad and a probe current of 50 pA. Energy dispersive X-ray spectroscopy (EDS mapping) results were achieved with a Dual X EDS system (Bruker) with a probe current of 250 pA. Data acquisition and analysis were done using Velox software (version 2.8.0).Reactions were performed in a continuous fixed-bed reactor (stainless steel) with an internal diameter of 10 mm. Typically, the cobalt-molybdenum sulfide catalyst (0.35 g) was physically mixed with K2CO3 (0.05 g, Sigma-Aldrich) and SiC (2.0 g, Sigma-Aldrich) and then loaded to the reactor. Before reaction, the catalyst was reduced in situ using a flow of H2 (50 ml min-1) at 400 °C for 8 h. After cooling to room temperature under a N2 stream, the reaction was started by switching to a gas mixture of H2/CO (molar ratio ranging from 1.0 to 2.0) with 6% N2 (internal standard). Typical reaction conditions are pressures between 8.7 and 14.7 MPa and temperatures between 340 and 380 °C. The gas hourly space velocity (GHSV) was varied from 4500 to 27000 mL g-1  h-1 by adjusting the flow rate of the feed gas. The reactor effluent was cooled and the liquid product was separated from the gas phase by using a double walled condenser at −5 °C. Details regarding product analysis are described in a previous publication from our groups [16]. The CO conversion (XCO), the product selectivity (Si) and yield (Yi) were calculated using Eqs. (1)–(3). (1) X C O = m o l e s   o f   C O i n f l u e n t - m o l e s   o f   C O e f f l u e n t m o l e s   o f   C O i n f l u e n t × 100 % (2) S i = m o l e s   o f   p r o d u c t   i × n u m b e r   o f   c a r b o n s   i n   p r o d u c t   i m o l e s   o f   C O i n f l u e n t - m o l e s   o f   C O e x f l u e n t × 100 % (3) Y = X C O × S i The activity data given in this study are the average for at least 6 h runtime and collected after 20 h, to ensure stable operation of the reactor. The selectivity of all products is carbon based and only data with carbon balances higher than 95% are reported here.The chain growth probability α was determined from the experimental data assuming an ASF distribution for the alcohols (Eq. (4)). (4) S n n = α n × ( 1 - α ) a Here, S n is the selectivity of the alcohols with a carbon number of n , n is the carbon number, and α is the chain growth probability. The value of α was determined by plotting l n ( S n n ) against n.Multivariable regression was used to quantify the effect of process conditions (T, P, GHSV and H2/CO ratio) on catalytic performance (Eq. (5)). (5) Y = a 0 + ∑ a i x i + ∑ a i i x i 2 + ∑ a i j x i x j Here x is independent variable (T, P, GHSV and H2/CO ratio) and Y is a dependent variable (selectivity and yield of C3+ alcohol), ai, aii, and aij are the regression coefficients and a0 is the intercept. The regression coefficients were determined using the Design-Expert (Version 7) software by backward elimination of statistically non-significant parameters. The significant factors were selected based on their p-value in the analysis of variance (ANOVA). A parameter with a p-value less than 0.05 is considered significant and is included in the response model.The cobalt-molybdenum sulfide catalysts with different Co contents were prepared by sulfurization of the corresponding cobalt-molybdenum oxide precursors using KSCN. The actual Co/(Co + Mo) molar ratio was determined by ICP-OES and ranged from 0 to 0.63 (Table 1). The textural properties of the sulfurized catalysts (without K addition) are depending on the Co content, see Table 1 for details. When considering the specific surface area, a maximum was found for Co-MoSx-0.13, with a value of 11.5 m2 g-1. This value is in the broad range reported in the literature for Co-MoSx catalysts (from single digit values to several hundred square meters per gram [35]), rationalized by differences in the Co and Mo precursors used and synthesis conditions. The observed reduction at higher Co amounts may be due to the formation of a segregated Co sulfide phase [36]. Similar trends were observed for the pore volume and pore diameters of the catalysts, viz. the highest value was found for catalyst Co-MoSx-0.13.The XRD patterns of the catalyst (without K addition) are shown in Fig. 1 . The MoSx catalyst shows broad diffractions at 2θ values of about 14°, 33°, 36° and 58°, which are associated with the (0 0 2), (1 0 0), (1 0 2) and (1 1 0) planes, respectively, of the 2H-MoS2 phase (JCPDS card No. 00-037-1492). Upon the addition of Co, the reflexes of the crystalline MoS2 phase disappear and new signals arise. These were identified as cobalt-containing species like CoS2 (JCPDS card No. 01-089-3056), CoMoS3.13 (JCPDS card No. 00-016-0439) and CoMoO4 (JCPDS card No. 00-021-0868). Of interest is the presence the CoMoS3.13 phase, which is known to be formed by partial substitution of Mo atoms at the edges of MoS2 sheets by Co. Mixed Co-Mo-S phases are generally thought to be active for higher alcohol synthesis by promoting carbon chain growth [6]. At high Co loadings, sharp reflexes from crystalline CoS2 and CoMoO4 are present, suggesting a higher abundance and larger nanoparticle sizes. Reflexes attributed to a Co9S8 phase, reported to be present at higher Co loadings, were not detected [30].H2-TPR measurements were performed for all sulfided Co-Mo catalysts and the profiles are given in Fig. 2 . The Co free MoSx catalyst displays two H2 peaks, a small one at 310 °C and a larger one at about 720 °C. The first peak is ascribed either to the presence of over-stoichiometric Sx species or to weakly bonded sulfur anions along MoS2 edges [37]. The high temperature peak is associated with more strongly bound sulfur anions located at the edges [38]. Another possibility is a phase formed by desulfurization of the MoS2 phase by elimination of basal sulfur, though not likely as temperatures higher than 830–1030 °C are required for this transition [39]. Upon the addition of Co, additional peaks become visible. The low temperature peak is shifted to lower temperatures (about 220 °C), indicating that the presence of Co leads to a weakening of the Mo-S bond [40]. A similar low temperature peak was also observed during H2-TPR measurements on supported Co-MoS2/Al2O3 catalysts for HDS reactions and associated with the presence of a Co-Mo-S phase [41]. The area of the first peak is reduced when adding more Co in the catalyst formulation. Besides, a new peak at an intermediate temperature (370–470 °C) appears, which is ascribed to a cobalt sulfide phase [41]. In line with this explanation is the observation that the area of this particular peak increases with increasing Co content. This suggests that for low Co/(Co + Mo) ratios, the Co atoms are dispersed at the edge of a MoS2 phase to form a Co-Mo-S phase, whereas higher Co amounts lead to the formation of Co sulfide species. These may be present as a single phase or closely interact with Co-Mo-S and MoS2 phases.The Raman spectra of the sulfided Co-Mo catalysts (without K) are shown in Fig. 3 . The unpromoted MoSx catalyst exhibits two peaks at 380 cm-1 and 405 cm-1, which are ascribed to the in-plane E1 2g and out-of-plane A1g vibration mode of the MoS2 layer structure [42]. These two bands are also detected in Co-MoSx-0.13, and the distance between the two bands, which is an indicator for the interlayer distance between the MoS2 stacked layers [15,43], is similar to that for the unpromoted MoSx catalyst. This suggests that, different with K [12], Co is not intercalated in the interlayer space of MoS2 phase, which is consistent with the H2-TPR result. For the catalysts with high Co contents, the two peaks disappear, and a new peak at 931 cm-1 emerges, associated with the formation of a β-CoMoO4 phase, which is in consistent with the XRD results. The intensity of the peak increases with increasing Co content.HRTEM was used to determine the morphology and microstructure of the catalysts. Representative images are displayed in Fig. 4 . The MoSx catalyst without Co shows a multilayer structure with a lattice spacing of 0.63 nm, corresponding to the (0 0 2) plane of the MoS2 phase (Fig. 4a) [44]. After the addition of Co, various Co-containing species were identified based on their specific lattice fringes. Examples are Co-MoSx, CoSx and CoMoO4 phases (Fig. 4b–f). The lattice fringe with a lattice spacing of 0.25 nm corresponds to the (2 1 0) plane of CoS2.Of interest is the observation of close contacts between the CoS2 and MoS2 phase for Co-MoSx-0.13 (Fig. 4b–c), indicating the presence of a CoS2/MoS2 interface. The presence of this interface has been reported to be beneficial for higher alcohol formation [45]. The phase with a lattice spacing of 0.63 nm may be either from MoSx or a CoMoS3.13 species. For catalysts with a higher Co content (e.g. Co-MoSx-0.37), a CoMoO4 phase is present (lattice fringe with a spacing of 0.68 nm (Fig. 4d)), consistent with the XRD analysis.With the catalyst characterization data available, the effect of the amount of Co on catalyst structure may be assessed. Unpromoted MoSx reveals a multilayer structure with long-range ordered MoS2 domains, in line with the literature data. After promotion with Co, Co-Mo-S and CoS2 phases are formed, which are considered possible active phases for higher alcohol synthesis (Co-MoSx-0.13). At higher Co contents, higher amounts of CoS2 and CoMoO4 species are present, which may have a negative effect on catalyst performance (vide infra).Finally, the K promoted version of Co-MoSx-0.13 (K-Co-MoSx-0.13), which is the best catalyst in terms of performance for higher alcohol synthesis (vide infra), was characterized in detail using XRD, HRTEM and STEM with EDS mapping to gain insights in changes in the structure upon the addition of K. The sample was prepared by physically mixing Co-MoSx-0.13 with K2CO3 followed by reduction with hydrogen and passivation (see experimental section).XRD spectra of K-Co-MoSx-0.13, together with MoSx and Co-MoSx-0.13 for comparison, are given in Fig. 5 a. The (002) reflex of K-Co-MoSx-0.13 at 13.3° is slightly shifted downfield compared to that of MoS2 (14.1°), indicating an expanded interlayer spacing due to the incorporation of K. A HRTEM image (Fig. 5b) of K-Co-MoSx-0.13 confirms the expanded interlayer spacing (0.77- 0.81 nm vs 0.63 nm for Co-MoSx-0.13, Fig. 4c) after K addition. The intercalation of K into the MoS2 structure leads to the formation of a KMoS2 phase, which was discussed in detail in our previous work [16] and is suggested to be essential for alcohol synthesis.The reflexes of CoS2, clearly visible in Co-MoSx-0.13, are absent in the XRD spectrum of K-Co-MoSx-0.13. New reflexes at 30.1°, 31.2° and 39.7°, identified as Co9S8 species (JCPDS card No. 00-003-0631) are present. The Co9S8 species are likely formed by reduction of CoS2, which is consistent with the H2-TPR results (Fig. 2). Representative reflexes of crystalline CoMoS3.13 are also present in K-Co-MoSx-0.13. The presence of both Co9S8 and CoMoS3.13 species in K-Co-MoSx-0.13 is confirmed by HRTEM images (Fig. 5c–d). Close contacts between the Co9S8 and K promoted (Co)MoSx phase were observed (Fig. 5b–d), in agreement with the observation of CoS2/(Co)MoSx interfaces in the unpromoted Co-MoSx-0.13 catalyst (Fig. 4b–c).A STEM dark field image combined with EDS mapping (Fig. S1) of K-Co-MoSx-0.13 shows that K, Co, Mo and S are uniformly dispersed in the catalyst. Such a homogeneous distribution is indicative for the presence of abundant Co9S8/K-(Co)MoSx interfaces in K-Co-MoSx-0.13.Benchmark experiments with all catalysts were performed at 360 °C, 8.7 MPa, a GHSV of 4500 mL g-1 h-1 and a H2/CO ratio of 1 in a continuous packed bed reactor set-up. These conditions were selected based on previous experience in our group on the use of MoS2 catalysts for higher alcohol synthesis [16]. Prior to reaction, the catalysts were promoted with K using a physical mixing method followed by an in situ treatment with H2. The same amount of K was used for all catalyst formulations. The experiments were performed for at least 6 h and the performance of the catalyst was the average over the time period from 20 h to final runtime and thus taken at steady state conditions in the reactor (Table 2 ).A typical example of the product selectivity and CO conversion versus the runtime is given in Fig. 6 (340 °C, 11.7 MPa, GHSV of 4500 mL g-1 h-1 and H2/CO ratio of 1.5 using the K-Co-MoSx-0.13 catalyst). It also shows the catalyst is stable for at least 100 h without co-feeding of sulfur.Typical reactions products are alcohols (methanol, ethanol, and C3+ alcohol), hydrocarbons (methane and higher ones) and CO2. The latter is formed by the water-gas shift reaction involving CO and water. The unpromoted K-MoSx catalyst provides a selectivity of 40.8% to alcohols and 24.8% to hydrocarbons at a CO conversion level of 25.6% (Fig. 7 ), which is typical for Mo-based catalysts [6]. Upon the addition of Co to the catalyst formulation, the CO conversion decreases, which may be due to the reduced availability of the active sulfided Mo-Co species by coverage with inactive CoMoO4 species and/or the presence of less active CoS2 species, as observed from XRD and HRTEM results.The selectivity is a clear function of the Co content. Alcohol selectivity reaches a maximum (47.1%) for the K-Co-MoSx-0.13 catalyst and decreases with higher Co loadings, see Fig. 7 for details. The selectivity to hydrocarbons (mainly CH4), shows a reverse trend, whereas the CO2 selectivity is about constant. The product selectivity at two other temperatures (340 and 380 °C) also shows a similar trend regarding the Co content in the catalyst formulation (Table S2).The effect of Co addition on the carbon distribution of the alcohols is given in Fig. 8 a. It shows that the amount of C3+ alcohols reaches a maximum at 59.0% for the K-Co-MoSx-0.13 catalyst and decreases at higher Co amounts. The individual distribution of alcohols for the unpromoted K-MoSx, K-Co-MoSx-0.13 and K-Co-MoSx-0.63 catalyst are depicted in Fig. 8b (the distributions for other catalysts are shown in Fig. S2) as Anderson-Schulz-Flory (ASF) plots. The unpromoted K-MoSx catalyst shows a large deviation for particularly methanol when considering an ideal linear ASF distribution. This is in line with previous findings of our group, rationalized by assuming an enhanced chain growth mechanism for C3+ alcohol using these types of catalysts [16]. After loading with Co, an even larger deviation for methanol and also for ethanol is observed for the K-Co-MoSx-0.13 catalyst. However, the deviation is less pronounced when further increasing the Co content (Fig. S2) and the K-Co-MoSx-0.63 catalyst shows an almost perfect linear distribution for the mixed alcohols including methanol. The carbon chain growth probability was calculated for the C2+ alcohols, showing a volcano-shaped curve with a peak for the K-Co-MoSx-0.13 catalyst (Fig. S3). Thus, alcohol selectivity and carbon chain growth are best for the K-Co-MoSx-0.13 catalyst, whereas higher Co contents lead to a higher hydrocarbon selectivity and a lower carbon chain growth for the alcohols.For comparison, and also to determine the role of Mo in the catalyst formulation, the catalytic performance of a K promoted CoS2 catalyst was also investigated. We first attempted to prepare the CoS2 catalyst by a similar procedure as used for the Co-MoSx samples (viz. sulfurization of the cobalt-oxide precursors using KSCN). However, Co3O4 instead of CoS2 was obtained (Fig. S4), indicating that Co-oxides are difficult to sulfurize using KSCN at the prevailing conditions. Therefore, CoS2 (Sigma-Aldrich) was used as the catalyst precursor, and after K addition and pretreatment (in situ reduction with H2 at 400 °C for 8 h) tested for higher alcohol synthesis (360 °C, 8.7 MPa, GHSV of 4500 mL g-1 h-1 and H2/CO molar ratio of 1). A very high hydrocarbon selectivity of 63.1% was achieved at a CO conversion of 1.3% (Table S3). Higher alcohols could not be detected in the liquid phase. The low CO conversion might be due to the presence of large crystallites (76 nm, from XRD data using Scherrer equation) and the lack of structural defects (Fig. S5). These findings are in line with experiments by Li et al., who reported that only C1-C4 alkanes and no alcohols were formed when using a K-CoSx on activated carbon catalyst (in which Co is present in the form of Co9S8 crystallites) [20]. Co9S8 species, formed by reduction of CoS2 were indeed detected after reaction (Fig. S5), in line with literature data [20].The unpromoted Co-MoSx-0.13 catalyst (without K) showed high CO conversion and very low selectivity for alcohols (< 2%) in comparison with that of K-Co-MoSx-0.13 (Table S4), indicating the important role of K for alcohol synthesis. Specifically, the presence of a KMoS2 phase (Fig. 5) is considered to be essential for alcohol synthesis, see also previous work from our group [16]. This is also in agreement with literature data revealing that the addition of K in MoS2 catalysts leads to lower hydrogenation rates while maintaining good CO insertion rates [8,9,46]. The obtained higher alcohols over the K modified catalyst are mainly composed of linear primary alcohols as well as branched alcohols like 2-methyl-1-propanol, 2-methyl-1-butanol, and 2-methyl-1-pentanol (Figs. S6–9). These branched alcohols were suggested to be formed via a β-addition process [47,48]. We have recently proposed that the linear primary alcohols are formed through CO insertion, while the branched alcohols are formed by CO insertion and CHx β-addition [16,49], see Schema 1 for details. n-Propanol is formed through both routes, supported by the high amount (> 97%) of n-propanol in total propanol fraction (Fig. S6) (Scheme 1 ).In the current investigation, the role of Co on product selectivity was investigated. Upon Co addition, the CH4 selectivity is lowered slightly from 17.7% for K-MoSx to 16.8% for the K-Co-MoSx-0.13 catalyst. A further increase in Co in the catalyst formulation leads to a gradual increase in CH4 selectivity (Fig. 7), suggesting a somewhat higher hydrogenation ability. The latter may be due to the presence of higher amounts of (K promoted) CoS2 species (Figs. 1, 2 and 4) in the catalysts at higher Co contents.The selectivity to alcohols in general and C3+ alcohols in particular shows an optimum for the K-Co-MoSx-0.13 catalyst and decreases with higher Co loadings (Figs. 7 and 9 ). These findings are rationalized by considering that the amounts of Co-Mo-S and CoS2 phases in the Co-MoSx-0.13 catalyst are highest and that these are preferred for higher alcohol synthesis. At higher Co contents, considerable amounts of CoMoO4 species are present which result in lower higher alcohol selectivity.The trend as given in Fig. 9 holds for the unpromoted (no K) catalysts. Analyses of a K-promoted catalyst (K-Co-MoSx-0.13) by XRD and HRTEM shows that the CoS2 phase, is reduced to Co9S8 (Fig. 5). Based on these findings, we propose that the catalytic performance of the K-MoSx catalyst is enhanced by the addition of Co due to the formation of cobalt sulfides (mainly Co9S8) and a K-promoted (Co)MoSx phase in close proximity. This assembly is given in Scheme 2 and shows a (K promoted) Co9S8 phase sandwiched between two K-promoted (Co)MoSx phases. The (K-)Co9S8 phase gives mainly hydrocarbons for syngas conversions, see results for the Mo free K-Cox provided in this manuscript and literature data [20]. This implies the presence of significant amounts of adsorbed CHx* (and higher carbon number analogs) on the surface of the Co9S8 phase. We assume that efficient transfer of such CHx* species from the Co9S8 phase to adsorbed CH3CHCH2O* species on the K-(Co)MoSx phase occurs, leading to branched alcohols (CHx β-addition mechanism). In addition, linear alcohols are formed by transfer of adsorbed CH3CH2CH2* on the Co9S8 phase to adsorbed CO on the K-(Co)MoSx phase.To determine the effects of process conditions on CO conversion and product selectivity (particularly C3+ alcohols), a total of 44 experiments were performed in the continuous set-up at a range of 340–380 °C 8.7–14.7 MPa, GHSV of 4500–27000 mL g-1 h-1 and H2/CO ratio of 1.0–2.0 for the best catalyst (K-Co-MoSx-0.13) based on the benchmark experiments. In the initial stage, one variable was changed within the range while the other variables were kept constant (Figs. S10–18). This allows for determination of the individual effects of a variable on the CO conversion and product selectivity. In a later stage all experimental data (Table 3 ) were used simultaneously to develop multivariable nonlinear regression models of the form given in Eq. (5). This approach allowed the identification of interactions between the variables (T, P, GHSV and H2/CO ratio) on the selectivity and yield of C3+ alcohol.The yield (%) and selectivity (%) of C3+ alcohol as a function of reaction conditions were successfully modeled and the results are given in Eqs. (6) and (7), respectively. (6) Y i e l d = 2.05 × P + 0.13 × T + 0.00034 × G H S V - 1.27 × R a t i o - 0.000021 × P × G H S V - 0.088 × P 2 - 3.91 × 10 - 9 × G H S V 2 - 51.51 (7) S e l e c t i v i t y = - 11.45 × P + 0.20 × T + 0.0037 × G H S V - 5.14 × R a t i o - 0.00017 × P × G H S V - 0.00029 × G H S V × R a t i o + 0.41 × P 2 - 1.86 × 10 - 8 × G H S V 2 + 20.77 The high F-value of both models (Tables S5–6) implies that the models are significant and adequate to represent the actual relationship between the response and the variables [50]. The models also reveal that interactions between parameters are significant (e.g. P × GHSV and GHSV × Ratio). The predicted values of C3+ alcohol yield and selectivity match well with the experiment data (Fig. S19–20, R2 = 0.92 for yield and R2 = 0.91 for selectivity).The effect of the pressure and GHSV on C3+ alcohol yield (Fig. 10 ) and selectivity (Fig. S21) are represented in response surface plots. It shows that intermediate pressure and GHSV are best for highest C3+ alcohol yield. This is confirmed by experiments in this regime, viz. a C3+ alcohol yield of 9.2% at 11.7 MPa, GHSV of 13500 mL g-1 h-1 (380 °C, H2/CO ratio of 1, Table 3, entry 11). The model also predicts that a relatively high temperature and low H2/CO ratio are also best for higher alcohol synthesis (surface plots not shown for brevity).The experimentally obtained C3+ alcohol selectivity at different CO conversion over the best catalyst (K-Co-MoSx-0.13) in this study is given in Fig. 11 , together with literature data for other Mo based catalysts. Details regarding reaction conditions are shown in Table S7. Literature sources providing alcohol selectivity only on a CO2-free basis were excluded since this leads to an overestimation of the actual C3+ alcohol selectivity and thus does not enable a fair comparison. The majority of the KMoS2-based catalyst reported in the literature are promoted by Co or Ni and are supported on activated carbon (AC), carbon nanotubes (CNT), mixed metal oxides (MMO) and Al2O3.It is clear that the best catalysts identified in this work (K-Co-MoSx-0.13) outperforms all existing Mo-based catalysts. In comparison with the Co free K-MoS2 catalyst reported previously by our groups (Table S7, entry 5), promotion with the appropriate amount of Co leads to higher selectivity and yield for C3+ alcohol.We have prepared a series of K-Co-MoSx catalyst with different Co contents to investigate the effect of Co promotion on product selectivity and particularly C3+ alcohol formation from syngas. The preparation of the Co-MoSx samples through sulfurization of cobalt-molybdenum oxide precursors leads to among others the formation of Co-Mo-S and CoS2 phases, the actual amounts being dependent on the Co amount in the catalyst formulation. The best performance was obtained using the K-Co-MoSx-0.13 catalyst. This catalyst contains the highest amounts of Co-Mo-S and Co9S8 phases, implying that these are preferred for higher alcohol synthesis. It is speculated that close contact between a potassium modified Co9S8 phase and a Co promoted Mo-S phases is beneficial for higher alcohol synthesis due to facile transfer of adsorbed CHx* species (and higher analogs) on the Co9S8 phase to oxygenated species on the Co promoted Mo-S phase to give branched higher alcohols and transfer of adsorbed CH3CH2CH2* on the Co9S8 phase to adsorbed CO on the K-(Co)MoS phase to give linear alcohols. Reaction conditions (T, P, GHSV and H2/CO ratio) were varied to study the effect on catalytic performance and models with high significance were developed. Highest C3+ alcohol yields of 7.3–9.2% and selectivities between 31.0–37.6% were obtained at a temperature of 380 °C, a pressure of 11.7 MPa, a GHSV of 13500–27000 mL g-1 h-1 and H2/CO ratio of 1 over the optimized K-Co-MoSx-0.13 catalyst. These results are the highest reported in the literature so far, and indicate the potential of such catalysts for further scale up studies.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. Xiaoying Xi: Investigation, Data curation, Formal analysis, Writing - original draft. Feng Zeng: Investigation, Data curation, Formal analysis, Writing - original draft. Huatang Cao: Investigation, Data curation, Formal analysis. Catia Cannilla: Investigation, Data curation, Formal analysis, Writing - review & editing. Timo Bisswanger: Data curation, Formal analysis, Writing - review & editing. Sytze de Graaf: Investigation, Data curation, Formal analysis. Yutao Pei: Supervision, Validation, Writing - review & editing. Francesco Frusteri: Supervision, Validation, Writing - review & editing. Christoph Stampfer: Data curation, Formal analysis. Regina Palkovits: Conceptualization, Supervision, Validation, Writing - review & editing. Hero Jan Heeres: Conceptualization, Funding acquisition, Supervision, Validation, Writing - review & editing.Xiaoying Xi and Feng Zeng acknowledge the China Scholarship Council (CSC) for financial support.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.apcatb.2020.118950.The following is Supplementary data to this article:

K-Co-MoSx catalysts varying in Co content were prepared to investigate the role of Co in this catalyst formulation for the synthesis of C3+ alcohols from syngas. The Co-MoSx precursors and the best performing K-doped version were characterized in detail and the amount of active cobalt sulfide and mixed metal sulfide (Co-Mo-S) phases were shown to be a function of the Co content. The catalysts were tested in a continuous set-up at 360 °C, 8.7 MPa, a GHSV of 4500 mL g-1 h-1 and a H2/CO ratio of 1. The highest alcohol selectivity of 47.1%, with 61% in the C3+ range, was obtained using the K-Co-MoSx catalyst with a Co/(Co + Mo) molar ratio of 0.13. These findings were rationalized considering the amount and interactions between cobalt sulfide and Co-Mo-S or MoS2 phases. Process studies followed by statistical modeling gave a C3+ alcohol selectivity of 31.0% (yield of 9.2%) at a CO conversion of 29.8% at optimized conditions.

O2/H2 fuel cells were invented by Schönbein [1] and Grove [2] in 1839 and they are still not widely used. The 4-electron oxygen reduction reaction (ORR) is a very spontaneous reaction in H2/O2 fuel cells (O2 + 2H2 → 2H2O), but its kinetics are sluggish on most electrode surfaces because it involves several electron transfer steps and consequently several activation barriers [3]. The ORR at the cathode represents the bottleneck in fuel cell performance. The cathode requires the presence of expensive catalysts. Most active catalytic materials contain Pd or Pt-based catalysts (for alkaline and proton exchange membrane fuel cells, respectively) [4–8]. Catalysts containing very low amounts of highly dispersed Pt group metals have been developed, but they still contribute substantially to the cost of most fuel cells. For this reason, the search for less expensive catalysts using earth abundant elements has been active for five decades [9–15]. Earlier catalytic materials belonged to the MN4 or MNx class involving metal-phthalocyanines, metal-porphyrins, and molecules alike (refer Figure 1 ). These molecular catalysts are cheaper than Pt group metals and exhibit rather high activities when immobilized on graphitic and carbon surfaces, but most of them do not have long-term stability in fuel cells electrolytes, especially in acid [16–20]. However they have served as models to establish reactivity descriptors that essentially indicate that the metal centers need to have a rather positive M(III)/(II) redox potential (ideally close to the O2/H2O reversible potential) and moderate M–O2 binding energies [21,22], essentially similar to those for Pt. On the other hand, their low stability has been improved by heat-treatment of intact MN4 complexes or by pyrolysis using ingredients containing the necessary elements C, N, and a metal, generally Fe. Most MN4 complexes have pyrrolic inner ligands, but pyrolyzed materials have pyridinic inner ligands, and they have been modeled recently using Fe(phen)2N2 chelates [23]. Many procedures have been reported to prepare these pyrolyzed materials involving heat treatments up to 1000 °C, and stability is linked to the method of preparation. Even though stability is crucial for practical applications, it is an issue that has not been addressed with the same emphasis compared with that for achieving high activity [24–27]. Hence, it is important to develop unified stability–activity relationships as guidelines for the development of realistic catalysts for fuel cells.Recently, the attention has been focused on two reactivity descriptors, the active metal site density ( S D [ m o l s i t e c m − 2 ] , or S D m a s s [ s i t e g c a t − 1 ] , or S D v o l [ s i t e c m − 3 ] ) and the catalytic turn-over frequency ( T O F [ s − 1 ] or [ e l e c t r o n s i t e − 1 s − 1 ] ), that is the electrons transferred per active site per second [28]. Increasing the value of one, or both of these descriptors, predicts highly active electrocatalysts. The combination of TOF and SD, together with the Faraday constant F, the number of electrons involved in the reaction, n, and τ CL the thickness of the catalyst layer, provide the kinetic activity of platinum group metal-free (PGM-free) catalysts at a specific potential [29,30]: J k i n [ A c m − 2 ] = n · F [ C m o l − 1 ] · T O F [ s − 1 ] · S D [ m o l s i t e c m − 2 ] or J k i n , m a s s [ A g − 1 ] = T O F [ e l e c t r o n s i t e − 1 s − 1 ] · S D m a s s [ s i t e g − 1 ] · e [ C e l e c t r o n − 1 ] or J k i n , v o l [ A c m − 2 ] = T O F [ e l e c t r o n s i t e − 1 s − 1 ] · S D v o l [ s i t e c m − 3 ] · e [ C e l e c t r o n − 1 ] · τ C L [ c m ] The combination of TOF and SD can be considered as a rigorous comparison between catalysts [28]. TOF can be estimated in different ways and it is very important to know how they are estimated. A typical technique to measure both SD, or better the gravimetric site density S D m a s s [ s i t e g c a t − 1 ] , and TOF, is represented by the ex situ low temperature (or cryo) CO chemisorption/desorption at −80 °C. CO rapidly and strongly adsorbs on oxygen-free Fe(II)-Nx sites [31]. Moreover, the amount of CO adsorbed is monotonic proportional to the ORR activity [32,33], i.e. one adsorbed CO molecule corresponds to one Fe(II)-Nx moiety at the surface of the catalyst. Thus, the measurement of the CO uptake, n CO , allows calculating the SD mass , which can be further used in combination with J kin to calculate the TOF. S D m a s s / C O [ s i t e g c a t − 1 ] = n C O [ m o l g c a t − 1 ] · N A [ s i t e m o l − 1 ] T O F C O [ s − 1 ] = J k i n , m a s s [ A g c a t − 1 ] · N A [ s i t e m o l − 1 ] F [ C m o l − 1 ] · S D m a s s / C O [ s i t e g c a t − 1 ] A strip protocol of cleaning the catalyst surface of oxygen, followed by a series of CO pulses to reach saturation of the active centers and the temperature program desorption allows a very precise measurement [32,33]. SD can also be estimated with an in situ electrochemical technique based on the adsorption of nitrite and electrostripping of NO through a 5-electron reaction on Fe(II)-Nx active sites, assuming that one NO molecule poisons one site. The determination of Q strip , that is the excess coulometric charge associated with the stripping peak, together with the number of electrons necessary to reduce the nitrite ion, n strip , and the specific surface area of the catalyst, S BET , allows determining the SD. NO2 – anions adsorption largely affects the ORR activity of the electrocatalyst, by poisoning the active sites. Thus, it is also possible to evaluate the TOF at a certain potential by measuring the kinetic mass current of the catalyst as the difference of the unpoisoned and poisoned kinetic mass current values, ΔJ kin ( J k i n , m a s s u n p o i s o n e d − J k i n , m a s s p o i s o n e d ). This technique has been developed by the group of Kucernak et al. [29,30,34], it requires a series of subsequent steps of cleaning, poisoning, and stripping of the reaction products from the catalyst layer for the determination of Q strip and ΔJ kin . S D m a s s / N O 2 − [ m o l   s i t e g − 1 ] = Q s t r i p [ C g − 1 ] n s t r i p · F [ C m o l − 1 ] T O F N O 2 − [ s − 1 ] = Δ J k i n [ A g − 1 ] F [ C m o l − 1 ] · S D m a s s / N O 2 − [ m o l   s i t e g − 1 ] Double layer (DL) capacitance can be used to estimate the area of electrodes. For pyrolyzed catalysts, the wetted surface area can be estimated from Electrochemical Impedance Spectroscopy (EIS) measurements [35], but it is very dependent on the type of carbon or graphitic materials, graphitic edges, and presence of carbon functionalities [36]. Thus, DL capacitances can vary from as low as 4 μF cm−2 (defect-free basal plane graphite) to 60 μF cm−2 for more heterogeneous carbon/graphitic materials and in the average, most materials exhibit DL-capacitances around 20–25 μF cm−2 [36].It is important to know how TOF values are estimated. A correct TOF estimation should only consider the active sites available for the reaction. An overestimation will lead to lower TOF values. For example, a catalyst containing Fe as active sites, where Fe bulk is not effective for several reasons: oxygen has no access to those sites but they are considered in the TOF calculation. There could be another reason for those sites not be active: they are partially covered or occupied by adsorbed intermediates resulting from the reaction and another: some of those active sites are in the wrong oxidation state at the particular electrode potential. It is widely accepted that only M(II) is active.This is illustrated in Figure 2 a and shows a clear volcano correlation for several MN4 molecular catalysts adsorbed on a smooth graphite surface. Hypothetically, it is assumed that all active sites are accessible to oxygen as the adsorbed MN4 molecules are lying flat on the graphite surface, and there are no MN4 molecules imbedded in the graphite. Those electrodes were made of graphite crystals, so porosity is very low or absent. NiPc shows low activity because the interaction between Ni and O2 is too weak. The opposite is true for CrPc. The classical arguments for the low activity of CrPc would be that most active sites are occupied by strongly bound oxygen intermediates. However, the most important factor contributing to the low activity of CrPc is that its Cr(III)/(II) redox potential is too negative compared to the potential at which ORR currents are compared. Thus, most Cr sites are in the inactive state Cr(III). However, if the currents are divided by the fraction of sites θ M(II) in the active state M(II) for all MN4 catalysts, the volcano correlation becomes a linear correlation and the activity per active sites increases going from weak oxygen binding catalysts (right) to strong binding catalysts (left). The same happens if the TOF values are plotted versus the M–O2 binding energy in Figure 2b, as log(i/θ M(II))E values are directly proportional to TOF values but with the slight difference that for strongly binding catalysts n = 4 (O2 reduction to OH–) and for weak binding catalysts n = 2 (O2 reduction to peroxide).In Figure 2a CrPc shows very low activity but exhibits the highest TOF in Figure 2b. The best catalyst illustrated in Figure 2a is FePyPz that shows a TOF value almost 5 orders of magnitude lower than the poor CrPc catalyst. Yang et al. [37] reported a series of TOF values for the activity of FePc for ORR in 0.1 M KOH using several carbon substrates and they report values that vary in the range from 0.5 to 2.8 at E = 0.868 V vs RHE. However, these values are underestimated because that particular potential is the formal potential of the Fe(III)/(II) couple of FePc. At this potential θ Fe(II) = 0.5, consequently the TOF values are underestimated by a factor of 2. Therefore, errors can be introduced if the potential for comparing ORR currents is close to the potential of that redox couple. For pyrolyzed materials, θ M(II) cannot be estimated easily as most of these materials exhibit no clear redox signals that can be attributed to the M(III)/(II) redox couple.The two in situ and ex situ techniques mentioned before let the estimation of reactivity descriptors of PGM-free catalysts with a high degree of precision, and with comparable results. Usually, TOF values estimated from CO chemisorption result slightly lower that those calculated by nitrite stripping because of the overestimation of the related SD, or different way to calculate the kinetic current in the two methods [33]. Moreover, also the difference between gas-phase accessibility (CO chemisorption) and electrochemical accessibility (nitrite reactivity) plays a role in the different values obtained by measuring the SD by in situ or ex situ methods. In fact, the electrochemical surface can match rather well the gas-phase surface for Fe–N–C materials with low S BET (specific surface area, that is, relatively low amount of micropores), although the two values can be quite different in the case of high S BET , where the micropore area is prevalent [33]. However, Fe–N–C materials have modest intrinsic catalytic activity, lower than Pt, especially in acid, obliging increasing the catalytic loading at the cathode to compensate for the overall activity [28,38–40]. In addition, thicker catalytic layers represent a limitation in terms of mass transport resistance, electronic resistance, and proton resistance [38], not only at RDE level but especially at MEA (membrane electrode assembly) level, when testing polymer electrolyte fuel cells [40–42]. In fact, when working at high power density, thin electrodes are required to limit mass-transport–related voltage losses [43]. Thus, enhanced proton transport properties of the active site are essential for a high TOF in acid [44]. Just as an example, Pt-based catalysts can easily reach TOF values ranging between 10 and 42 e– site−1 s−1 at reported conditions [28,43], whereas Fe–N–C materials are at much lower order of magnitude, not overcoming 2 e– site−1 s−1 at reported conditions [33,45].Different groups have investigated MN4 macrocyclic complexes not subjected to any heat treatment. They provide simple models to identify some reactivity descriptors because active sites are clearly identified, especially the metal centers in the MN4 moiety [21] as showing well-defined CV redox peaks attributed to metal-centered redox processes. The charge under those reversible peaks allows the accurate determination of the amount or electroactive catalysts present as S D [ m o l s i t e c m − 2 ] . Most reports agree that they lack the long-time stability/durability required for fuel-cell performance, especially in acid. However, Cao et al. [46] reported that FePc can exhibit higher activity than Pt/C in alkaline media and a stability higher than 1000 cycles, when anchored on carbon nanotubes via a pyridine axial ligand (FePc-Py-SWCNT) (refer Figure 1). Yang et al. [37] have studied FePc as well and directly deposited on different nanocarbons. The activity depends strongly on the type of carbon substrate used. Figure 3 a illustrates the TOF values for FePc on different carbon supports and the highest TOF value is observed when C450, a 3D nanoporous C with macropores of 450 nm, is used [37]. The low stability of FePc-based catalysts having high TOF values could be due to the fact that highly reactive sites for ORR can also be highly reactive to other species and/or intermediates such as peroxide and OH radicals that are generated faster than less active catalysts attacking those sites, besides carbon corrosion [35]. Similar mechanisms have been described for pyrolyzed metal–nitrogen–carbon catalysts [33,46].Yang et al. [37] calculated these TOF values according to T O F [ s − 1 ] = i [ A c m − 2 ] 4 · Q F e , a c t i v e [ C c m − 2 ] where i is the current at 0.868 VRHE and Q Fe,active represents the amount of electrochemically active centered Fe ion, estimated from the area of the Fe(III)/(II) redox peak determined by cyclic voltammetry under N2-saturated 0.1 KOH. Q Fe,active depends on the morphology of the carbon support (specific surface area, pore structure, and roughness of the surface of the carbon), which affects somehow the amount of FePc deposited on the support itself at equal deposition procedures [37].As explained before, those TOF values in Fig. 3 are probably underestimated because the number of Fe(II) active sites could be lower than the total sites N (mol cm−2). In fact, at that particular potential, a fraction of the sites are in the oxidation state Fe(III) as 0.868 VRHE is too close to the Fe(III)/(II) formal potential of the catalyst. A rough estimation using a formal potential of 0.868 VRHE using the Nernst equation (estimated from the peak of CV curves reported, which coincides with the potential used for comparing activities) indicates that only ca. 50% of the catalyst is active as Fe(II). Fe(III) in alkaline media does not catalyze ORR as those sites are strongly binding OH– ions [21]. However, the order of activity totally changes after a chronoamperometry at 0.868 VRHE, showing that the most durable catalyst (that is the one that lasted longer before reaching the 50% current loss) was the one with the lowest TOF value (FePc/KBC) [37]. Interestingly, the results denote a shorter durability with TOF increasing. This effect can be explained considering that a low TOF value implies a slower formation of reaction intermediates during ORR, such as HO2 –, which can partially hinder the active site. Thus, degradation is limited compared with catalysts with higher TOF values.Cao et al. [46] did not provide any TOF, SD values, or CV profiles that could allow the estimation of Q. The enhancement of the activity of FePc when using a pyridine back-ligand seem to be associated to the electron-withdrawing effect of pyridine that would shift the Fe(III)/(II) redox potential in the positive direction and push the catalyst up towards the top of the volcano correlation of (logi) E versus the Eº′Fe(III)/(II) of the catalyst [21]. The beneficial effect of an electron-withdrawing axial ligand on Fe phthalocyanines has been demonstrated using several FePcs attached to carbon nanotubes [48,49] substrates and gold (111) [50]. These axial ligands seem to mimic the action of similar ligands in enhancing the catalytic activity for ORR of cytochrome-c in the respiratory chain of aerobic life [47,51].C–N materials show ORR activity even in the absence of any transition metal. Chakraborty et al. [52] have proposed a method for evaluating the active SD of metal-free nitrogen-doped carbon using catechol as an adsorbate. These catechol provide well-defined redox peaks that facilitate an indirect estimation of the mass-specific active SD (SD mass ) evaluated from the electrical charge involved in these redox processes according to: S D m a s s [ active site   g − 1 ] = Integrated CV area  [ A V ] · N A [ s i t e m o l − 1 ] n · s c a n r a t e [ V s − 1 ] · F [ C m o l − 1 ] · m [ g ] where N A is the Avogadro number, n the number of electrons, m is the catalyst's mass. The rest of terms have the usual meaning. The electrochemical surface area can be estimated according to (using the normalizing factor of 611 μC cm−2 as charge of unit surface area on graphite [53]): E S A [ c m 2 g − 1 ] = C a t a d s c h a r g e [ C ] 2 · m [ g ] · 611 [ μ C c m − 2 ] Chakraborty et al. [52] double checked their electrochemical surface area estimation using the equation mentioned previously using the BET-specific surface area (S BET ) and multiplied it by the total pyridinic nitrogen percentage, as reported by Guo et al. [53] and found comparable results suggesting the accuracy of the equation aforementioned. This is important for intact and pyrolyzed catalysts as nonmetallic sites can contribute to the catalytic process.Most intact MN4 complexes do not exhibit long-term stabilities that can be compatible with fuel performance. This is less critical in alkaline media, and there are some reports that might promise some success in this sense if the FeN4 complexes are attached to carbon nanotubes directly or linked via pyridinic axial ligands.The combination of TOF and SD can be considered as a rigorous comparison between catalysts [28]. There are different ways of estimating TOF values and the values obtained depend on the method used. This is relevant to both intact and pyrolyzed materials. SD should consider only active sites that are available for ORR. Bulk sites do not count. When using TOF values, special care needs to be taken in estimating the amount of M(II) active sites present under operating conditions (electrode potential) because some complexes could exhibit M(III)/(II) redox potentials close to the operating potential.Finally, it has been suggested recently that intact complexes, which possess pyrrolic N inner nitrogens are not representative of real active sites present in MNx pyrolyzed catalysts, which predominantly pyridinic N inner ligands. In this case, complex having phenanthroline inner ligands can serve as better models for future studies.The authors contribute equally to the work by discussing and writing the manuscript, drawing the figures, along with approving its final version.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.J.H.Z. acknowledges the funding from Anillo Project ACT 192175 Chile and Fondecyt Project 1181037. S.S. and P.A. acknowledge the funding provided by the Staff Mobility for Training & Teaching between Programme and Partner Countries within the Program Erasmus+/KA1 Higher Education action KA107 (International Credit Mobility, years 2017 and 2018) for the reciprocal visits @ UCI (in 2019) and POLITO (in 2020), respectively.

There has been a significant progress toward the development of highly active and stable platinum group metal-free (PGM-free) electrocatalysts for the oxygen reduction reaction (ORR) in polymer electrolyte fuel cells, promising a low-cost replacement for Pt group electrocatalysts. However, the success of such developments depends on the implementation of PGM-free electrocatalysts that are not only highly active but importantly, they also exhibit long-term durability under polymer electrolyte fuel cell operating conditions. This manuscript is an overview of the current status of a specific, most advanced class of PGM-free electrocatalysts: transition metal–nitrogen–carbon ORR catalysts. We present an overview for the understanding of catalysts’ performance descriptors for metal–nitrogen–carbon materials.

Isobutene (2-methylpropene) is a vital base chemical extensively used as a building block for the synthesis of a vast number of intermediates in the chemical industry. It is an essential precursor for the synthesis of various oxygenates like methyl tert-butyl ether (MTBE), ethyl tert-butyl ether (ETBE), and methacrylates which are used as octane enhancing additives in gasoline [1]. Isobutene is also widely used in the polymer industry for the production of butyl rubber [2]. The estimated increasing demand for isobutene soon insists on the need for alternative synthesis pathways other than conventional naphtha steam cracking and crude oil fluid catalytic cracking methods. In this regard, the catalytic dehydrogenation of isobutane acquires much significance due to the available low-cost feedstocks [3]. However, the endothermicity of this process requires an elevated operational temperature to obtain high yield of isobutene.Chromium and platinum-based catalysts have been extensively studied for the dehydrogenation of isobutane (DHisoB). Pt-Sn/Al2O3 and CrOx/Al2O3 systems are well explored and already implemented in the industry a few decades back. Even though these catalytic systems are giving satisfactory results, they are suffered from some disadvantages. A part of chromium ions that exist in carcinogenic Cr6+ in alumina supported chromium catalyst causes severe environmental threats [4]. Also, possible sintering of Pt nanoparticles and high-cost limits Pt-based catalysts to some extent [5]. Moreover, the catalyst deactivation due to coke deposition is unpreventable at stringent reaction conditions. Hence, the development of an environmentally friendly and cost-effective promising catalyst for the non-oxidative dehydrogenation of isobutane becomes imperative.Catalyst support has a significant role in defining the activity performance. Especially for a dehydrogenation reaction, the support should be thermally stable to survive the rigorous reaction conditions. Moreover, the limited acidity can evade undesired C-C cracking and alkane isomerization reactions. High surface area and uniform pore size distribution of the support will enhance homogeneous metal dispersion [6]. While considering all these aspects, the nonreducible metal oxide, alumina, serves as an excellent candidate for the dehydrogenation reaction owing to its high thermal stability as well as mechanical strength. Moreover, state of the art drawn from the existing literature evinced alumina as the most preferred support for DHisoB, including various industrial applications [7–9]. Meanwhile, the commercialized catalysts have been modified to overcome the current difficulties and to improve activity by adding different metal oxides as active components [10]. Recently, Uwe et al. have proved that bare alumina itself is active for the reaction with a 30% yield towards isobutene due to the coordinatively unsaturated Al sites on the surface [11]. The supported VOx, GaOx, and MoOx materials on alumina have improved the conversion and selectivity for DHsioB [6].Alumina is acidic, and hence it should be modified to control the acidity. Researchers have successfully attempted to reduce the acidic properties of alumina by combining with ZnO and MgO, which has improved the dehydrogenation activity and catalyst stability [12]. The addition of alkali metals can also contribute to this by selectively poisoning the acidic sites and thus hinder the coke formation [13,14]. Shingo et al. have established that the addition of a small amount of iron can improve the activity, selectivity, and stability of Pt/Al2O3 catalyst for the DHsioB. NH3-TPD studies have proved a considerable reduction in influential acid sites after Fe addition [8], which is in good agreement with the investigations of Kania et al. [15]. These prominent acidic sites prompt alkene hydrogenolysis and decrease isobutene selectivity [16]. Apart from these, iron oxide-containing activated carbon has also served as a suitable catalyst for the dehydrogenation of C4-C5 hydrocarbons [17].Inspired by these described observations, in this contribution, we illustrated the synthesis of chromium-free catalytic system based on iron-doped mesoporous alumina system. To solve the above-discussed problems, the catalyst is separately modified with an alkali metal, phosphorous as well as noble metal via dry impregnation method. The synthesized materials are characterized for their topology, morphology and chemical properties by an array of instrumentation techniques including powder X-ray diffraction (PXRD), N2 physisorption analysis, scanning electron microscopy (SEM), transmission electron microscopy (TEM), Raman spectroscopy, X-ray photoelectron spectroscopy (XPS), temperature-programmed reduction (H2-TPR), temperature-programmed desorption (NH3-TPD) and thermogravimetric analysis (TGA). The materials displayed promising catalytic activity towards the non-oxidative dehydrogenation of isobutane. Iron doped alumina based catalyst systems were explored for different reactions. However, in the current study Ag, K and P are used as promoters to enhance the catalytic activity of isobutane dehydrogenation and is reporting for the first time up to our knowledge.The chemicals with more than 99% purity were used as received from the respected company for material synthesis. Fe doped mesoporous alumina catalyst was prepared via method reported by Bing Yan et al. [18]. In a typical synthesis, roughly 1.5 g of F-127 (Sigma Aldrich) and 0.36 g Fe(NO3)3.9H2O (Merck) was added to 20 mL anhydrous ethanol (99.9%, CSS) and vigorously stirred for 4 h and labeled as ‘A.’ Meanwhile, to a stirring mixture of 15 mL anhydrous ethanol and 1.6 mL concentrated nitric acid (69–70%, Thomas baker), 12 mmol aluminum isopropoxide (Aldrich) was added and labeled as ‘B.’ Both A and B were combined using 3.0 mL ethanol to transfer the solution B. This final mixture was stirred continuously for 8 h and the obtained residue was dried in the oven at 60 °C for 48 h. Further, the solid was ground well and calcined at 700 °C for 5 h with 1 °C min−1 ramping rate. The final powder was represented as MesoFeAl.For the synthesis of promoted catalysts, 1 wt% of K, P, and Ag were dry impregnated over 2 g MesoFeAl from a concentrated solution of the precursors KNO3 (Merck), orthophosphoric acid (88%, Merck) and AgNO3 (99.995%, Alfa Aesar) respectively. These solids were sintered at 550 °C for 2 h with 1 °C min−1 and denoted as KMesoFeAl, PMesoFeAl, and AgMesoFeAl. For the comparison study, the pristine support mesoporous alumina (MesoAl) was also synthesized by following the above procedure without adding the Fe precursor. Later, Fe was impregnated from the nitrate precursor over MesoAl and represented as FeMesoAl.The catalytic activity of as-synthesized mesoporous alumina catalysts was measured for the non-oxidative dehydrogenation of isobutane in a continuous flow mode fixed bed reactor system equipped with an Inconel (nickel‑chromium based alloy) made reactor tube of 8*11*480 (ID:OD:L) mm dimension. The reactor tube was heated in a furnace fitted with two zones maintained at the same reaction temperature. In a single measurement, 300 mg catalyst sieved into 1.2 -1.7 mm grain size pellets were loaded at the center of the reactor tube in between quartz wool to form a catalyst bed. High temperature stable ceramic beads were used to fill the remaining space. The bed temperature was monitored continuously with K type thermocouple. A mixture of isobutane and Ar taken in an equal ratio was continuously passed into the catalyst bed using 5890E series Brook's make mass flow controllers at 400 h−1 GHSV. The reaction temperature was varied between 400 and 600 °C for the study in different runs. The dehydrogenated products were analyzed in every 30 min time interval with a Thermo Scientific Trace 1110 gas chromatograph equipped with both FID coupled with Alumina plot column and TCD coupled with Porapak Q as well as Molecular sieve columns. The major components in the effluents were n-butane, 2-butenes, 1,3-butadiene, propene, propane, ethene, ethane, methane and hydrogen. No carbon oxides were detected. Conversion of isobutane, and selectivity for isobutene, werecalculated using the equations reported elsewhere [10].Crystalline features of the synthesized materials were collected by powder X-ray diffraction analysis in PANalytical X'pert Pro dual goniometer diffractometer. The data were collected with a step of 0.008° (2θ) and a scan rate of 0.5° min−1 at room temperature. The radiation applied was Cu Kα (1.5418 Å) with a Ni filter, and the data was obtained using a flat holder in Bragg-Brentango geometry. Nitrogen physisorption was analyzed for the textural properties of the materials from Quantachrome Quadrasorb SI using the Brunauer-Emmett-Teller (BET) model. N2 sorption data were examined at −196 °C after degassing the samples at 300 °C and the relative pressure of P/P0 = 0.05–0.3 was selected to calculate surface area. Horiba JY LabRAMHR800 Raman spectrometer coupled with a microscope in reflectance mode was used with a 628 nm excitation wavelength for Raman studies. X-ray photoelectron spectra were acquired from the pelletized materials on the Thermo Scientific K-Alpha+ instrument using micro-focused and monochromated Al Kα radiation with energy 1486.6 eV.Temperature programmed reduction experiments were carried out in a Micromeritics Autocem II 2920 chemisorption analyzer (USA). In a single TPR analysis, 0.05 g catalyst was pre-treated at 400 °C inside the furnace coupled with the instrument with 30 cm3 min−1 of 10% O2/He controlled by Brooks make mass flow controllers. Then, the sample was reduced with 5% H2/Ar at a 30 cm3 min−1 flow rate while ramping the sample temperature from 50 to 1000 °C at 10 °C min−1. Hydrogen consumption was quantified using a thermal conductivity detector. Acidity of the materials were determined in the same model instrument. In a typical TPD study, the samples were degassed at 400 °C under Helium at 30 cm3 min−1 flow before each run. Afterwards, the sample temperature was brought down to 50 °C to adsorb the probe molecule at 10% NH3/He (30 cm3 min−1) for the acidic site evaluation. The desorbed gas was analyzed from 100 to 1000 °C with 10 °C min−1 ramp rate. The quantification was made from the resulting TPD profiles.A dual-beam scanning electron microscope FEI company made and Quanta 200 3D model operated at 30 kV was used to collect scanning electron microscopy images. A high-resolution transmission electron spectroscopy study was performed in JEM2100 multipurpose instrument operated at 300 kV. The sample was prepared by dispersing the powder in isopropyl alcohol followed by drop-casting on carbon‑copper mesh (200 μm size) and silicon wafer respectively for TEM and SEM study. Coke deposition over the spent catalyst was determined with the help of the Perkin Elmer instrument. TG analysis was performed under the air atmosphere and the carbon combustion temperature was taken from the exothermic event in DTA.A conventional fixed bed reactor was used to study the catalytic activity of the mesoporous alumina catalysts. The effect of different promoters, as well as reaction parameters like temperature on the activity performance, was investigated through various experiments. The main products were obtained from the dehydrogenation, isomerization, and cracking processes occurred during the reaction (Scheme 1 ). Coke deposition was also detected near the catalyst bed at the end of each experiment.Catalyst promoters have been utilized on an existing catalyst to improve the product yield. Therefore, the effect of different promoters on the selective dehydrogenation of isobutane to isobutene was performed at 600 °C. All experiments were conducted at 400 h−1 GHSV concerning isobutane and argon mixture taken in 1:1 ratio. The results are depicted in Fig. 1 as well as in Table S1. For the comparison study, the reaction was carried out over MesoAl also. Interestingly, as can be seen, the bare support exhibited appreciable activity under the given reaction conditions with the highest selectivity towards total dehydrogenating products (STDP), which include isobutene, 1-butene, 2-butenes, propene, and ethene. However, the final conversion of isobutane (Xisobutane) and selectivity for isobutene (Sisobutene) were low. The incorporation of Fe simultaneously decreases STDP and increases isobutane conversion. It can be realized from the activity performance given in Fig. 1 that the addition of small amount of promoters only slightly affected the dehydrogenation reaction pattern at 600 °C. Potassium promoted catalyst (KMesoFeAl) showed excellent selectivity towards the dehydrogenation reaction while conversion was comparatively low. Conversely, the acid-treated material (PMesoFeAl) was less selective for isobutene with a better isobutane conversion. Surprisingly, the noble metal promoted AgMesoFeAl displayed slightly higher yield of isobutene, Yisobutene (Table S1) at the studied reaction parameters. Isobutane conversion achieved was 36%, with 32% selectivity towards isobutene. It is essential to highlight the considerable STDP over the same catalyst.To investigate the effect of the synthesis method adopted for Fe incorporation in MesoAl support on the non-oxidative dehydrogenation reaction, FeMesoAl was prepared by dry impregnation method. Moreover, the activity performance of this catalyst was compared with the in situ prepared MesoFeAl catalyst. The activity results represented in Fig. 1 clearly illustrates that FeMesoAl shows the least selectivity towards the dehydrogenation products. The reason for this observation was analyzed with the material properties and explained in the respective sections. It may be interpreted in terms of the amount of coke deposited on the surface, which is quantified from TG analysis. From the above results, AgMesoFeAl was selected as a representative catalyst for further analysis.Fig. S1 shows isobutane conversion and isobutene selectivity as a function of reaction temperature over AgMesoFeAl catalyst. At 400 °C, Xisobutane was insignificant with the least isobutene selectivity due to the formation of other dehydrogenated products. The main process that occurred at this temperature would be isomerization of the formed alkene to n-butenes. On further increasing the reaction temperature up to 500 °C, conversion slightly improves accompanied by the steady increase in the product selectivity. A sharp increase in reactant conversion is observed above 500 °C. Contrastively, the selectivity slowly increases with the reaction temperature and remained constant after 550 °C. Thus, Fig. S1 illustrates that 36% isobutane is converted with 32% selectivity towards isobutene at 600 °C. The formation of other alkenes is also in line with the isobutene selectivity pattern when plotted against reaction temperature (Fig. S1). At higher temperatures, isobutene was further cracked into lighter hydrocarbons. This results in coke formation which is deposited on the catalyst surface and blocks the active sites. Post characterization of the spent catalysts supports this observation. This activity trend with temperature may also be correlated to the particle size distribution over the alumina support during the TOS analysis.The stability of a catalyst under the reaction conditions has a major significance while approaching commercialization. Hence, the time on stream study of AgMesoFeAl was executed at an optimized reaction temperature; 600 °C. Fig. 2 further depicts the long term performance of AgMesoFeAl. It was perceived from the study that at an early time on stream, the conversion is 30% with appreciable selectivity for dehydrogenated products. Gradually, Xisobutane improves during this initial 4 h phase and correspondingly Sisobutene decreases. The most top performance over this catalyst was achieved as 32% isobutene selectivity and 36% conversion of isobutane under the studied parameters. After 4 h, a steady reduction in conversion (30%) as well as selectivity (30%) is observed. Subsequently, this activity remained constant above 10 h. Followed by this initial phase, nearly stable activity was maintained up to 60 h.The main significance of the present work is the maintenance of overall product yield over 60 h under the drastic reaction conditions. Additionally, here dehydrogenation is practiced with low space velocity. Major reasons for the decrease in activity after 10 h and the maintenance of isobutene yield in TOS are explained in the material characterization section based on the particle distribution and carbon deposition on the catalyst.All synthesized mesoporous alumina catalysts were well studied for the structural-textural properties, and surface morphology. The acivity results presented in the previous section are correlated to establish the relationship between the performance and material characteristics.N2-sorption was used to characterize the textural features of the calcined promoted catalysts and pure MesoAl support. Total surface area (SBET), pore-volume (Vp), and average pore-size values were calculated applying the BET method and presented in Table 1 . According to the IUPAC classification system, the N2-sorption isotherms given in Fig. 3 indicate type IV with H1 hysteresis loop typical of mesoporous materials. However, FeMesoAl possesses slight features of the H2 loop with wide pore size distribution. These are characteristic of bottlenecked pores [19]. A close analysis of isotherms of the promoted catalysts proves that even after the impregnation process, the materials maintained mesoporosity. Pore size distribution curves derived from the N2-desorption process shown in Fig. S2 also supports the presence of mesopores. Additionally, it may be mentioned here that small pores are unfavorable for the migration of coke deposited on the active sites towards the acidic sites [20].All mesoporous materials exhibited an appreciable specific surface area. Surprisingly, the total surface area of MesoAl increases after Fe doping. However, iron impregnation on mesoporous alumina support may block the pores of FeMesoAl in the rich presence of iron oxide resulting in a slight decrese in surface area from 162 to 153 m2/g. The average pore diameter (Dp) considerably reduces in the synthesized catalysts compared to the pristine support MesoAl due to the entering of the active metal species into the pore channels. The uniform mesoporous structure and noticeably large pore volume of these materials can significantly contribute towards catalytic activity. AgMesoFeAl also possess remarkable surface area with comparable pore dimensions with other samples. Fig. 4 presents the X-ray diffractions of the synthesized mesoporous alumina-based catalysts. The diffraction pattern of all samples indicates the formation of γ-Al2O3 phase with JCPDS card no. 10–0425. The peaks observed at 31.9, 37, 39.5, 45.9, 60.5 and 67° are attributed to the diffraction planes 220, 311, 222, 400, 440 and 511 respectively [21]. For MesoFeAl, no peak linked to iron is found suggesting high dispersion of Fe during the in situ synthesis, and the formed iron oxide domains are too small to be detected. However, Fe has indeed detected in XPS and elemental mapping analysis. Moreover, Kobayashi et al. reported that at low content, Fe2O3 could form a solid solution with γ-Al2O3 [8]. Although, the iron impregnated sample FeMesoAl exhibited peaks at 2θ values 24.3, 33.3, 35.9, 41.1, 49.6, 54.3, 57.6, 57.9, 62.7, and 64.4° corresponding to α-Fe2O3 phase (JCPDS card no. 33–0664). This observation stipulates that only a diminutive amount of Fe is doped in the alumina framework and also, iron oxide species are poorly dispersed on mesoporous alumina support due to the ex situ synthesis method. The excess amount of Fe in FeMesoAl would oxidize in the air during the calcination process to form iron oxide and separately crystallizes on the surface.When promoters are added in the MesoFeAl sample, the corresponding peaks are not identified due to the high dispersion of these species on the alumina matrix. Except in FeMesoAl, a small shift in the diffraction angle to lower 2θ values are noticed in other solids compared to the pristine support MesoAl. This could be because of the changes that occurred in the lattice parameters, which might arise from the proper doping of the elements into aluminum site [22]. The difference in ionic radii of Fe3+ (0.645 Å) and Al3+ (0.535 Å) might cause the broadening of γ-Al2O3 reflections. Compared to other materials, AgMesoFeAl best resembled in crystalline structural features with MesoAl proving the perfect incorporation of Fe and Ag. This could be a reason for the observed faintly higher activity of AgMesoFeAl. No silver oxide peaks were found in this sample, because of the minimal size and greater dispersion in the MesoFeAl matrix. Hence it can be concluded that stable phases of Fe2O3 and Al2O3 are formed in the studied catalysts after sintering at high calcination temperature.Further examination of the structural and crystalline properties was conducted with Raman analysis of the annealed samples. Fig. S3 shows the Raman spectra of FeMesoAl, MesoFeAl, and AgMesoFeAl samples. FeMesoAl revealed peaks at 228 (A1g), 504 (A1g), 247(Eg), 297 (Eg), 416 (Eg), 620 (Eg) cm−1 corresponding to the Raman modes of α-Fe2O3. Additionally, some rare Raman peaks reported by a few researchers for the hematite phase are also identified at 669 and 836 cm−1 [23]. Surprisingly, MesoFeAl does not exhibited a single peak attributing to iron oxide, demonstrating either high dispersion of Fe in alumina matrix or it is undetectable in Raman analysis. This observation further supports PXRD and HRTEM studies.Material morphology and proper dispersion of the elements contribute to a great extent towards the catalytic performance. The selected AgMesoFeAl catalyst was analyzed with TEM, and the images collected are shown in Fig. S4. The results exhibit a mesoporous structure which are in consistent with the type of isotherms obtained from the N2 sorption analysis. A sheet-like morphology is visible for the fresh AgMesoFeAl catalyst with evenly distributed metal nanoparticles. High-resolution TEM analysis provided well-resolved lattice fringes with 0.2 nm spacing characteristic of the interplanar distance corresponding to (104) plane of the α-Fe2O3 phase. The SAED pattern of this sample with concentric rings and bright spots represented typical of a nanocrystalline material with an average particle size of 5.6 nm. Small clusters of iron oxide particles on the surface are observed in FeMesoAl, as represented in Fig. S5 demonstrating agglomeration and poor dispersion. The segregated α-Fe2O3 crysatallites are clearly emerging in the PXRD pattern shown in Fig. 4. This could be due to the different synthesis strategies adopted for the material. Therefore, all other catalysts are superior to FeMesoAl in the studied selective dehydrogenation reaction. For example, AgMesoFeAl exhibited better dispersion of metal nanoparticles than FeMesoAl, which is contributing to the catalytic activity.Elemental mapping of fresh AgMesoFeAl as a respect for its best catalytic properties was performed using HRTEM, and the results are given in Fig. S6. The images revealed uniform distribution of the elements over the surface. This further indicates fine dispersion of the catalyst supporting PXRD as well as Raman results. SEM analysis was used to determine the surface morphology of the catalyst, and it does not exhibited any particular morphology as represented in Fig. S7. The particle shape remained the same even after the non-oxidative dehydrogenation reaction. The EDX data also confirms the presence of all elements in the catalyst. No elements other than Ag, Fe, Al, and O were present, which signify the excellent purity of the sample. Therefore, it may be perceived that the proper dispersion of active metal species on the mesoporous alumina matrix can facilitate the dehydrogenation activity. These observations declare the appropriate incorporation of the metal species in alumina consistent with the structural features of the optimized catalyst. Time-dependent DHsioB experiments are carried out over the AgMesoFeAl catalyst to get insight into the morphology changes happening on its surface. The results are depicted in Fig. S4 and explained in the post reaction analysis.Acidic-basic properties of the materials are crucial in defining the catalytic activity. In view of this, temperature-programmed desorption experiments were conducted with NH3 as a probe molecule to measure the number of acid sites present in the catalysts. The uptake of NH3 is propotional to this value. The TPD profiles are given in Fig. S8 and the acidity values measured are presented in Table 1. According to the existing works from the literature, the samples exhibit two significant peaks at around 150, and 450 °C corresponding to NH3 desorbed from the weak and medium Lewis acidic sites, respectively [24]. It is observed from the TPD profile and the acidic strength values shown in Table 1 that bare support MesoAl (0.02 mmol/g) and FeMesoAl (0.07 mmol/g) are least acidic in nature. The in situ incorporation of metal ions enhanced the acidity of the catalysts. Iron doping to MesoAl has considerably improved the acidity to 0.34 mmol/g. As reported by Carvalho et al., strong acid sites cause hydrogenolysis of alkenes [16], which are absent in the studied materials. Isobutane mainly activates on the acidic centers and promotes the conversion. On the other hand, desorption of isobutene may inhibit at the medium and strong acidic centers, which can increase the cracking reactions [20]. The in situ iron-doped catalysts show a comparable amount of medium acidic sites, which matches well with the activity trend proving the role of catalyst acidity. The ratio between the number of active metal sites to the number of acidic sites is significant in defining catalytic performance.The redox properties of the catalysts are another factor affecting the catalyst performance. Therefore, the materials were measured with H2 probed temperature-programmed reduction analysis. Under the reducing atmosphere, the amount of consumed hydrogen for all sintered catalysts is quantified from the TPR profiles (Fig. S9) and tabulated in Table 1. AgMesoFeAl exhibits the highest H2 consumption among the promoted catalysts (1.49 mmol/g) due to the additional reduction of silver oxide, which occurs at low temperature. Except in FeMesoAl, no other materials demonstrated reduction peaks correspond to the iron oxides proving the high dispersion of these species on alumina. On the contrary, the iron oxide species in FeMesoAl is reduced in different steps as Fe2O3 → Fe3O4 → FeO → Fe [25]. Hence, TPR further corroborates the poor dispersion of Fe over mesoporous alumina in FeMesoAl unlike other catalysts. Peak broadening was detected due to the multistep reduction of iron oxide. The highly exposed iron oxide crystallites can lead to further cracking of formed alkenes which result in low selectivity towards the desired isobutene as well as other dehydrogenated products (Fig. 1). Therefore, the highest coke deposits are found on this catalyst (26.2%) as shown in Table 1. Quickly reducible silver oxides in AgMesoFeAl might make it favorable for high-temperature dehydrogenation.To understand the surface species involved, XPS analysis was conducted ex situ for the fresh and spent catalysts, and the results are represented in Fig. S10. All spectra were binding energy (BE) shifted by carbon correction using the C 1 s core level value (284.8 eV) of adventitious carbon on the surface. Spectra recorded for Ag 3d exhibited (Fig. S10a) main peaks at 367.8 and 373.7 eV constitute for 3d5/2 and 3d3/2 of Ag2O in Ag+ state [26]. The presence of AgO cannot be rejected or confirmed owing to its high instability under ultra-high vacuum [27]. Al 2p spectra, given in Fig. S10c contained 2p3/2 peak at 74.5 eV assigned to Al3+ in mesoporous alumina. XPS peak appeared at 78 eV is attributed to anhydride Al2O3 [28]. The doping of metal into alumina lattice can affect the chemical states of Al and O, and the BE values will be changed [29]. O 1 s core level XPS spectra recorded is shown in Fig. S10d for the samples before and after the reaction. A major broad peak at lower BE value (530 eV) represents the lattice oxygen collectively contributed from all metal oxides. The surface Al-O species will be present at around 531.4 eV. Subsurface O species form silver may also contribute towards this peak [30]. The isolated –OH groups give rise to a peak at 532.7 eV. It is well known that any chemisorbed oxygen species like water molecules appear at higher BE. These values fall in slightly higher regions compared to the reported data. A close inspection of Fe 2p core-level spectra of fresh AgMesoFeAl in Fig. S10b hints the presence of 2p3/2 of Fe3+ state at BE, 712 eV accompanied by satellite peak located at 720 eV [31]. A small peak shift observed indicate that the catalyst surface is reduced after the reaction.Silver oxide can be reduced to metallic silver under the high-temperature dehydrogenation atmosphere in the presence of isobutane. Probably the Ag0 would drive the reaction. However, no peaks correspond to Ag metal are detected in XPS may be due to its high susceptibility for atmospheric oxidation. Interestingly, a well-known phenomena called voltage induced differential charging [32] might happened in the core level spectra of the studied samples. Consequently, distortion of peaks and shift of the measured position of the peaks corresponding to the core levels could occur. This is usually observed on the non-uniformly conducting sample surface [32]. The splitting of peaks observed in the current study may be attributed to this phenomenon. The catalytic activity is not directly correlated to the XPS analysis although gives the idea about the surface composition as well as oxidation states of each element in the catalyst.Post reaction analysis of the catalysts was carried out to discover the structural and chemical changes that occurred during the high-temperature isobutene synthesis. The X-ray diffraction patterns of all catalysts after the reaction were collected and given in Fig. S11. The crystallinity of the samples remained intact even after the reaction at 600 °C. Crystalline phases are analyzed after 6 h as well as 60 h of DHsioB reaction at the optimized reaction temperature and presented in Fig. 5 .TEM images of spent AgMesoFeAl catalyst are recorded to study the changes that occurred in the catalyst morphology. The material retained sheet-like mesoporous structure even after 6 h of dehydrogenation reaction at 600 °C with a small increment in particle size to 7.7 nm. It is clearly seen as slightly bigger particles in the TEM image represented in Fig. S4. However, after 60 h of TOS study, more agglomerated metal particles are observed (Fig. S4), and the average particle size increases to 16.9 nm. The SAED pattern given in the inset corroborated the separation of crystalline metal oxides and is also distinctly perceived from the XRD pattern of AgMesoFeAl after 60 h as presented in Fig. 5. Moreover, thermogravimetric analysis performed in the presence of air quantified 14.4% coke deposition on the catalyst after on-stream analysis. It is observed as carbon nanotubes in the TEM image depicted in Fig. S4 after 60 h reaction. Close analysis of the TEM images illustrates some of the active metal particles are entrapped in the nanotubes, which would also be a reason for the activity decline. The Raman spectrum of this sample shown in Fig. S12 confirms the formation of both sp2 and sp3 hybridized carbon formed during the high-temperature dehydrogenation process. This coke formation mainly leads to slight deactivation of AgMesoFeAl over 60 h of stability test. However, the material maintained mesoporosity throughout the reaction.Coke formation is one of the chief concerns involved in the isobutane dehydrogenation at high temperatures. The type of coke formed and the blocking of active sites by coke deposition can cause severe catalyst deactivation. Hence, AgMesoFeAl spent catalyst after 60 h of on stream study was investigated with Raman spectroscopy to study the nature of coke. The spectrum given in Fig. S12 proved the presence of D and G bands at 1330 and 1590 cm−1, respectively which are attributed to the disordered and pristine graphene bands. In addition to this, peaks correspond to Fe2O3 are also emerged, which were not visible in the fresh catalyst. These observations could be correlated to PXRD and TEM results evidencing phase separation as well as carbon deposition happening on the catalyst surface that leads to the deactivation during TOS analysis.Since carbon formation is observed as one of the main reason for catalyst deactivation, the quantification of coke deposited over the spent catalyst becomes inevitable. Therefore, thermogravimetric analysis was carried out for all spent catalyst and depicted in Fig. S13. The thermogram of AgMesoFeAl-TOS after on stream DHsioB result is also represented. The amount of coke for each catalyst was quantified and tabulated in Table 1. The pristine support MesoAl exhibits the least coke deposition, which might be due to the lowest isobutane onversion and less C-C cracking reactions occurred during the dehydrogenation reaction. The table clearly shows PMesoFeAl (12%), as well as AgMesoFeAl (12.7%), have lower coke deposits compared to other promoted catalysts. The maximum C-C cracking occurred on FeMesoAl (26.6%), which is higher than both AgMesoFeAl after 60 h on stream study (14.4%) and MesoFeAl (16.2%). This observation illustrates the importance of in situ Fe doping in the mesoporous alumina. Moreover, the results obtained from TG analysis can be directly correlated to the dehydrogenation reaction trend. Coke formation continues even after the initial period which is evidenced from the thermogravimetric analysis of spent catalyst after 6 h (12.7%) and 60 h (14.4%). But the coke deposition process is very slow. This might be a reason for the stable activity of AgMesoFeAl after the initial period. In addition, FeMesoAl exhibited a meager isobutene yield due to the cracking of formed alkene under severe reaction conditions. Therefore, maximum carbon deposition was observed for this sample. Additionally, sometimes carbon deposition can also contribute to maintain the alkene selectivity through deactivating the extreme active sites. McGregor et al. have studied butane dehydrogenation over VOx/Al2O3 and observed coke encapsulation over the catalyst after the reaction at 700 °C. Unlike typical catalyst deactivation, in this case the catalytic activity remained the same indicating that carbon species can catalyze the reaction [33]. Appreciable activity towards carbon-based systems are also reported for the catalytic dehydrogenation of propane by Liu et al [34]. Interestingly, nitrogen-doped carbon nanotubes exhibited efficient isobutene yield for the conversion of isobutane without any oxidant [35]. These results can be correlated to the current study and the possibility of carbon nanotubes to catalyze the dehydrogenation process cannot be ruled out. This could also be a reason for the stable activity of AgMesoFeAl in the TOS analysis.In summary, mesoporous alumina-based catalysts synthesized by the sol-gel method along with the material characterization correlated to their activity performance for non-oxidative dehydrogenation delivered essential insights to the synthesis of isobutene. The overall yield of isobutene over MesoAl was improved after metal incorporation. The proper doping of iron in the alumina matrix and the optimum surface acidity are viewed as the chief properties to inhibit cracking reactions and thus affecting isobutene selectivity. AgMesoFeAl was selected for further studies as a representative of the metal incorporated material. The catalyst surface was reduced after the reaction. Particle agglomeration followed by carbon nanotube development occurred over the continuous gas phase high-temperature dehydrogenation reaction, and isobutene formation are slightly hindered. To conclude, the promoted materials exhibited nearly the same activity and can be a promising catalyst for the selective dehydrogenation of isobutene. AgMesoFeAl catalyst provided a stable yield of isobutene during the long time reaction under the drastic reaction conditions with low space velocity.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 thank the Council of Scientific and Industrial Research (CSIR), New Delhi, India, for fellowship. Financial support from Mission Mode Project (HCP-0009) on catalytic methodologies for dimethyl ether funded by CSIR, India, and GAP-324426 project funded by the Department of Science and Technology-Science and Engineering Research Board (DST-SERB), India are greatly acknowledged. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2020.106263.

Production of isobutene is commercially consequential and highly demanding from the end-use industries being a key platform molecule as well as an intermediate for a variety of value-added chemicals. Traditionally, isobutene is prepared via steam cracking and fluid catalytic cracking methods. However, the catalysts used in these conventional methods have disadvantages like coke formation, sintering, etc. In this study, the catalytic non-oxidative dehydrogenation of isobutane over acidic, alkaline, and noble metal promoted mesoporous iron-doped catalysts was investigated. Iron doping has a significant function in controlling isobutene selectivity. The synthesis method is crucial to achieve successful metal doping in the mesoporous alumina matrix. Promoted catalysts exhibited a notable difference in isobutane conversion with a marginal change in dehydrogenation selectivity. Silver promoted catalyst showed slightly higher isobutene yield due to the optimal catalytic properties. Thiscatalyst was stable for a considerable duration, and coke deposition, as well as particle agglomeration, were observed to faintly inhibit the catalytic activity.

In the past centuries, human society has made tremendous technological and economical progress, which were accompanied by the rapid growth in world population [1]. The resulting challenge of fulfilling the increasing demand in food supply has always been a universal concern. As an efficient solution to this problem, intensive agriculture strongly relies on the industrial manufacturing of synthetic fertilizers, which require a large-scale supply of ammonia as feedstock. Fritz Haber [2–4] demonstrated the possibility of a catalyzed synthesis of ammonia from hydrogen and nitrogen under high temperature and pressure. Later Carl Bosch realized the industrial scale production of ammonia from its elements through the Haber–Bosch process [5,6].Although many catalyst systems have been proven to be active for ammonia synthesis [7,8], only two of them showed a potential high enough for application in the industrial production of ammonia. One of these catalyst systems is the promoted Ru-based catalyst supported on carbon materials [9–11] or metal oxides [12–14]. However, the high manufacturing costs make Ru-based catalysts hard to compete with the other catalyst system [15,16], i.e. the industrially widely applied iron catalyst found by Alwin Mittasch. The catalyst precursor of the iron-based catalysts is routinely prepared by fusing iron oxides with other oxide additives consisting of “structural” promoters like Al2O3 and CaO, and “electronic” promoters like K2O [17–20]. Prior to ammonia synthesis, the fused iron oxide, which contains additional oxide promoters, is activated by a multistep-reduction process in a H2/N2 mixture [21–23]. The iron oxide phase in the catalyst precursor has proven to be a crucial factor that influences the phase composition and the morphology of the reduced catalyst and determines the performance of the catalyst in ammonia synthesis [24,25]. Besides the conventional fused magnetite, wuestite has shown promising potential as an alternative catalyst precursor [25–28]. In the late 1980s, Liu et al. [27,29–32] have reported that catalysts based on non-stoichiometric wuestite Fe1-xO as precursor show superior activity in ammonia synthesis compared to the magnetite-based catalysts. Please note, that although magnetite and wuestite are the typical industrial oxide precursors for Fe-based ammonia synthesis catalysts, Fe2O3-derived ammonia synthesis catalysts are widely applied in academic research [33–35].In addition to the iron oxide precursor, the oxide promoters also play a decisive role not only in the formation of the active catalyst but also on the performance. Alumina is considered as one of the important structural promoters for ammonia synthesis catalyst [18,36]. For magnetite-based catalysts it is reported, that alumina is present in the reduced catalyst in the form of a thin Al2O3 layer [37,38] on the surface or segregated as both, FeAl2O4 and Al2O3 [39]. The presence of alumina or/and a Fe-Al solid solution [40] can prevent the activated iron from sintering, which would otherwise lead to the formation of larger crystals [18]. In the wuestite precursor, FeAl2O4 can also be formed via the solid reaction between FeO and Al2O3 [41]. However, due to the different crystal structure of FeO and FeAl2O4, the distribution of Al2O3 in the wuestite precursor is not as uniform as in magnetite precursors [25]. Consequently, it is assumed that alumina is not the only structural promoter and other oxide promoters such as SiO2 and ZrO2 are required [25]. Nevertheless, for wuestite-based precursor alumina is believed to participate in the restructuring of the surface of the reduced catalyst [25,42,43]. CaO is another structural promoter for catalysts based on magnetite as precursor [44,45]. In the reduced catalyst CaO segregates to the space between the Fe crystallites [46,47]. Furthermore, CaO increases the surface area and activity of the activated catalyst as well as promotes its resistance against impurities in the reactant gas [20]. For the wuestite precursor, CaO additionally inhibits the disproportionation reaction at low temperature of Fe1-xO which would form magnetite and metallic iron [25].Besides the structural promoters, K2O is the most important electronic promoter for ammonia synthesis catalysts based on both, magnetite or wuestite precursors [18,25]. Here, an enrichment of potassium on the surface during the reduction of the iron oxide precursors is taking place [25,46,48]. K2O is hydrolyzed to strongly basic potassium hydroxide during the reduction, which enables the formation of amphoteric metal oxohydroxides with alumina and iron oxide [20]. The basic iron oxides act as a binder to the other oxide promoters as well as positively influence the reduction kinetics. In the activated catalyst K exists in the form of anhydrous KOH rather than a metallic adsorbate [20]. The promoting effect is attributed to the ability of the active anhydrous KOH species to enhance the dissociative adsorption energy of nitrogen, which is described as rate determining step of ammonia synthesis [20,49]. Furthermore, the basic KOH species reduces the adsorption energy of ammonia and, therefore, prevents the catalyst from self-poisoning caused by adsorption of the formed ammonia, especially at high reaction pressures.In the present work, we focus on wuestite-based ammonia synthesis catalysts. To investigate the effect of different promoters on the performance of the catalyst, a series of wuestite-based catalysts containing different oxide promoters were synthesized, characterized and tested in ammonia synthesis. An industrial wuestite-based catalyst was used as reference. As we will show, disproportionation is the dominant chemical feature during the reductive activation at lower pressures and it can be strongly influenced by the use of promoters. Furthermore, the promoters as well as elevated pressure let all reductive events collapse into one reduction process that retains the phase disposition generated during fusion. This allows for the complete reduction of wuestite and the formation of “ammonia iron” below 500 °C.The catalyst samples (Table 1 ) were prepared according to the recipe of the applied industrial catalyst in a lab-scale electric arc furnace. Raw materials were mixed together as fine powders, and the mixture was placed inside a melting pot. The furnace chamber was evacuated and the synthetic air pressure was set to the desired value. A voltage was applied, creating an electric arc. When the melting process was finished, the melt was cooled down and crushed by jaw crusher. Finally, the granules were sieved in order to obtain the desired size fraction.For the TEM investigation, the wuestite grains were crushed and the resulting powders were dispersed in ultrapure ethanol and sonicated for 5 min. The experiments were performed with a transmission electron microscope (TEM) JEOL ARM 200 F operating at 200 kV, equipped with a double spherical aberration correctors, and GATAN Oneview and Orius cameras. For scanning TEM (STEM) studies, a high‐angle annular dark‐field (HAADF) detector was used, which maximized the collection of incoherent scattered electrons.SEM investigations were conducted using Hitachi microscope, operating at 15 kV, and which is equipped with a secondary ion and electron detectors.The BET surface areas and BJH pore size distributions were determined by measuring N2-physisorption isotherms at −196 °C with a Quantachrome QUADRASORB evo MP set-up. For regular measurements of the air-stable precursors the sample surfaces were cleaned from water and other potential adsorbates by degassing them at 100 °C for 12 h in vacuum. The air sensitive catalysts after reduction were transferred without air contact and directly measured without any thermal pretreatment. BET surface areas were calculated from data collected in a p/p0 range between 0.05 and 0.3. Adsorption and desorption isotherm were measured at a p/p0 range between 0.05 and 0.95 and used for the determination of BJH pore size distributions.The quasi in situ XRD data were collected in Bragg-Brentano geometry using a STOE Theta/theta X-ray diffractometer (CuK α1+2 radiation, secondary graphite monochromator, scintillation counter) equipped with an Anton Paar XRK 900 in situ reactor chamber. The samples were reduced in the in situ chamber with a heating rate of 3 °C min−1 until the desired target temperature was reached, followed by rapid cooling (20 °C min−1) and XRD measurement at 25 °C. Subsequently, the sample was heated again with 20 °C min−1 until reaching the previous target temperature, where the original TPR ramp of 3 K min−1 was resumed until the final reduction temperature of 850 °C was reached. The gas feed was mixed by means of Bronkhorst mass flow controllers, using 20 % H2 in helium at a total flow rate of 100 N mL min−1. The effluent gas composition was monitored with a Pfeiffer OmniStar quadrupole mass spectrometer. Ex situ XRD measurements of post mortem samples were performed in Bragg-Brentano geometry on a Bruker AXS D8 Advance II theta/theta diffractometer, using Ni filtered CuK α1+2 radiation and a position sensitive energy dispersive LynxEye silicon strip detector. The diffraction patterns were analyzed by whole powder pattern fitting using the TOPAS software (version 5, ©1999−2014 Bruker AXS).Temperature-programmed reduction (TPR) experiments were performed in a custom-designed set-up equipped with stainless-steel tubes, a fixed bed reactor (quartz glass, U-tube) and an on-line thermal conductivity detector (TCD) for monitoring the H2 consumption. The TCD (Emerson X-stream) was calibrated by reducing a known amount of CuO. A molecular sieve containing tube was installed ahead of the detector as water trap.For a measurement 100 mg of catalyst precursor (particle fraction 250−425 μm) were reduced by heating it to 900 °C in a total gas flow of 75 N mL min−1 (20 % H2, 80 % Ar) applying a linear heating rate of 3 °C min−1. The Monti-Baker criterion was in a range of 100–125 depending on the sample [50]. The Mallet-Caballero criterion was in a range of 5–6.5 K depending on the sample [51].The ammonia synthesis tests were conducted in a commercial all stainless-steel flow set-up (Integrated Lab Solutions Gmbh) equipped with a guard reactor, a synthesis reactor, and an on-line IR-detector for NH3 and H2O (Emerson X-stream) for quantitative product gas analysis. For a detailed description of the set-up see [52].For a measurement 3 g of Fe-based precursor (particle fraction 425−560 μm) were diluted with 3.9 g SiC (average particle size 154 μm). The catalyst bed was placed in the synthesis reactor between pure SiC and held in position by glass wool plugs at the entrance and exhaust of the reactor. After intensive purging of the reactor until water content was stable at almost zero, the sample was reduced by heating it in a gas flow of 858 N mL min−1 (75 % H2, 25 % N2) with a temperature program up to 500 °C at an elevated pressure of 30 bar. The precursor was reduced in three temperature steps with different heating rates: from room temperature to 250 °C with 1.2 °C min−1, 250−400 °C with 0.3 °C min−1 and 400−500 °C with 0.2 °C min−1. Afterwards, the conditions were kept constant for ca. 4.5 h. In total the reduction procedure took 24 h.For catalytic testing, the temperature was kept at 500 °C while the total gas flow was adjusted to 357 N mL min−1 (75 % H2, 25 % N2). The pressure was increased from 30 bar up to 90 bar in three steps of 20 bar. Each step was performed with a pressure ramp of 1 bar min−1 (1 h per step) and after reaching the elevated pressure it was kept constant for 40 min before starting the next step. After reaching 90 bar, the temperature was reduced to 400 °C with a rate of 1 °C min−1 and kept constant for 22 h. Afterwards the catalyst was heated again with 1 °C min−1 to 500 °C and measured for 14 h before cooling it back to 400 °C and measuring it for 22 h. This was repeated in total two times. At the end of the measurement the pressure was released and the catalyst was cooled down to room temperature, while the reactor was flushed with nitrogen. The tested sample was removed inside of a glovebox to allow further characterization of the catalyst in a reduced form. The activity of the catalysts is given as relative NH3 synthesis activity, where the effluent mole fraction of NH3 of the different catalysts was normalized to the initial effluent mole fraction of NH3 of the industrial catalyst FeO-04.Four different Fe1-xO-based precursors were investigated (Table 1) including three laboratory produced samples (FeO-01, FeO-02, FeO-03) and one industrially applied catalyst (FeO-04). The samples differ in their degree of promotion (Table 1). One laboratory sample (FeO-01) is unpromoted. FeO-02 is promoted with K and Al which reflect the most common promotors for all Fe-based ammonia synthesis catalysts [18]. FeO-03 has the same K and Al content as FeO-02 and is additionally doped with Ca as it is known to be one of the most important promoters for wuestite-based precursors [25]. The industrial FeO-based catalyst (FeO-04) contains a package of different promoters that are present in different amounts. The amounts of K, Al and Ca in FeO-04 are similar to FeO-03.The precursors were activated by a reductive pretreatment as described in the experimental section to form the actual catalyst. The reduction of the precursors is accompanied by water formation and initiates the production of ammonia (Figure S1). Following the reductive activation, the catalytic activity towards ammonia synthesis was measured at a pressure of 90 bar at two different temperatures (400 °C and 500 °C) (Figure S2).The reduction behavior of the Fe1-xO precursors varies strongly with the applied reduction conditions (Fig. 1 a). Under the pressure of 30 bar the samples exhibit quite similar reduction profiles. The peak shape during reduction is asymmetric for all catalysts indicating overlaying reduction steps and/or higher order kinetics. Furthermore, the rate maximum is shifted to higher temperatures with increasing degree of promotion. Temperature programmed reduction (TPR) analysis of these catalysts at atmospheric pressure reveals kinetic resolution and a splitting of the reduction profile and for the catalysts with lower degree of promotion intermediate phases between reduction steps become stable. Due to the lower pressure and the lower H2 amount in the gas phase the reduction potential is lower compared to the activation procedure at higher pressures as evidenced by the TPR measurements. In addition, the position of all reduction peaks is shifted to higher temperatures at lower hydrogen partial and total pressures. Although for pure wuestite only one reduction signal would be expected (Fe1-xO → Fe), the reduction profile of the unpromoted sample FeO-01 is split into three overlapping peaks. This arises from the consecutive reduction of wuestite and the disproportionation products. With increasing promotion the peak splitting decreases. While FeO-02 that is doped with K and Al shows two overlapping peaks, FeO-03 and FeO-04 exhibit only one visible reduction peak. This observation highlights that phase formation during reduction is strongly affected by the promoters acting on the catalysts synthesis as well as on the ammonia synthesis.As expected, the comparison of the catalytic activity of the samples that is presented in Fig. 1b displays an enhancement of the activity with an increasing degree of promotion. The addition of K and Al (FeO-02) almost doubles the activity of the wuestite-based catalyst in comparison to the unpromoted sample (FeO-01). A further significant activity boost is achieved by the additional presence of Ca (FeO-03), which leads to an even three times higher activity compared to the unpromoted sample. With the addition of several different promoters the multi-promoted industrial catalyst (FeO-04) exhibits by far the highest activity and still leaves a significant gap to the sample with 3 promoters (FeO-03). The massive influence of promotion is especially visible by comparing the multi-promoted industrial catalyst (FeO-04) to the unpromoted sample (FeO-01). The multi-promoted Fe1-xO-based catalyst is more than 5 times as active as the unpromoted analogue.A major contribution to this increase in activity can be assigned to structural promotion as can be seen by the BET surface areas and mesoporous pore volumes of the reduced and tested catalysts (Fig. 1b, Table S1). The addition of K and Al (FeO-02) and the subsequent introduction of Ca (FeO-03) leads to an increasing surface area and pore volume quite comparable to the increase of the overall catalytic activity. It should be noted that this behavior does not exclude an influence of other promoting effects like the improvement of the reaction kinetics, where K is known to be important. In general, Al and Ca are known as structural promoters, which preserve the Fe nanostructures from sintering [27,32]. Furthermore, they cause an increase of the surface area of the catalyst and, therefore, an increase of the total number of active centers. By comparing the BJH pore size distribution of the samples after ammonia synthesis the difference in the effect of structural promotion becomes visible (Figure S3). While FeO-01 exhibits only little mesoporosity the addition of K and Al (FeO-02) leads to the formation of a small mesoporous pore volume with a broad pore size distribution. A significant effect can be seen by the addition of Ca (FeO-03). A clear mesoporous structure can be observed with a defined maximum centered around 15 nm. The pore size distributions of FeO-03 and FeO-04 are almost identical. Thus, Ca seems to have a major role as a structural promoter and leads to the formation of a mesoporous network within high performance ammonia synthesis catalysts. Besides structural promoting, the addition of further promoters in FeO-04 may have a primary effect on the ammonia reaction kinetics, while the combination of K, Al and Ca in FeO-03 are believed to be the major constituents that lead to the total surface area and mesoporous nanostructure of the industrial catalyst.As mentioned before the four samples vary in their reduction behavior during the TPR at 1 bar. In order to understand associated changes of the iron phases, XRD measurements were performed before the reduction, after the reduction and in the middle of the reduction process at 500 °C in a hydrogen-containing atmosphere. Due to the low time resolution (ca. 20 h per scan), all XRD measurements were performed at room temperature to avoid ongoing reduction of the sample during the data collection, which renders this technique quasi in situ. The intermediate target temperature of 500 °C was chosen according to the minima of the TPR profile of the sample FeO-01, while 850 °C was the maximum temperature accessible with the setup. At the minima of the TPR profile, the reduction rate is the lowest and thus the necessary interim cooling/re-heating phases should have the smallest possible impact on the reduction profile. Furthermore, it may be expected that potential intermediate phases have their maximum concentrations at these points. Although the minima are less resolved for FeO-02 or no TPR minima could be found for FeO-03, the same temperature program was applied to all samples for the sake of comparability.Before reduction, the XRD patterns of all samples exhibit a distinct wuestite (Fe1-xO) phase with a small amount of α-Fe, while after reduction only an α-Fe phase is present for all samples (Fig. 2 ). A difference for the samples can be observed with the XRD patterns in the middle of the reduction progress at 500 °C. It is possible to see the formation of a magnetite (Fe3O4) phase in different amounts for all samples before iron is fully reduced at higher temperatures. It should be noted that the XRD pattern of sample FeO-03 after reduction to 500 °C is peculiar in showing an unusually strong 220 reflection (60.5°) of the wuestite phase. Peak shape and broadening exclude the possibility of bad sampling statistics, which could cause significant intensity deviations in the case of highly crystalline phases. Furthermore, the cubic crystallographic symmetry and the lack of directing mechanical forces during the experiment rule out preferred orientation effects as a possible explanation. Thus, we interpret this surprising change of the relative intensities as a true structural effect. A Rietveld refinement was obtained after allowing the occupation of tetrahedral interstitial sites in the wuestite crystal structure by Fe atoms [53]. In the refined model, about one third of the iron atoms resided on the new tetrahedral positions, while the rest occupied the normal octahedral sites. Whether this unusual, modified wuestite phase is directly stabilized by the promoters in FeO-03, or whether it occurs generally as an intermediate during reduction and was only accumulated into noticeable amounts due to the delayed disproportionation/reduction kinetics, remains open.The formation of magnetite is caused by thermal disproportionation of the wuestite into α-iron and magnetite. When the amounts of the different iron oxide phases are compared (Table 2 ) it can be seen that the unpromoted sample FeO-01 exhibits a much higher amount of magnetite at 500 °C compared to the two other promoted samples at this temperature. This shows that the promoters are stabilizing the metastable wuestite phase and inhibit thermal disproportionation (Fig. 2), i.e. minimizing the amount of magnetite at the final reduction temperature such that the individual events occur at lower temperature and coincide.The thermal disproportionation can explain the peak splitting of the reduction peaks, which is strong in the unpromoted sample. Due to the disproportionation into α-Fe and Fe3O4 the precursor turns into phase mixtures of Fe1-xO, α-Fe and Fe3O4. This leads to a reaction network during reductive activation (Fig. 3 ). Hence, parts of the Fe1-xO phase are directly reduced, while other parts of this phase disproportionate into iron and magnetite. The newly formed Fe3O4 is also reduced at higher temperatures compared to the original wuestite phase. It can be speculated that Fe3O4 forms during its reduction a new Fe1-xO phase as an intermediate step, which is subsequently reduced or disproportionated. It is further possible that Fe1-xO phases with different x values are formed. This could also explain the change of their reduction rate. At the applied higher temperatures all reactions can happen simultaneously as indicated by the overlaying TPR profiles.Peak splitting in TPR decreases with increasing degree of promotion until only one reduction peak is present. However, disproportionation can still not be fully excluded as even the most promoted model sample FeO-03 still exhibits a detectable amount of magnetite as shown by the XRD patterns in Fig. 2f. Thus, the single reduction peak of the FeO-03 precursor can result from an overlap of different reduction events.Please note, the quasi in situ XRD analysis was limited to a maximum measurement angle of 90° 2θ. Due to this limitation only three accessible reflections that can be assigned to bcc Fe could be obtained. The absence of further reflections has limited the peak profile analysis to consider only isotropic size broadening. Nonetheless, this analysis hints to a slight anisotropy in the peak profiles corresponding to α-Fe (Fig. 4 ) [54]. The peak width misfit is most pronounced for the 200 reflection. As opposed to the quasi in situ analysis, this effect of anisotropic peak broadening is more obvious in the diffraction patterns that were measured of all FeO catalysts after ammonia synthesis. However, these catalysts have been investigated ex situ. To illustrate the effect of anisotropic peak broadening, the XRD results of FeO-03 after ammonia synthesis are presented as a structural example in Figure S4a. The data was fitted with an isotropic profile in the absence of a crystal structure model. Furthermore, a fit that includes a crystal structure model (i.e. Rietveld refinement), reveals an additional mismatch of the calculated relative intensities for the spent FeO-03 catalysts (Figure S4b). Since the bcc crystal structure of α-Fe has no internal degrees of freedom, i.e. it accommodates only one atomic site that is fixed to a special position, except the thermal displacement parameter, this result can be taken as evidence that the real structure of the iron phase must be more complex than simple α-Fe. Phenomenologically, the observed intensity distribution could be approximated either by allowing the thermal displacement parameter to take physically implausible negative values or by assuming that additional electron density is residing on interstitial positions. It should be noted here that diffractometer misalignment or beam spill effects were explicitly ruled out as potential causes for the observed intensity mismatch.From the applied Rietveld analysis of the experimental data of the wuestite precursors that have been measured by the quasi in situ XRD approach lattice parameters a and domain size values L Vol-IB were extracted which are presented in Table 3 . However, due to limitations to the isotropic fit model, these values should be interpreted only in terms of a trend rather than absolute values. The domain size values suggest that the addition of K, Al reduce sintering, while the addition of Ca seems to increase the Fe lattice parameter after reduction at 500 °C. It should be noted that the peculiarity of the lattice parameter occurs in the same scan as the “interstitial wuestite” phase occurred and vanishes with higher reduction temperatures.These different possible reduction pathways render any detailed analysis difficult. While the formation of magnetite can only originate from the disproportionation of wuestite, the α-Fe can come from the disproportionation of wuestite as well as from the reduction of magnetite. In addition, it is possible that the full amount of all Fe phases cannot be detected by XRD due to the absence of translational symmetry or the formation of too small crystalline domains. For example, for sample FeO-01 (4 FeO → Fe3O4 + Fe) that exhibits a large amount of magnetite a α-Fe to magnetite weight-% ratio of 1–4 it could be stoichiometrically expected. However, the actual ratio is only around 1–9. This indicates that probably not all α-Fe in the sample is detected. Despite these uncertainties it is still possible to conclude on a few trends, in particular, when the amount of phases is strongly changing as it does for the aforementioned amount of magnetite. It is also possible to observe that the FeO-02 sample is at 500 °C more reduced indicated by a higher amount of α-Fe compared to FeO-03 sample. This can be explained by the presence of Ca in the FeO-03 sample, which slows down the reduction process of wuestite [55]. CaO crystallizes in the same crystal structure than wuestite and, therefore, it can be incorporated into the wuestite lattice [25]. This could form CaxFe3-xO4 solid solutions, which inhibits the disproportionation process. This is in agreement with the findings of Li et al. [55], who have shown the positive effect of CaO on thermal stability of wuestite structure under base pressure of 1.33 Pa. Meanwhile, the formation of CaxFe3-xO4 species will hinder the reduction rate in comparison to the wuestite precursor without Ca. A similar hindering effect of Ca on the magnetite-based precursor reduction has been reported by Liu et al. [32]. As the Ca content is sufficiently low the precipitation of the spinel phase could be avoided. It has been highlighted that phase purity of the oxide precursor is a necessary requirement for a good ammonia synthesis catalysts [25].To investigate how the early stages of disproportionation influences the structure on the local scale transmission electron microscopy investigations were conducted. Transmission electron microscopy allows to investigate the morphology and structure of the catalysts at the (sub)-nanometric scale and is thus complementary to bulk averaging characterization techniques. In order to study the morphological and structural changes that occur during the reduction on the individual particles, quasi in situ TEM experiments were conducted [56,57]. This later allows to investigate catalyst particles before and after activation at elevated pressure and temperature, which are relevant for ammonia synthesis catalysts.The reduction of the wuestite was first followed for the FeO-03 precursor to exemplify the presence of local disproportionation events for a fully promoted sample by submitting the sample to a H2/Ar mixture of 3 to 1 at 10 bar and 365 °C. Fig. 5 shows typical and identical wuestite particles before (Fig. 5a) and after the quasi in situ experiment (Fig. 5b). Before quasi in situ activation (Fig. 5a), surface near diffraction contrast indicates the formation of defective structures within the particles which are characteristic of wuestite-type materials. After reductive treatment (Fig. 5b), a cracking of the identical particle occurred that can be best described by the formation of a hedgehog-like structure, which is accompanied by the formation of porosity and the outgrowth of multiple nanoplatelets [58,59]. High-resolution (HR)-TEM imaging and corresponding Fast-Fourier transform (FFT) analysis (Fig. 5c-d) of the nanoplatelets denote the growth of polycrystalline particles with a nanoparticular structure. After treatment at 365 °C the wuestite structure is still present in the sample. In addition, α-iron and magnetite are formed. Thus, this observation corroborates the thermal disproportionation in the early stage of the reduction process, as previously established by XRD and TPR even for the fully promoted samples.Additional areas of the FeO-03 sample after reduction were investigated, which show the formation of complex structures and the local inhomogeneity of the sample. Fig. 6 presents one example of a detailed HRTEM analysis. Polycrystalline particles are observed including the presence of Fe3O4 (Fig. 6b–d) and Fe (Fig. 6e–g) phases. In addition, Moiré patterns (Fig. 6a and e) as a result of overlapping lattice planes indicate the presence of turbostratic disordered layers, which can also highlight the formation of defective phases. The complexity of the atomic structure is further corroborated by analyzing the surface layer of such a particular aggregate, including amorphous layers (Fig. 6h), short-range ordered surface structures (dashed ellipse in Fig. 6i) and surface roughness (Fig. 6i). The observation of amorphous and short-range ordered phase may also be in line with the missing Fe content calculated from Rietveld analyzed XRD patterns. Elemental analysis indicates that the promoters (Al, K, Ca) are in close contact with iron phases (Figure S5).From the above results, it is apparent that the temperature of 365 °C is insufficient to fully reduce the wuestite precursor even at elevated pressure. In order to corroborate the TPR results showing that at elevated pressure and higher temperature (above 400 °C) also the unpromoted sample can be fully reduced quasi in situ investigations of FeO-01 were conducted. Fig. 7 shows TEM images on the FeO-01 sample before (Fig. 7a) and after (Fig. 7b) quasi in situ reduction experiment at 10 bar of H2/N2 (3/1) mixture at 470 °C showing that at elevated pressure the unpromoted sample is fully reduced even at the local scale. The wuestite precursor (Fig. 7a) indicates a particle-like morphology with a length of about 500 nm. Similar to the pristine FeO-03 sample, localized diffraction contrast is observed which is distributed all over the particle. After the exposure of the precursor to a H2/N2 (3/1) mixture at 10 bar and 470 °C (Fig. 7b), a drastic morphological change occurred which is expressed by the formation of elongated polycrystalline iron phases. This is confirmed by HRTEM imaging (Fig. 7c), which shows the presence of (011) α-Fe planes of different crystals. The HRTEM also shows the formation of Moiré patterns between the agglomerates of particles, as a result of the interference of overlapping (011) lattice planes indicating thin layers. Furthermore, the particles exhibit a dark contrast (see arrows in the Fig. 7c), which is indicative of the formation of strains on the particles.In addition, the structure of the sample before and after the quasi in situ reduction was investigated using selected area electron diffraction (SAED). Ring patterns were observed in both before (Fig. 7d) and after (Fig. 7e) quasi in situ reduction, indicating a polycrystalline structure with a small crystallite size. The analysis confirms that the full reduction of wuestite precursors to Fe occurs at elevated pressures even for the unpromoted sample at lower temperatures which corroborates the TPR results presented in Fig. 1a.It should be noted that we have conducted similar reduction experiments for FeO-02 by quasi in situ TEM (see Figure S7). As demonstrated for FeO-01 (Fig. 7) the sample is reduced and similar morphological and structural features were detected.Additionally, the microstructure of the FeO-03 sample before reduction (Fig. 8 a) as well as after reduction and ammonia synthesis (Fig. 8b-c) was investigated by scanning electron microscopy (SEM). A complex microstructure is observed with the presence of different grain sizes (Fig. 8a). Furthermore, grain boundaries can be observed which indicate the presence of a defective structure. The bulk morphology of the wuestite precursor is similar to the one after ammonia synthesis. However, a sponge-like structure is formed with voids distributed all along (see areas highlighted by arrows in Fig. 8c). This is in line with the BJH results, which showed a clear mesoporous structure for the wuestite sample promoted with K, Al and Ca. This may be related to the reduction of the particles, which generates the porosity and increases the surface area of the particles as confirmed by the BET surface area.In order to investigate the microstructure and chemical composition on the sponge-like structure formed after NH3 synthesis, SEM-EDX mapping were conducted (Fig. 8d–I and Figure S6). The EDX mapping has revealed the presence mainly of Fe. However, O, Al, K, Ca were also detected, and are more concentrated into the grain boundaries. Similar localization of the promoters in the wuestite grains before reduction were observed (see Figure S8). Therefore, the promoters could act as binders between the larger grains and thus help to stabilize the sample morphology during the reduction/activation steps.In summary, the reduction of wuestite-based precursors occurs in a complex reaction network involving reduction and disproportionation events. Our results show positive effects of K, Al and Ca as the main promoters on the performance and reduction behavior of wuestite-based ammonia synthesis catalysts. They have a significant influence on improving and stabilizing the catalyst nanostructure that is defined during the fusion and retained during the reduction of the precursors, whereas the promoters seem to be concentrated in the grain boundaries of the formed α-iron bulk crystals. Furthermore, they narrow the reduction and disproportionation events of metastable wuestite which allows a more direct reduction with less amounts of magnetite at the final activation temperature. The presented results support the formation of a structure of defective and disordered nanoplatelets within a mesoporous network as the origin of “ammonia iron” in comparison to bulk α-iron. This enhances by far the surface area of the catalyst. These global improvements from the promoters lead to a more active catalyst for ammonia synthesis. Especially Ca seems to play a major role on defining the reduction process and, therefore, also the mesoporous network and the surface area of the resulting catalyst. We note that the difference between normal iron and ammonia iron mostly occurs on a mesoscopic scale. Typical spectroscopies as Mössbauer or EXAFS would not detect such differences [60]. XRD is slightly sensitive in its line shapes [61] being peculiar in active catalysts. The information content of these anomalies precludes a distinction of mesoscopic defect from possibly present additional local defects for which TEM gave some hints. Jan Folke: Writing - original draft, Investigation, Validation, Formal analysis, Visualization. Kassiogé Dembélé: Writing - original draft, Investigation, Validation, Formal analysis, Visualization. Frank Girgsdies: Investigation, Formal analysis, Visualization, Writing - review & editing. Huiqing Song: Investigation, Validation, Formal analysis, Writing - original draft. Rene Eckert: Investigation, Writing - original draft, Writing - review & editing, Project administration. Stephan Reitmeier: Writing - review & editing, Resources, Supervision. Andreas Reitzmann: Writing - review & editing, Resources, Supervision. Robert Schlögl: Conceptualization, Methodology, Writing - review & editing, Resources, Supervision. Thomas Lunkenbein: Conceptualization, Methodology, Writing - original draft, Writing - review & editing. Holger Ruland: Conceptualization, Methodology, Writing - original draft, Writing - review & editing, Project administration.The authors declare no conflict of interest.The authors would like to thank the Max Planck Society for financial support. Furthermore, we thank Birgit Deckers for the graphical artwork.Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.cattod.2021.03.013.The following are Supplementary data to this article:

Ammonia synthesis remains one of the most important catalytic processes since it enables efficient hydrogen storage and provides the basis for the production of fertilizers. Herein, complementary bulk and local analytical techniques were combined to investigate the effect of selected promoters (Al, K, Ca) on the reduction of wuestite into α-iron and their catalytic performance for ammonia synthesis. The use of promoters appears to have a positive effect on the wuestite-derived catalyst in ammonia synthesis. The promoters seemingly act as a binder for wuestite grains and impede the reduction and disproportionation events of wuestite precursors resulting in an increased catalytic performance. This effect is associated with an increase of surface area and mesoporosity. The study delivers new insights into the interplay of structure and promoters in wuestite-based catalysts.

Data will be made available on request.The industrial revolution has made life very convenient by regularly upgrading the technologies, however, it’s still dependent on natural fuels such as natural gas, coal and petroleum [1–3]. On one hand, the combustion of these fuels causes serious environmental pollution, while on the other hand, its reservoirs are limited, which compels researchers to explore new energy sources [4,5]. In this regard, ideas have been projected to overcome the said problem and the solution seems to be in fuel cells, metal-air batteries, and alkaline water electrolysis [6]. Complicated modules have been designed, but their basics are quite simple, such as two electrodes assembly where oxygen evolution reaction (OER) takes place at the anode, while hydrogen evolution reaction (HER) at the cathode [7,8]. Sluggish kinetics is the big hurdle keeping away the process from practical application. OER is a four-electron transfer reaction, while HER is 2e− a transfer reaction, so OER needs a higher overpotential [9]. In the previous few decades, serious efforts have been made to design such a catalyst that enhances kinetics as well as stability in different media. Benchmark for OER is RuO2 / IrO2, which shows high activity in any media [10]. However it is highly unstable in acidic/basic media and oxidized to RuO4 / IrO3 and gets dissolved in the solution [11]. The comparative stability of IrO2 is higher than RuO2 but still both are precious and far away for large–scale usage [12]. Therefore, researchers investigated alternatives of RuO2 and IrO2. The possible candidates are those species that mimic the electronic environment of those species. The closer ones to those are transition metal oxides/hydroxides and doped transition metal oxides and hydroxides.Remarkable efforts have been made to study transition metal oxides (TMO) such as perovskite, spinel and layer structures having remarkable OER activity [13]. The easy synthetic approach, low capital cost, environment-friendly, high OER activity and stability in alkaline media make them an attractive choice for researchers. TMO has a rich blend of TM with variable oxidation state and coordination environment, which is fine-tuneable for OER [14]. Layer structures have also been given considerable attention, especially for their boosting activity upon doping. Recently, there has been noteworthy interest in 2D and layered 2D materials, particularly for their electrocatalytic application [15].Non–oxide electrocatalysts such as CoC, CoP, Co2N, Co2P, CoC2, Co3N, Co3S4, Co4N, Co9S8, Ni3N, Ni3C2 (C = Se, S and Te) and organometallics have also been reported for having high OER or HER activity and stability, but overall they are suffering from high over potential [16,17].Layer structures are usually transition metal hydroxide M(OH)2 and oxyhydroxides MOOH, and almost both have significant OER activity [18]. In a typical structure, TM is located in the center and oxygen anions at the corner of octahedrons forming [MO6] as a repeating subunit, while H+ is sandwiched between layers. Metal oxyhydroxides are of many kinds. Subbaraman et al. reported the same trend (Ni > Co > Fe > Mn) for LDH especially 3d TM such as Co, Mn, Ni, Fe hydr(oxy)oxide as in perovskite [19]. According to their analysis, the Ni–OH optimal bonding is the causative agent for high OER activity. Contrary to this statement, Corrigan et al. observed that introducing Fe and increasing Fe content boosted the OER performance of NiOOH. Trotochaud et al. also thoroughly studied Fe influence in NiOOH, and concluded that 25 % addition of Fe caused a 30-fold increase in its conductivity, which consequently enhanced its OER activity reflected from dropping overpotential to 200 mV. But still, the high robust activity cannot solely be explained by co-precipitation and a slight increase in conductivity [20]. Bode et al. proposed the oxidation of Ni+2 to Ni+3 by deprotonation during charging/discharging in basic media [21].Compiling all the reported data in the literature, it can be inferred that incorporating Fe into the NiOOH increases its OER activity because Fe acts as an active site for the reaction. In the current work, we incorporated different amounts of Fe such as 10, 20, 30 and 40 % and successfully doped it in NiOOH sheets. The electrochemical analysis indicates that the OER activity increases with an increase in doping amount and reaches to peak position upon 30 % doping and then drops down upon further incorporation of Fe. This may be explained based on exposed active sites to OER which are gradually covered upon further addition consequently giving a volcano plot.Nickel (II) acetate tetrahydrate. (C4H6NiO4·4H2O), Iron (III) nitrate nonahydrate (Fe(NO3)3·9H2O), Nafion® (5 wt% in lower aliphatic alcohols and water, contains 15–20 % water), isopropyl alcohol (C3H8O), ethanol (C2H5OH) and N,N-Dimethy formamide (C3H7NO) were bought from Sinopharm Chemical Reagent Co. Ltd. Throughout the course of experimental work the chemicals were used as collected without subjecting to any purification. The ultrapure Millipore water (18.2 MΩ) were used in all experiments.Typically, NixFexOOH sheets were prepared via a single-step approach. C4H6NiO4·4H2O and Fe (NO3)3·9H2O with different molar ratios were dissolved in a mixture of H2O and DMF and sonicated for 30 min then transferred to Teflon Lined Autoclave (TLA) and heated at 150  degreesC for 12 h. The prepared catalyst is cooled down at room temperature and washed several times with DI water. The synthetic detail is given in the Supplementary Table 1 (S1, ESI†). According to our knowledge, this synthetic route for NixFexOOH sheets is novel.The electrochemical analysis was made using an IM6 electrochemical workstation (Zahner, Germany) with a three-electrode system. The prepared NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH, Ni0.7Fe0.3OOH and Ni0.6Fe0.4OOH nanosheets/C were used as the working electrodes. Platinum wire was used as the counter electrode, and Ag/AgCl electrode was used as the reference electrode. All potentials were referenced to the reversible hydrogen electrode (RHE) through RHE calibration. Before any analysis, the electrolyte cell was purged with O2 gas for 30 min to keep a saturated vapour pressure. During analysis, the mild flow of O2 was continued to ensure H2O/O2 equilibrium at 1.23 vs RHE. Approximately 4 mg of the catalyst was dispersed in a mixture containing 0.75 mL of water, 0.25 mL of isopropanol alcohol, and 20 μL of 5 wt% Nafion (Aldrich). After ultrasonication for 1 h, suitable microliters of the ink were decorated on a glassy carbon rotating disk electrode (Pine Instruments). Polarization curves were recorded in 1 M KOH solution with a rotation rate of 1600 rpm and a scan rate of 5 mV s−1. Electrochemical impedance spectroscopy was performed over a frequency range from 100 kHz to 0.5 Hz with a sinusoidal voltage amplitude of 5 mV. Accelerated stability tests were conducted by applying cyclic sweeps between 1.2 and 1.7 V versus the RHE in 1 M KOH electrolyte at a scan rate of 100 mV s−1. Polarization curves were recorded at the end of cycling with a rotation rate of 1600 rpm and a scan rate of 5 mV s−1.Transmission electron micrographs (TEM) images were obtained by using Hitachi H-7650 instrument with an acceleration voltage of 100 kV. The composition of the samples was analyzed by inductively coupled plasma atomic emission spectrometry (ICP-AES, Atomscan Advantage, Thermo Jarrell Ash, USA). The surface valance band spectra were collected by high-resolution X-ray photoemission spectroscopy (XPS) on the nanosheet monolayers. These were recorded on an ESCALAB-250 spectrometer having a monochromatic Al Kα X-ray source (hν = 1486.6 eV), with a spot size of 500 μm. X-ray diffraction (XRD) characterization was performed using a Philips X’Pert Pro X-ray diffractometer with a monochromatized Cu Kα radiation source and a wavelength of 0.1542 nm.Typically, the NixFexOOH nanosheets were synthesized via a single-step novel approach, to the best of our knowledge, the method has not been reported yet. The whole process is schematically represented in Fig. 1 . The full detail is given in Table S1†. The nanosheet powder was vacuum dried and then dispersed in ethanol via ultrasonication. The nanosheet powder dispersion was dropped onto the copper grid coated with a carbon layer and air-dried for TEM measurements.Both TEM and SEM reveal that NiOOH nanosheets and Fe-doped sheets are of average size 400–500 nm Fig. 1-A, B, C, D and E. Additionally, the doping didn’t change the morphology as well as the sizes of the sheets. The retention of morphology is due to the fact that Ni − O and Fe − O are approximately the same and comparable to that of NiOOH as reported by Nilsson and Bill et al. in their computational study. STEM-energy dispersive X-ray (STEM-EDX) elemental mapping images of the Ni0.7Fe0.3OOH nanosheets clearly indicate that Ni, Fe and O are homogeneously distributed (Fig. 1-F). The composition of the samples was analysed by inductively coupled plasma atomic emission spectrometry, the data shows that the molar ratio of the prepared catalyst was 0.9:0.1, 0.8:0.2, 0.7:0.3 and 0.6:0.4 for Ni and Fe, respectively.The XPS data confirms the presence of Fe in NiOOH sheets. As shown in Fig. 1-G, H the peak at 856.1 and 873.7 refers to Ni 2P3/2 and 2P1/2 respectively, while the peak at 709.6 refers to Fe2p3/2. Zhenzhi Yin et al & Runze He et al finding is relatable with our analysis. [22,23] All the XPS peaks show that both Ni and Fe were in the oxidized form, the comparative XPS full range plot is given as Fig. S2 ESI†. XRD pattern of the samples (Fig. S3 ESI†) obtained using the standard procedure shows the crystallinity of the material before and after doping it also indicates clear peak shifting after Fe doping. Although the peak left shifting due to doping is ∼ 0.5 degree, it still cannot be ignored because it strengthens the incorporation of Fe to NiOOH sheets and does not undergo phase changes. It may act as a substituting agent which may also be confirmed from the TEM and SEM images. The representative SEM images and EDX of Ni0.7Fe0.3OOH are given in Fig. S4 ESI† [24,25].Before the electrochemical analysis, the prepared catalyst is dispersed in a mixture of isopropanol, water and Nafion. A typical mass ratio among the Nafion, catalyst and water is 1:1:1000. The mixture is then sonicated to get a uniform catalyst ink. The catalyst ink is then dropped on the glassy carbon electrode (GCE). The electrode is then subjected to get dried under an ambient environment. The electrocatalytic OER activity of the Ni0.7Fe0.3OOH nanosheets were analysed in 1.0 M KOH solution at room temperature using a standard three-electrode system. Before analysing the polarization curve the working electrode was stabilized by repeated cyclic voltammograms. Then the polarization curve was recorded (LSV). The OER activity of all the prepared catalysts was compared. Fig. 2 -A shows polarization curves recorded by linear sweep voltammetry (LSV) at a slow scan rate of 5 mV s−1. The ohmic potential drop (iR) losses that were created from the electrolyte solution resistance were all corrected before comparison.Interestingly, the Ni0.7Fe0.3OOH nanosheets revealed a much lower onset potential and higher current density relative to the NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH and Ni0.6Fe0.4OOH nanosheets. The first peak at 1.38 V versus RHE (observed for all catalysts) is attributed to the characteristic interconversion of Ni2+/Ni3+mediated by OH−. OER performance can be approached by checking the Tafel slope, which holds a paramount position. Generally, the smaller value of the Tafel slope indicating high OER activity provided that the data is at the same overpotential increment. The Tafel plot for Ni0.7Fe0.3OOH gave a value of 44.8 mV/decade (Fig. 2.B), which was much smaller than that determined for NiOOH (118.1 mV/decade), Ni0.9Fe0.1OOH (72.2 mV/decade), Ni0.8Fe0.2OOH (66.8 mV/decade), and Ni0.6Fe0.4OOH (68.8 mV/decade). The Tafel slope value for Ni0.7Fe0.3OOH is comparable to the data published by Jiawang Li et al (35.3 mV/decade) [26] and Meng Li et al 41.9 mV/decade for Fe based electro catalyst [27]. A low value of the Tafel plot (Fig. 2-B) specifies a superior OER ability with lowering overpotential [28].For most of the catalysts, the electrocatalytic performance is usually a function of exposed electrochemical active surface area (ECSA) [29]. ECSA was not directly measured here due to the difficulties in measuring. Instead, the double layer capacitance (C dl ) is implemented to imitate the ECSA, because they are linearly related. To explain the difference among the prepared material, C dl was calculated for each sample by CV method in the non-Faradic region, which is a voltage region commonly positioned at the open circuit potential (OCP) of an electrode, assuming that obtained current is specifically from DL charging-discharging. C dl was calculated by using equation (1). (1) i c = υ C dl Where i c stands for charging current and υ for the scan rate. Fig. 3 indicates the ECSA of all the prepared materials, which clearly reveals that there is a negligible change that occurred in the ECSA of the samples after incorporating the Fe. The electrocatalytic performance at fixed overpotential (300 mV) of all the catalysts was directly compared and the current densities are summarized and given on the left side of Fig. 2-C. It is obvious that the Ni0.7Fe0.3OOH nanosheets showed the highest current density of 56.5 mA cm−2 which was 22.6, 6.27, 4.18 and 5.13 times higher than those of NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH and Ni0.6Fe0.4OOH nanosheets /C, respectively. The high activity of the NiFe DLH is due to its H2O and hydroxide permeable nature so that these species reach layers below the electrode/electrolyte interface. This free moment of a hydroxide ion and water species is the key factor for high OER activity. Electrical impedance spectroscopy (EIS) analysis of all the synthesized catalysts was conducted to probe into the deep insight of the OER kinetics. Fig. 2-D indicates that the samples experience a semicircle in the high-frequency range related to charge transfer resistance (Rct). In comparison to NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH and Ni0.6Fe0.4OOH nanosheets /C, the Ni0.7Fe0.3OOH nanosheets show the minimum Rct of 27 Ω, which signify the enhanced charge transfer kinetics due to incorporation of Fe in the NiOOH sheet. Mohammad Etesami et al. find the same trend for NiFe-based electrocatalysts compared with commercial Pt/C [30]. The sample without Fe shows the highest Rct of 174 Ω. The graph reveals that Rct values decrease as the Fe content increases reaching a minimum value and then increasing following the volcano plot [31].Stability is another key indicator to estimate the performance of an electrocatalyst. CV cycling is generally used to estimate the stability of supercapacitor materials because their real working condition is dependent on charging and discharging [32]. However, in water splitting, the OER is usually performed at a constant current or voltage. In order to imitate the more likely real working condition, chronopotentiometric analysis was used to estimate the stability of the Ni0.7Fe0.3OOH. Fig. 4 -A shows that at 10 mA cm−2 current density, the catalyst was quite stable for almost 15 h. Jiawang Li reported a nitrogen-doped FeNiOOH electrocatalyst for OER and showed comparable chronopotentiometry analysis. Polarization curves of the Ni0.7Fe0.3OOH before chronopotentiometric analysis and after are given as inset of Fig. 4-A. The comparative histogram of overpotential of all the prepared catalysts as a function of different Fe content at fixed current density (10 mA cm−2) indicates that Ni0.7Fe0.3OOH has the best performance among all the catalysts given in Fig. 4-B.In summary, Fe-doped NiOOH nanosheets were successfully synthesized by a single-step approach. Ni0.7Fe0.3OOH nanosheets are represented as a model OER catalyst with significantly enhanced OER activity. It requires an overpotential of 265 mV to generate 10 mA cm−2 current density. The robust OER activity is due to the incorporation of Fe, which increases the material's conductivity and provides active sites for OER. The current density produced at fixed overpotential (300 mV) for Ni0.7Fe0.3OOH was 56.5 mA cm−2 which was 22.6, 6.27, 4.18 and 5.13 times higher than those of NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH and Ni0.6Fe0.4OOH nanosheets /C, respectively. We believe that this approach would promote the encouraging possibilities for designing highly efficient catalysts based on engineering the active sites. Fawad Ahmad: Writing – original draft, Conceptualization, Investigation. Asad Ali: Writing – review & editing, Supervision, Project administration. Jiaqian Qin: 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 work was supported by the Collaborative Innovation Center of Suzhou Nano Science and Technology, MOST of China (2014CB932700), NSFC (21603208, 21573206), Key Research Program of Frontier Sciences of the CAS (QYZDBSSW- SLH017), Anhui Provincial Key Scientific and Technological Project (1704a0902013), Major Program of Development Foundation of Hefei Center for Physical Science and Technology (2017FXZY002), and Fundamental Research Funds for the Central Universities. Fawad is also grateful for a generous CAS-TWAS president’s fellowship.This manuscript accompanies supporting information (Table S1†, XPS Fig. ES2†, XRD Fig. ES3†, SEM images of Ni0.7Fe0.3OOH Fig. S4 ESI†) available free of cost. Supplementary data to this article can be found online at https://doi.org/10.1016/j.rechem.2023.100808.The following are the Supplementary data to this article: Supplementary data 1

Electrochemical water splitting to generate oxygen and hydrogen is a key process for several energy storage and conversion devices. Developing low-cost, robust, efficient, and earth-abundant electrochemical catalysts for the oxygen evolution reaction (OER) is therefore holding a paramount position. Herein, we report the doping process to prepare two-dimensional Fe-doped NiOOH nanosheets with the tuneable molar ratio of Fe ranging from 0 to 0.4 by using a single-pot synthetic approach. Among the obtained nanomaterials, the Ni0.7Fe0.3OOH nanosheets/C exhibited greatly enhanced electrocatalytic performance toward OER in alkaline media (1.0 M KOH), with an overpotential of 265 mV to afford 10 mA cm−2 current density. The current density produced at fixed overpotential (300 mV) for Ni0.7Fe0.3OOH was 56.5 mA cm−2 which was 22.6, 6.27, 4.18 and 5.13 times higher than those of NiOOH, Ni0.9Fe0.1OOH, Ni0.8Fe0.2OOH and Ni0.6Fe0.4OOH nanosheets /C, respectively. Moreover, the nanosheets were able to retain excellent performance for over 15 h without obvious degradation. The Tafel slope for Ni0.7Fe0.3OOH was 44.8 mV/decade. Therefore, this approach has opened a new possibility for designing highly efficient catalyst-based active sites.

Proton exchange membrane fuel cell (PEMFC) is a fascinating sustainable energy technology that converts the chemical energy of hydrogen into electricity to power clean electric vehicles [1–3]. Massive Platinum group metals (PGM) are needed to catalyze the sluggish oxygen reduction reaction (ORR) at the cathode, which constitutes the major cost barrier in the large-scale application of PEMFC [4,5]. Hence, many efforts have been aimed at seeking PGM-free catalysts with decent ORR activity and stability as low-cost alternatives [6–8].Among PGM-free catalysts that have been developed, metal–nitrogen–carbon catalysts (M–N–C, M = Fe, Co, Mn, etc.) stand out because of their most promising fuel cell performance [9–15]. Their initial fuel cell performances could currently approach those of Pt-based cathodes at low loadings. For example, Shui and co-workers reported a Fe–N–C catalyst with highly dense Fe–N4 active sites achieving the highest peak power density (P max) of 1.18 W cm−2 under 2.5 bar H2–O2 [15]. However, the stability of these highly active PGM-free catalysts in real fuel cell conditions is far from satisfactory. Specifically, their performance loss following the first 100-h test typical reaches 40–80%. The limited stability must be addressed to make PGM-free catalysts commercially viable. Therefore, the focus of continued research and development is turning from activity enhancement to the degradation mechanisms and targeted mitigation strategies.An in-depth understanding of the nature of the active sites is a prerequisite for diagnosing the cause of instability. M–N–C catalysts are generally synthesized by pyrolyzing a precursor containing iron, nitrogen, and carbon elements. The obtained catalysts might possess three types of active sites including nitrogen-doped carbons (denoted as NxCy), nitrogen–carbon encapsulated metal nanoparticles (denoted as M@NxCy), and nitrogen coordinated metal atoms embedded in carbon substrate (denoted as M–Nx or MNxCy) [16–24]. After extensive experimental and computational studies [25–29], a consensus has been reached that M–Nx sites are most active in acid media, though some studies showed that M@NxCy structures also exhibited exceptional ORR activities. The overall fuel cell performance depends on the active site density (SD), the intrinsic activity of a single site (turnover frequency, TOF), as well as the accessibility to the reactants [30,31]. The decrease of any would result in the performance decay. In this context, four main suspicious degradation mechanisms of PGM-free catalysts were proposed previously: 1) carbon oxidation, 2) demetalation, 3) protonation and 4) micropore flooding [32–34]. The root-cause of instability is under extensive debate partially due to the structural complexity of M–N–C catalysts. In addition, the protocols to study stability vary with labs, which may also bring different results. For example, the rotating disc electrode (RDE) technique is often used for simplicity, which is different from the fuel cell environment [35]. The atmosphere, temperature, potential range also affect the conclusions.Recently, with the advancement of synthesis of ultra-pure materials [36–38] and operando characterization techniques [39–41], the research community is active and fruitful to identify the most likely degradation mechanisms by well-designed experiments. In this review, a survey of the increasing understanding of the degradation mechanisms of PGM-free catalysts is provided, with the major advances highlighted. Based on the mechanism understanding, the strategies for improving the stability of PGM-free catalysts are also summarized.The automotive applications of PEMFC involve various working conditions, including routine steady-state operation, rapid cycling, and frequent starts and stops. The term stability refers to the ability to maintain performance at constant current/voltage conditions while durability refers to the ability to maintain performance following a voltage cycling accelerated stress test [32]. Most fuel cell level investigations have focused on stability, while only limited studies on voltage cycling durability. As shown in Fig. 1 , the stability was typically measured by chronoamperometry at the cell voltage around 0.4–0.6 V [15,42]. Two decay rates may be observed: a fast decay lasting about 15–20 h accounting for the major performance loss, followed by a much slower decay lasting up to the end of test. The major task of stability studies is to identify the cause of the rapid initial performance loss.The oxidation of the carbon support leads to the modification of the carbon surface and even worse the disintegration of active sites, which can be divided into two types: electrochemical oxidation and chemical oxidation. Electrochemical oxidation of the carbon is triggered by electrochemical potential and thermodynamically possible above 0.207 V versus standard hydrogen electrode (SHE) [39]: (1) C + 2H2O → CO2 + 4H+ + 4e−, E 0 = 0.207 V (2) C + H2O → CO + 2H+ + 2e−, E 0 = 0.518 V However, these reactions generally require an overpotential of at least hundreds of millivolts. Such high potentials usually appear in the case of uncontrolled and transient starts/stops (ST/ST) of PEMFCs [43]. Mayrhofer and co-workers observed intensified ORR activity decay of a Fe–N–C catalyst with positively shifted potential range and elevated temperature by RDE tests [39]. They further made in situ observation of carbon oxidation of the catalyst in a modified scanning flow cell (SFC) system with differential electrochemical mass spectroscopy (DEMS) [39]. They found the onset potentials of carbon oxidation to CO2 and CO were about 0.9 and 1.2 V, respectively (Fig. 2 ). Using identical location-scanning transmission electron spectroscopy (IL-STEM) and IL-energy dispersive X-ray (EDX) spectroscopy analysis, they revealed a 2D shrinkage of 5–15% of the catalyst particle dimensions and a 20% reduction in iron content after 5000 cycles performed between 1.2 and 1.5 V at 50 °C in 0.1 M HClO4 (Fig. 2b–d). The Fe dissolution signal was directly proportional to the rate of carbon oxidation (Fig. 2e). This study demonstrated the direct correlation between and carbon oxidation/Fe leaching and ORR activity decay at high electrochemical potential. Note that the above experiments were performed in a de-aerated electrolyte. The presence of O2 is not expected to play a direct role in electrochemical carbon corrosion since the reactant is water rather than O2. However, the aerobic environment may trigger the formation of hydrogen peroxide (H2O2), an undesirable byproduct of ORR process.In addition to electrochemical carbon oxidation at high potentials, chemical oxidation by H2O2 is another widely accepted cause of instability of PGM-free catalysts. There might be two paths of H2O2 attack: 1) direct oxidation by H2O2; 2) H2O2 decomposes into reactive oxygen species (ROS) via Fenton reaction, and then ROS attack the catalysts. A proposed mechanism for the direct attack was provided by Schulenburg and co-workers [44]. In this work, it was suggested that the H2O2 can directly attack the N ligands to which the metal center is bound. While this mechanism could explain some parameter changes in Mössbauer spectra of the degraded catalysts, the follow-up investigations were rare.The attack by ROS was firstly suggested by Dodelet and co-workers [45]. These ROS, in particular hydroxyl free radical (·OH), readily add to unsaturated aliphatic or aromatic compounds on the carbon support thus forming oxygenated surface groups, which were believed to be detrimental to the activity. It should be noted that the ROS would also induce the oxidative degradation of the Nafion membrane, such as thinning, pin-hole formation and ionic conductivity loss, which might ultimately contribute to the stack failure [46,47]. The mild surface oxidation was systemically investigated by Choi and co-workers with RDE technique (Fig. 3 ) [48]. They showed that ex situ exposure to H2O2 in the acid medium would selectively oxidize a fraction of top-surface carbon atoms via Fenton-like reactions in the presence of surface iron sites, without the formation of volatile CO or CO2. Such a mild surface oxidation would not change the structure of FeNxCy moieties but weaken the O2-binding, thus decreasing their single-site activity and 4e − selectivity. This type of deactivation was reversible, as the activity and selectivity could be partially recovered after electrochemical reduction to remove some oxygen groups. It was also found that peroxide-treatment in alkaline medium did not modify the ORR activity nor selectivity of the catalyst. This discrepancy suggested the formation of ROS from peroxide and surface iron sites was pH-dependent and primarily supported the pathway of indirect attack by ROS.Although only mild surface oxidation occurred with ex situ exposure to H2O2, the fate of the catalysts during PEMFC operation would be rather complex. Recently, Maillard and co-workers reported unexpected carbon corrosion coupled with significant demetalation at low potential of 0.3–0.7 V in the presence of O2 [49]. Compared with load cycling in Ar-saturated electrolyte, the activity loss was four times higher in O2-saturated electrolyte. This type of carbon corrosion was not controlled by classical electrochemical carbon corrosion but due to ROS produced between H2O2 and Fe sites via Fenton reactions. However, different from the only mild oxidation of the surface from ex situ H2O2 treatment, additional irreversible carbon corrosion, i.e. forming volatile products CO and CO2, also occurred. A significant fraction of the atomic Fe–Nx sites was leached from the catalyst, which either exited the cathode layer at 25 °C or precipitated as iron oxide particles at 80 °C. Although the above studies were conducted using RDE technique, the iron oxide formation from some Fe–Nx sites could also be transposable to the PEMFC operating conditions at 80 °C. In this report, the mechanism of iron leaching was left quite vague but strongly related to the irreversible carbon corrosion. The overall decay of the activity was caused by both the decreased active site density (SD) and the decreased single-site activity (TOF), with the latter due to the mild surface carbon oxidation. However, it was currently difficult to quantitatively distinguish the contributions of these two factors, because of the technical barrier of accurate measurement of SD under the interference of iron oxide.The metal leaching from PGM-free catalysts has been investigated for a long time [50,51]. The main challenge lies in the complex heterostructure of the catalyst and the lack of operando characterization methods. Therefore, it is difficult to identify the type of metal moieties that are easily leached out and whether it is responsible for performance degradation.The metallic Fe or Co particles without the protection of graphitic layers are generally believed with negligible ORR activity and are readily dissolved in the acid PEMFC environment according to the Pourbaix (E-pH) diagrams (Fig. 4 ), which illustrate the thermodynamic stability of metals in aqueous solutions [52,53]. Contrary to the common belief, Mayrhofer and co-workers showed that the operando Fe leaching from iron particles was potential-dependent in the SFC system mentioned in Section 2.1 [39]. Fe ions were released from Fe–N–C at low potential (<0.7 V) while no No Fe signal was detected above 0.8 V. They further elucidated that the ex situ acid-wash could not completely remove the iron particles due to a relatively high open circuit potential (~0.9 V) leading to the formation of insoluble ferric species, whereas these particles dissolved at the low potential due to operando reduction to soluble ferrous cations [40]. The Fe leaching from inactive iron particles will not cause a decrease in activity, and it may even improve the activity as new mesopores are created to expose more active sites [54]. However, the leached metal ions would have negative effects on the PEMFC performance, such as the carbon oxidation via Fenton reaction and the deterioration of the membrane and ionomer in the catalyst layer [55].The demetalation of active Fe–Nx sites directly decreases the performance. Besides the destruction of Fe–Nx sites via irreversible carbon corrosion as discussed above, there were very few reports discussing the direct demetalation of the Fe–Nx sites. In 2005, Coutanceau and co-workers reported the deactivation of a FePc-based catalyst during the ORR was due to a substitution of the central iron by two protons, leading to inactive H2Pc [56]: (3) FePc + 2H+ → H2Pc + Fe2+ It was suggested that Fe2+ was oxidized to Fe3+ in the presence of O2, which reduced its ionic radius thus making it less stable in the macrocycle. More recently, Dodelet and co-workers proposed that the specific demetalation of Fe–N4 catalytic sites located in the micropores was at the origin of the initial activity loss based on a systematic study at fuel cell level [57]. The Fe–N4 sites were calculated thermodynamically stable in stagnant acidic conditions, but those in the micropores would demetalate under the quick water flow running into the micropores. The flux of water, which was driven by the humidified air streaming through the working cathode, would continuously shift the thermodynamic equilibrium between Fe ions and Fe–N4 sites towards the direction of demetalation according to Le Chatelier principle. The degree of demetalation was measured by Mössbauer spectroscopy and showed a similar trend to the relative current density measured at 0.6 V (Fig. 5 a), which strongly supported this degradation mechanism. The electro-oxidation of the catalyst surface was ruled out to play a major role in the initial fast performance decay based on the following two reasons. First, the activity decay was found irrespective of the value of the potential applied during chronoamperometry (Fig. 5b), while a large difference in the carbon support oxidation currents was observed (Fig. 5c). More importantly, electro-oxidation would only result in more hydrophilic micropores and less accessible active sites, thus the loss of Fe–N4 sites would not be detected by Mössbauer spectroscopy. It was also noticed that a fraction of Fe–N4 sites survived after 50 h, which were recognized as Fe–N4 sites located in the mesopores. The dynamic equilibrium in the mesopores could be established and a longer stability term was expected. These findings suggest that increasing the proportion of active sites located in mesopores might be a way to improve the stability of PGM-free catalysts. Wu and co-workers provided electron microscopic evidence for the demetalation of Fe–N4 sites (Fig. 5d–g) [58]. Using high angle annular dark field STEM (HAADF-STEM) and electron energy loss spectroscopy (EELS), they observed the formation of iron clusters and the break of Fe–N bonds on a single-atom Fe–N–C catalyst that had worked for 100 h.There were several Fe@NxCy structures also reported highly active in acid medium. Although their activities were still inferior to the atomic Fe–Nx sites, their fuel cell stabilities seemed better. For example, in 2011, Zelenay and co-workers reported a series of PGM-free catalysts featuring metal-containing particles encapsulated in onion-like graphitic carbon nanoshells [59]. The most active catalyst could serve stably for 700 h at a cell voltage of 0.4 V. Bao and co-workers produced a catalyst of Fe nanoparticles encapsulated within the compartments of pea-pod like CNTs that also demonstrated long-term stability in PEMFC [60]. More recently, Mukerjee and co-workers synthesized a Fe–N–C catalyst with exclusive Fe@NxCy structure and devoid of any Fe–Nx sites that showed high activity (half-wave potential E 1/2 = 0.77 V) and a 4e − selectivity in acidic media [61]. The membrane electrode assembly (MEA) made with this catalyst was even more durable compared with the state-of-the-art Pt/C MEA. The exact origin of the durability of M@NxCy was still under investigation. A possible advantage of this structure was that Fe does not directly participate in the reaction, thus potentially eliminating any Fenton reactions involving exposed iron ions. Nevertheless, Jaouen and co-workers compared the stability of M–Nx sites and M@NxCy structures (M = Fe, Co) and found the former were more robust to both demetalation and carbon corrosion than the latter during load-cycling (0.6–1.0 V) or ST/ST process (1.0–1.5 V) [62]. However, this study was performed in an Ar-saturated electrolyte. Further clarification of the operando stability of these two types of active sites requires more intensive investigation.The mechanism of protonation was first proposed by Popov and co-workers to explain the difference in stability between the active sites of pyridinic nitrogen and graphitic (quaternary) nitrogen [63]. Two catalysts were synthesized with different pyrolysis temperatures. One pyrolyzed at 800 °C containing both pyridinic and graphitic nitrogen showed higher initial activity but much lower stability. The other pyrolyzed at 1100 °C only containing graphitic (quaternary) nitrogen showed the opposite. It was hypothesized that the lone electron pair of pyridinic-N could be protonated in the acidic PEMFC environment and thus formed an inactive pyridinic-N–H group. In contrast, the graphitic-N possessed no extra electron to be protonated, explaining its higher stability. The protonation mechanism provided a plausible explanation for the initial rapid performance loss of PGM-free catalysts in PEMFC. However, this degradation mechanism was criticized by Banham and co-workers [32]. Based on the protonation hypothesis, the pyridinic-N would be protonated very fast in the RDE test given the more available protons in the liquid electrolyte. This should have resulted in very similar activities of these two catalysts, which had not been observed in the experiment.Afterward, in 2011, Jaouen and co-workers provided a variant of the protonation mechanism: N protonation and anion binding [64]. It was proposed that the TOF of Fe–N4 sites could be tuned by the chemical state of the adjacent basic N-groups on the catalytic surface. The protonation of the basic N-groups did not decrease the TOF of the Fe–N4 sites, while the subsequent anion binding on the protonated N sites neutralized their basicity and ultimately resulted in activity decay. The activity could be restored after the anions were removed thermally or chemically. It was further proposed that protonation of basic N-groups was rapid but the anion binding was delayed in PEMFC due to restricted mobility of the sulfonate groups. Therefore, slow anion binding by polymeric anions was believed to be an indispensable reason for the initial rapid decay of PGM-free catalysts in PEMFC. However, in a report published in 2016, the authors abandoned this anionic neutralization hypothesis after several unsuccessful attempts to modify the first decay behavior by tuning the properties of the proton providers, such as replacing part of Nafion with hydrous ruthenium oxide [42].Based on intensive studies [65,66], Dodelet and co-workers had demonstrated that the active M–Nx sites were mainly located in the micropores of carbon matrix. Therefore, the water flooding of micropores would be detrimental to fuel cell performance by impeding the transport of oxygen. In 2015, micropore flooding was proposed by Dodelet and co-workers to explain the superior stability of a catalyst obtained after the pyrolysis in Ar at a high temperature of 1150 °C [67]. This catalyst possessed the best degree of graphitization of the carbon support and loss of heteroatoms (N and O), which was proposed to render the carbon support highly hydrophobic and would, therefore, reduce the possibility of water flooding. This mechanism also corroborated the negative correlation between stability and the micropore percentage in a catalyst. In a follow-up report, they further proposed that the micropore flooding was mainly responsible for the initial performance loss of PEMFC [42]. The rate of flooding depended on the hydrophilicity of the micropore walls, which would gradually change from hydrophobic to hydrophilic on the first 15 h of PEMFC operation due to the slow carbon electro-oxidation. The surface oxidation not only decreased the current density at 0.6 V by inducing mass transfer problems but also impeded the kinetics of electron transfer at 0.9 V. The Fenton reaction via H2O2 and iron ions was excluded as the major cause of the rapid initial performance decay, because they produced a series of catalysts with different level of iron content and found their decay patterns were similar.It is also important to clarify the difference between the micropore flooding and catalyst layer flooding. The latter can occur at high relative humidities (RH)/current densities due to the accumulation of water in the catalyst layer, which blocks mass transport pathways thus leading to performance loss. This type of degradation is reversible with periodic water drainage but not intrinsic to the catalyst. Recent work showed that the drainage ability of the catalyst layer had a negligible influence on the stability of Fe–N–C fuel cells [68].More recently, Chen and co-workers conducted a systematic study to investigate micropore flooding in situ before and after fuel cell stability tests (Fig. 6 ) [69]. The results cast doubt on micropore flooding as a major contributor to instability. The degree of micropore flooding was monitored by the changes in the double layer capacitance obtained by cyclic voltammetry (CV). It was analyzed that the micropores are partially wetted at beginning of life (BOL). After the stability test, a small (~8%) degree of additional catalyst layer wetting was observed, which could not account for the significant performance loss ( × 10 reduction in activity). The micropore flooding hypothesis was rejected primarily because the performance decay was largely kinetic, rather than from mass transport loss due to flooded micropores. The reversible catalyst layer flooding was also ruled out as no performance recovery was achieved after the dry-out protocol. Later, Dodelet and co-workers agreed that the micropore flooding mechanism should be abandoned and proposed the new hypothesis of specific demetalation, as discussed in section 2.2.As discussed above, carbon oxidation by Fenton-like reaction and demetalation might be the most likely degradation mechanisms of PGM-free catalysts. Therefore, effective improvement of the stability may be achieved by eliminating Fenton reaction, increasing the corrosion resistance of carbon support, as well as mitigating demetalation.The iron ions are criticized for their catalysis of the decomposition of H2O2 into ROS. Other transition metals, such as Cr, Mn, Co, Ni, Cu, Zn, are not powerful Fenton's reagents [70]. Therefore, the M–N–C catalysts based on these non-Fe metals hold the potential to alleviate Fenton reaction [71–74]. However, the ORR activities of the PGM- and Fe-free catalysts are generally poorly competitive with the iron counterparts, which restricts their application in real fuel cells. Given the ORR activity is a product of SD and TOF, increasing the density of M–Nx active sites in a Fe-free catalyst is a pathway to enhance the activity.Shui and co-workers optimized the Co content of a series Co–N–C catalysts to achieve the highest density of Co–N4 active sites [75]. The best catalyst exhibited a P max of 0.83 W cm−2, approaching that of Fe–N–C. As shown in Fig. 7 a, the higher stability of Co–N–C over Fe–N–C was observed. Wu and co-workers developed a surfactant-assisted confinement pyrolysis strategy to fabricate a Co–N–C catalyst with doubled Co–N4 site density, which showed fuel cell performance comparable to Fe–N–C [76]. A 100-h stability test at a cell voltage of 0.7 V using 1 bar H2-air is shown in Fig. 7b. Although significant performance decay was observed, the authors wrote that the stability was commendable at such a relatively high voltage, compared to other PGM-free catalysts. They also reported a single-atom Mn–N–C catalyst with increased active site density, achieving encouraging activity and stability [70]. Xing and co-workers synthesized a novel Cr–N–C catalyst with atomic Cr–N4 sites and investigated its catalytic ability toward Fenton reaction [77]. This catalyst exhibited a considerable half-wave potential of 0.74 V, and more impressively its stability was superior to the Fe–N–C counterpart. Although the Cr–N–C catalyst showed high H2O2 production, the ROS formation was suppressed according to the color reaction with ABTS (2, 20-azinobis (3-ethylbenzthiazoline-6-sulfonate) (Fig. 7c and d). The low catalytic ability toward Fenton reaction could thus remedy the shortcoming of high H2O2 production. In addition, PGM-based M–N–C catalysts with atomic M−N4 (M = Ir, Rh, Pt and Pd) were successfully synthesized by Shui and co-workers [78]. The Ir–N–C catalyst also exhibited higher durability than typical Fe–N–C catalysts.It is worth noting that some metal-free catalysts have shown sufficient stability in PEMFC [79,80]. However, their performances were still at a relatively low level. Further improvements may render them applicable in certain scenarios.As discussed in Section 2.2, Fe@NxCy structures have the advantages of indirect participation of iron and thus the potential to eliminate Fenton reactions. The iron nanoparticles also catalyze the graphitization of surface carbon layers thus increasing their resistance to oxidation. Recently, a growing number of researchers resort to this type of catalyst to achieve a stable PGM-free fuel cell cathode with acceptable performance [81–84]. The density of metal nanoparticles and the number of graphitic layers on their surface have a great impact on the performance. However, this type of catalyst is usually produced by pyrolysis, hence the controllable synthesis of this material with high-density nanoparticles and well-defined graphitic layers remains a challenge [85].Because H2O2 is the source of ROS, decreasing the production of H2O2 or breaking down the produced H2O2 might be effective to mitigate the attack of ROS. Liu and co-workers made a combination of Pt–Co nanoparticles and Co–N–C catalyst and found the interaction between Pt–Co and Co–N4 sites improved ORR activity and durability [86]. This combination led to a 4-electron transfer, which suggested the completed conversion from O2 to H2O instead of H2O2. It was speculated that intermediate H2O2 was transferred from the Co–N4 site to the Pt–Co nanoparticles through a reverse spillover. As shown in Fig. 8 , theoretical calculations revealed the subsequent thermodynamic favorable breakdown of H2O2 over the strained Pt (111) surface, which therefore served as a powerful H2O2 scavenger. Similarly, Jaouen and co-workers reported the addition of 1–2 wt% Pt to a Fe–N–C catalyst could significantly improve its fuel cell stability [87]. Introducing ultralow-loading of Pt to PGM-free catalysts may provide a way to achieve a balance of cost and performance [88]. The remaining challenge is to find applicable low-cost H2O2 scavengers. Eliminating ROS might be another way to circumvent Fenton reaction. Ramani and co-workers previously demonstrated that the addition of CeO2 nanoparticles to the proton exchange membrane effectively scavenge ROS thus mitigating their degradation [89]. The utility of CeO2 as a ROS scavenger to improve the stability of PGM-free catalysts, however, remains to be further explored [90].To date, most well-performing PGM-free catalysts were built on highly microporous carbon supports with low degree of graphitization, which resulted in the drawbacks of low conductivity and poor corrosion resistance. In this regard, growing active sites on highly graphitic carbons such as carbon nanotubes (CNTs) and graphene may have the chance to improve stability [91–94]. Chen and co-workers fabricated a catalyst with abundant 3D porous graphene-like structures, which hosted a dense population of accessible active sites [95]. After 5000 load cycles in N2 environment, this catalyst retained 90% of its initial fuel-cell performance, attesting its high resistance to carbon corrosion. Kang and co-workers successfully embedded high-density Fe–Nx active sites into CNTs (Fig. 9 ) [96]. When compared with a state-of-the-art ZIF-derived Fe–N–C catalyst, the new Fe–N/CNT catalyst demonstrated much enhanced stability.Fluorination of the carbon is also known to be beneficial to the stability towards oxidation [97]. Zhou and co-workers developed a surface fluorination strategy to boost the stability of the Fe–N–C cathode (Fig. 10 ) [98]. The catalyst surface was covalently grafted of a hydrophobic trifluoromethylphenyl (Ar-CF3) group, which could effectively prevent the catalyst layer flooding. More importantly, the electron-withdrawing property of Ar-CF3 group lowered the Fermi level of carbon matrix, thus increasing the reaction energy barrier and decreasing the rate of carbon oxidation. The intrinsic hydrophobicity and timely removal of excess water were also suggested to decrease the H2O-involved carbon oxidation rate. The demetalation was suppressed, which might be attributed to the decrease of carbon corrosion. As a result, the fluorinated catalyst could deliver a stable current density of 0.56 A cm−2 at cell voltage of 0.5 V up to 120 h. Note that the surface fluorination had an adverse effect on the BOL performance, however, this strategy was laudable as the final current density was much higher than that of original Fe–N–C.The primary strategy is to remove unprotected free metals in the catalysts. It could be achieved by post-acid wash or by fabricating pure-phase materials. As the mechanism of direct demetalation from the atomic M–Nx site was proposed most recently, it is not surprising there was little work discussing the targeted mitigation strategies. A survey of the literature suggests a possible way is to redesign the atomic structure of the active site, which should exert atomic-level protection to the metal center or increase the strength of the metal-N bonds [99]. Shui and co-workers grafted of a Pt atom onto the iron center through a bridging oxygen molecule, creating a new active moiety of Pt1–O2–Fe1–N4 (Fig. 11 a) [100]. This structure indeed exhibited considerably improved fuel cell performance and stability compared with the untreated Fe–N–C (Fig. 11b). The Pt1–O2– cap was speculated to avoid Fenton-like reaction and strengthen the Fe–N coordination. Another hotspot of the electrocatalysis is the double-atom catalyst, which features a N-coordinated double-atomic metal center. Li and co-workers designed and synthesized a catalyst with Fe–Co dual sites embedded in nitrogen-doped carbon [101]. This catalyst exhibited state-of-the-art ORR activity (E 1/2 = 0.863 V) and fuel cell performance (P max = 0.98 W cm−2 at 2 bar H2–O2). Surprisingly, the catalyst showed negligible performance loss after 100-h H2-air single cell operation. Several follow-up studies reporting similar diatomic structures such as Co–Zn atomic pairs also pointed to their enhanced stability [102,103]. However, the exact reason for the superior stability of double-atom catalysts was seldomly discussed. If these results were convincing, we speculated the larger radius of the diatomic center might decrease the tendency of demetalation.Several possible mechanisms have been previously proposed to account for the instability of PGM-free catalysts in PEMFC, including 1) carbon oxidation, 2) demetalation of metal sites, 3) protonation of active sites with possible subsequent anion binding, and 4) micropore flooding. After decades of extensive research, carbon oxidation and demetalation are generally accepted as primary degradation mechanisms, while the latter two mechanisms are strongly challenged. Carbon oxidation can be either triggered by high electrochemical potential >0.9 V or caused by oxidative attack by H2O2/ROS in a wider potential range. Mild surface oxidation may not destruct the nearby M–Nx active site but decrease its TOF via weakening the O2-binding. More severe carbon corrosion with the formation of volatile CO or CO2 will lead to the disintegration of active sites. Another type of specific demetalation of micropore-hosted M–N4 sites straightforward degrades the activity by decreasing the number of active sites. The leached metal ions, particularly iron ions, could in turn catalyze the ROS formation from H2O2 via Fenton reaction.With the deepening understanding toward the degradation mechanisms, the researchers are seeking solutions to the poor stability with increasing vigor. The mitigation strategies include eliminating Fenton reaction, enhancing carbon corrosion resistance, and alleviating demetalation. To eliminate Fenton reaction, the direct participation of iron should be avoided. In this regard, the researchers are developing PGM- and Fe-free catalysts (such as Co–N–C, Mn–N–C) and the catalysts of encapsulated M@NxCy structures with their performances comparable to the Fe–N–C counterparts. Introducing H2O2/ROS scavengers is also a promising strategy to alleviate Fenton reaction. To enhance carbon corrosion resistance, highly graphitic CNTs and graphene could be used as robust supports for active sites. Fluorination appears also beneficial the stability towards carbon oxidation. The mitigation of direct demetalation has been seldomly reported in the literature. A direction is atomic-scale redesign of the active site with stronger M–N bonds.Despite the significant progress, the long lifetime of high-performance PGM-free catalysts has remained an elusive target. Further understanding the degradation mechanisms at fuel cell levels requires advanced in situ characterization techniques with more localized resolutions. Several key questions remain to be answered and the efforts may focus on 1) identification of the durable and non-durable sites among FeNxCy moieties, 2) exploration of the factors affecting demetalation and remediation strategies, 3) investigation of the reason why some catalysts (for example, Fe–Co double-atom catalyst) are so stable in fuel cell operation.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 Thousand Talents Plan of China, the National Natural Science Foundation of China (Grant No. 21673014 and 21975010).

While Platinum group metals (PGM) free catalysts are promising alternatives to expensive Pt as the cathode catalyst in proton exchange membrane fuel cells, their rapid degradation must be addressed for the commercial feasibility. This review provides a historical survey of the possible degradation mechanisms of PGM-free catalysts. Decades of extensive studies confirm that carbon oxidation and demetalation are primarily responsible for the instability, whereas the mechanisms of protonation and micropore flooding are strongly criticized. Based on the mechanism understanding, the mitigation strategies for improving stability are discussed in detail. Finally, some directions to achieve high-performance and durable PGM-free catalysts are proposed.

Water still remains an essence of life, however with continuous discharge of waste into water bodies, access to clean and potable water has continued to dwindle. The detection of various waste organic contaminants such as pharmaceutical, dyes, pesticides, personal care products in surface, ground and drinking water is of major challenge globally due to their detrimental effects (Du and Zhou, 2021). For example, the consumption of water containing organic dyes like methylene blue (MB), rhodamine B (RhB) and methyl orange (MO) can cause eye irritation, bladder cancer and respiratory problems (Fernández et al., 2010; Rai et al., 2005; Tan et al., 2015). With the rapid growth of textile industries, one of the major industrial sources of organic dyes owing to>100 000 tons of production of dyes per year, thus it is vital to monitor and treat these industrial waste dyes before reaching the environment (Abdi et al., 2017; Gupta et al., 2013; Gupta and Suhas, 2009; Holkar et al., 2016; Katheresan et al., 2018). Of major concern, are pharmaceutical products that have been widely used in various fields including households, agriculture and medicine. In medicine, products such as penicillin, ciprofloxacin, tetracycline and sulfamethoxazole are used as antibiotics to treat bacterial infections (Huang et al., 2021; Mo et al., 2017). The production of antibiotics increases daily due to their high demand for bacterial infection prevention or to cure diseases. The production of penicillin per year was reported to be approximately 28,000 tons, thus making it 68 % of the global consumption of antibiotics (An et al., 2015; Du and Liu, 2012). The presence of some of these antibiotics are of great concern due to their serious health effects such as vomiting, nausea, acute renal failure and diarrhea (Orimolade et al., 2020).Other water contaminants which have been detected in different water bodies include pathogenic bacteria such as escherichia coli (E.coli), staphylococcus aureus (S. aureus), pseudomonas aeruginosa (P.aeruginosa), enterococcus faecalis (E.facelis) and other microbes. Inappropriate disposal of sewage and animal waste are the most common sources of faecal matter in the environment. The discharge of these waste materials from the environment into different water bodies such as rivers, lakes, oceans and streams does not affect only chemical oxygen demand (COD), biological oxygen demand (BOD) and turbidity of the surface water but also increases the number of various pathogenic pollutants (viruses and bacteria) existing in them (Pandey et al., 2014). In 2020, approximately>12 % of the global population was reported to be drinking water containing a substantial amount of unsafe pathogens. Drinking water containing these harmful pathogens can be lethal and cause some serious health issues and waterborne diseases such as diarrhoea, polio, typhoid and malaria etc (Pandey et al., 2014). According to the World Health Organisation (WHO) and standards, the allowed recommendable limitation concentration of organic dyes and bacteria should be below 1 ppm and 0 CFU/100 ml in drinking water, respectively (Katheresan et al., 2018; Masekela et al., 2020; Mahlaule-Glory et al., 2019). Thus, it is crucial to maintain the level of organic dyes, pharmaceutical and pathogenic bacterial within permissible limit so as to provide clean drinking water to humans and protect the environment.Several water technologies and bacterial disinfection techniques including chlorination, chlorine dioxide, ozonation, ultraviolet light (UV), adsorption, membrane filtration and coagulation have been developed to maintain the level of waste water pollutants (organic dyes, bacteria and pharmaceuticals) within a safe level (Hassan et al., 2012; Masekela et al., 2022b; Saucier et al., 2017; Sirés et al., 2014; Sirés and Brillas, 2012a). However, these methods suffer from several limitations including generation of secondary toxic waste, high cost maintenance, incomplete removal of wastewater pollutants, poor recyclability and the use of toxic chemicals. Chlorination is one of the most popular and inexpensive bacterial disinfection techniques for the removal of all micro-organisms present in drinking water. Even though this technique is relatively less expensive, it produces harmful toxic by products such as trihalomethanes (THMs), haloacetonitriles (HANs), haloacetic acids (HAAs) etc (Xiang et al., 2018). These disinfection by products (DBPs) have negative impacts on human health as they can cause intestinal cancer. Additionally, chlorination with other methods like adsorption and filtration partially removes pharmaceuticals from wastewater, since 60 % of pharmaceutical residues remain even after treatment (Orimolade and Arotiba, 2020; Sirés and Brillas, 2012b; Xiao et al., 2015). Furthermore, adsorption and membrane filtration generate secondary toxic waste pollutants, thus require additional treatment which is very expensive (Gupta et al., 2012). Therefore, it is very important implement methods which are highly effective, economical and can completely degrade a majority of the wastewater pollutants into less harmful by products.Advanced oxidation processes (AOPs) such photocatalysis and piezocatalysis have been used as effective methods for complete destruction of organic waste pollutants into less harmful by products. Photocatalysis and piezocatalysis uses generated strong oxidants such hydroxyl radicals (•OH) and superoxide anion (•O2 –) to completely decompose organic pollutants under the influence of visible light and ultrasonic vibration, respectively (Chen et al., 2020; Koe et al., 2020; Li et al., 2019; Liang et al., 2018; Wu et al., 2018a). Unlike other conventional methods, AOPs completely oxidise organic waste pollutants into less harmful by products such as carbon dioxide (CO2) and water (H2O). In the photocatalytic degradation process, one main disadvantage is the fast recombination of electrons and holes (X. Liu et al., 2020). Over the past years, several methods such as metallic or non-metallic doping, formation of heterojunction and composites have been employed to enhance their photocatalytic activity, however effective electron-holes separation still remains a problem (Alex et al., 2019; Kanhere et al., 2014; Qi et al., 2017; Wang et al., 2017; Yong Zhang et al., 2019). Consequently, a piezo-electric field that is built within semiconductors has been shown to effectively separate charge carriers (electron and holes) to prevent recombination reactions. Recently, piezoelectric perovskites (ABO3) structure materials have been employed as an alternative way for better separation of charge carriers (e- and h+) (Y. Feng et al., 2018; Fu et al., 2021; X. Li et al., 2021, 2021; Liu et al., 2021; Y. Liu et al., 2020; J. Wu et al., 2020; J. Zhang et al., 2019).Piezoelectric materials are known as smart materials which produce electric charges under the influence of applied mechanical vibration. These smart materials also tend to exhibit inverse piezoelectric effect, like the generation of mechanical stress under the influence of applied electric field (Xu et al., 2018). The generated electric charges on the opposite site of piezoelectric materials tends to form an electric field across the material. The built in electric field significantly separates the charge carrier (e- and h+) which further reacts with dissolved water and oxygen molecules to generate reactive oxygen species (hydroxyl and superoxide radicals), which are responsible for the breakdown of organic waste water pollutants (Y. Feng et al., 2018; Mushtaq et al., 2018; J. Wu et al., 2020).Among the numerous piezoelectric materials which have been used as piezocatalysts for catalytic degradation of organic waste pollutants present in wastewater, (BaTiO3) has grabbed more attention as a piezocatalyst due to its excellent piezoelectric properties and biocompatibility (Kumar et al., 2019a; Ray et al., 2021). Besides that, it is a lead-free piezoelectric material thus making it more appropriate to be applied in environmental applications. Previously, BaTiO3 as a lead free piezocatalyst has been widely used in sensors. However, recently piezo-photocatalytic applications of BaTiO3 as a piezo-photocatalyst has attracted more attention in environmental wastewater treatment (Aksel and Jones, 2010; Ray et al., 2021). Therefore, this review article gives an overview of the recent applications of BaTiO3 as a piezo-photocatalyst for the catalytic breakdown of organic dyes, bacteria and pharmaceutical pollutants. Moreover, the concept of piezocatalysis, photocatalysis, different fabrication methods, relevant piezoelectric properties and modification methods of BaTiO3 are discussed in detail.Photocatalysis and piezocatalysis processes are regarded as advanced oxidation processes. These two processes have been widely used in many applications including water splitting, bacterial disinfection, degradation and wastewater treatment (Mengying et al., 2017). The concept of piezocatalysis is similar to that of photocatalysis, the only difference lies on the triggering source to generate reactive oxygen species (ROS) which participate in redox reactions to degrade organic pollutants. In photocatalysis, a light source is usually utilized in the presence of a semiconductor (photocatalyst) to generate electron-holes pairs. The semiconductor absorbs the irradiated UV light with high energy thus resulting in electron excitation from valence band (VB) to conduction band (CB) leaving holes behind as displayed in Fig. 1 .As shown in Fig. 1, the photo-generated electron-hole pairs move on separate active sites of the semi-conductor to initiate redox-reactions which generates reactive oxygen species (ROS) responsible for the decomposition of organic waste pollutants. Unfortunately, the rate of electron-holes recombination is very fast which limits the application of semiconductors for photocatalysis. However, in piezo-photocatalysis, an internal voltage is generated under ultrasonic vibration with a built-in-electric field within the semiconductor. The in-built electric field piezo-semiconductors assists in the separation of photo-generated charge carries thus improving the photoactivity of the semiconductor. As shown in Fig. 2 (a) and (b), the separated charge carries due to piezoelectric effect at the opposite surfaces generates free radicals through redox reactions.Piezo semiconductor materials under the influence of applied pressure have been shown to behave like electrocatalytic reactors. The free electric charges (electrons and holes) at opposite sides of these materials tend to act as anode and cathode (Liang et al., 2018). The reaction (1) and (3) shows the formation of reactive oxygen species from free electric charges on the opposite sites of piezoelectric semiconductor materials. As shown in equation (2) and (3), the free positive charges react with water to form hydroxyl radicals (•OH), whereas negative charges react with free oxygen molecules to form superoxide radicals (•O2 –). These reactive oxygen species (•O2 – and •OH) are regarded as strong oxidants and are responsible for the degradation of organic dyes and bacterial disinfection. (1) BaTiO3 (Piezoelectric material) + ultrasonic vibration → BaTiO3 (e- + h+) Negatively charged surface of piezoelectric material (2) h+ + H2O → •OH + H+ (3) h+ + OH– → •OH Positively charged surface of piezoelectric material (4) e- + O2 → •O2 – Unlike photoelectrocatalysis which is another type of electrochemical advanced oxidation for wastewater treatment, this process requires an external high voltage to reduce the rate of electron-holes recombination. Instead of using an external voltage, piezoelectrics materials are used to produce an internal voltage under mechanical vibration. The most extensively used are lead based piezoelectric materials such as lead zirconate (PZT). However, PZT contains about 80 % of the lead (Pb) content thus limiting their use in various applications (Panda and Sahoo, 2015). Due to lead being toxic, it is very important to develop piezoelectric materials which are lead free. Over the past few years, barium titanate (BaTiO3) has been given more attention as a lead free piezoelectric material for the production of piezoelectricity under mechanical vibration. Furthermore, recently BaTiO3 has been widely used as one of the piezo semiconductors in piezo-photocatalytic wastewater treatment applications.Piezo-photocatalytic experiments using powder catalyst involves dispersing a certain amount of the catalyst into a contaminated solution. Since powder catalyst offers high surface to volume ratio, the solution mixture consisting of the catalyst is stirred for a certain time, normally for 30 min to reach an adsorption–desorption equilibrium in the dark. Thereafter, the solution mixture gets exposed to a light source. Some of the important parameters which need to be considered during conducting piezo-photocatalytic experiments in suspension systems includes the type of the light source (solar or UV light), UV light power, Ultrasonic power, UV intensity, the amount of the material used (dosage), reaction time and pH of the solution. Recently, a majority of the piezoelectric semiconductors such as BaTiO3 have been modified to convert their absorption from UV region to visible region (reduce their band gap) to utilize the visible light as a source of light, which constitutes of 43 % of the solar energy. For instance, the activity of piezoelectric semiconductors like ZnO and BaTiO3 were tested under different light sources such as sunlight and artificial visible light (Xenon lamp 1000 W, which emits visible light in the wavelength between 400 and 800 nm). Under solar light irradiation, the total organic carbon (TOC) results showed complete mineralization of phenol at lower concentrations as compared to artificial visible light irradiation (Pardeshi and Patil, 2008).The type of the vibration normally employed in piezo-photocatalysis process is ultrasonic vibration. Ultrasonic excitation, can be used to induce piezoelectric materials to produce piezoelectric potential, which can effectively encourage the deterioration of organic dyes. However, under stress the generated free carriers will move in a specific direction to their end two poles and shield the piezo-potential, reducing the driving force. As a result, to maintain the electric field during the piezocatalysis process, continual oscillation is necessary. The ultrasound has the capacity to deliver continuous stress as a physical expression of mechanical energy (Lu et al., 2022). It is important to note that prolonged ultrasonic vibration will have both sonochemical and piezoelectric effects on materials that are made of piezoelectric components (Torres et al., 2008). The sonochemical effect can also help in the degradation of organic or inorganic wastewater pollutants.The issues associated with powder piezo-photocatalyst such as low separation efficiency, poor recovery and regeneration ability, could be resolved by fabricating piezo-photocatalysts supported on the substrate to produce thin film electrodes. Typically, powder catalysts are separated from aqueous solution via filtration and centrifugation process, thus time consuming and some of the catalyst residue might remain in the solution and lead to secondary pollution. Piezo-photocatalyst thin film electrodes offer a good recoverability and recyclability, unlike powder catalysts. However, thin film electrodes during degradation process do not offer the full contact with the solution as compared to powder catalyst. Due their limited contact with the solution or low surface area, thin films exhibit slow degradation rate and low degradation efficiency. Besides that, growing interest is being shown in thin films with nanostructures that are directly formed on the surface of the substrate and are particularly susceptible to exposure to the dye solution. The degradation of organic pollutants via piezo-photocatalytic processes can be illustrated in Fig. 3 . As shown in the Fig. 3 experiment, the prepared piezo-photocatalyst thin film is dispersed into a solution containing organic pollutants, thereafter exposed to light and ultrasonic irradiation. Just like piezo-photocatalytic experiment in suspension system, the parameters such as; the distance between the thin film electrode and light source, distance between thin film electrode and ultrasonic probe, ultrasonic power and light source need to be considered. Recently, floatable thin films are designed, which freely moves atop the water offering better utilization of sunlight. Unlike, steady thin film which requires photoreactor and external light source. Furthermore steady thin film requires a specific platform to control the distance between the light source and thin film electrode, which obviously raises the cost of scalable water purification (Yaozhong Zhang et al., 2019).Another form of using thin film electrode is via sono (piezo)-photoelectrocatalytic degradation processes. Sono(piezo)-photoelectrocatalytic processes is a combination of sonocatalysis/piezocatalysis, photocatalysis and electrocatalysis. These processes have not yet been extensively investigated. In this experiment, light irradiation, ultrasonic vibration and bias voltage is applied on the surface of the thin film electrode. The degradation experiment is conducted using potentiostat/galvanostat, the prepared piezo-photocatalyst thin film is employed as a working electrode in the presence of a reference (Ag/AgCl) and counter electrode (platinum wire). Generally, the fabricated thin film electrode is positioned vertically opposite to the ultrasonic probe and light source (Fig. 4 ).BaTiO3 is one of the highly applied ferroelectric materials which exhibit piezoelectricity under any form of mechanical vibration. It belongs to the perovskites family (ABO3), whereby A denotes a Barium (Ba) atom and B is a Titanium (Ti) atom. The crystal structure of BaTiO3 consists of Ti4+ atoms co-ordinated to six oxygen atoms to produce octahedral cluster’s (TiO6) and Ba2+ co-ordinated to twelve oxygen atoms to form (BaO12) clusters (Fig. 5 ). As shown in Fig. 5, Ba atoms are situated at every corner position, O atoms at face centred positions and Ti atoms at the centre.Barium titanate can exist in different crystal structures such as cubic, tetragonal, orthorhombic and rhombohedral depending on the theta angles and phase transition temperature. The major distinction between cubic and tetragonal phases of BaTiO3 lies on the slight shift of theta angles of octahedral (TiO6) clusters from 90° to ≈ 93.3°, whereas the orthorhombic and rhombohedral phase occurs in the theta angles from approximately 89.9° to ∼ 85.7°(Itoh et al., 1985). The BaTiO3 crystal structures undergoes three different phase transitions under different temperatures. At temperature between 26.85 °C and 46.85 °C, cubic crystal structures transform into tetragonal structures, and to orthorhombic at approximately –23.15 °C to 6.85 °C, then ultimately to rhombohedral at temperatures around −73.15° C and –33.15 °C (Acosta et al., 2017; Oliveira et al., 2020). The band energy gap of each crystal structure of BaTiO3 plays a significant role in the photocatalysis process. The cubic crystal structure of BaTiO3 has a theoretical direct band energy gap of 4.68 eV, while orthorhombic, tetragonal and rhombohedral exhibit an indirect band energy gap of 5.06, 4.73 and 5.06 eV, respectively (Oliveira et al., 2020; Piskunov et al., 2004). Amongst these crystal structures, tetragonal-BaTiO3 (t-BaTiO3) has the lowest band energy gap than other phases (orthorhombic and rhombohedral). Owing to its low energy band gap (t-BaTiO3) when compared to other phases, this makes it a suitable piezo semiconductor for photocatalytic degradation of organic waste pollutants present in wastewater. Moreover, due to its well-positioned valence band, it also an important material in water splitting for hydrogen production.Furthermore, BaTiO3 has a wide band energy gap just like other metal oxides such as TiO2, ZnO, SnO2, WO3 and BiVO4 etc., and its photoactivity is limited by the recombination of photogenerated charge carriers (e- and h+), which occurs rapidly (Demircivi and Simsek, 2019). However, unlike normal semiconductors, BaTiO3 is considered also as a piezo semiconductor which produces an internal piezo electric field under mechanical vibration. The induced built-in piezoelectric field separates the photogenerated charge carries, thus reducing their rate of recombination (X. Liu et al., 2020).Several methods have been proposed to improve the photocatalytic performance of other metal oxide semiconductors such as metal/non-metal doping, formation of several metal oxide based composites, synthesis tailoring to attain certain morphology with improved photocatalytic activity and heterojunction formation with other semiconductors (Ray et al., 2021). These modification methods have been reported to help spatial charge separation and mitigate against fast recombination of photogenerated holes. However, to achieve effective degradation performance, it is proposed that the surface of the semiconductor must be loaded with a reduction cocatalyst and oxidation cocatalyst to achieve long live charge separation and speed up photogenerated hole transfer (G. C. Zhang et al., 2019). The next section, hence discusses the method of preparation and other modification strategies researchers have adopted to improve the performance of BaTiO3. Since the discovery of BaTiO3 during World War II (1941–1944) (Bouzidi et al., 2019), there has been a progressive development of BaTiO3 using different preparation approaches including sol–gel, hydrothermal/solvothermal, co-precipitation, mechanochemical and solid-state method. These synthesis methods have an impact on the physical and chemical characteristics of BaTiO3. Thus, it is critical to select appropriate preparation methods, since piezo-photocatalytic activity greatly depends on the physical and chemical properties of BaTiO3.Hydrothermal synthesis is one of the popular methods for the fabrication of powdered BaTiO3 since it is not expensive and can form stable and pure materials. This method involves a reaction between Barium (Ba) and Titanium (Ti). During their synthesis the most widely used precursors include barium chloride (BaCl2), barium hydroxide (Ba(OH)2), TiCl4 and TiO2 materials. The hydrothermal reactions take place in an autoclave at temperatures above 100 °C. Several parameters including reaction time, temperature, solvents and solution pH can influence the morphology, particle size and crystal structure of BaTiO3. Xia et al. prepared BaTiO3 nano/microcrystals using commercial titanium dioxide (TiO2) and Ti(OH)4 as a titanium (Ti) precursor mixed with Ba(OH)4 as barium (Ba) precursor (Xia et al., 1996). A very well crystalline and dispersed BaTiO3 with a crystallite size < 100 nm was formed when Ti(OH)4 gel and Ba(OH)2 solution were used as precursors. The results showed that the starting precursors also had a strong impact on the morphology and crystallite size of the prepared BaTiO3. Furthermore, the hydrothermal reaction temperature had a strong influence on the crystal structure. As shown in Table 1 , the lattice constant “a” slightly decreased with an increase in reaction temperature. According of the study conducted by Wen et al., it was found that lattice parameter a can affect photocatalytic activity of the semiconductor (Wen et al., n.d.). The photocatalyst (TiO2) anatase material with same composition, morphology, phase, and surface states but different lattice parameter ‘’a’’ were employed for photocatalytic degradation and photo-reduction of toluene and Cr(VI), respectively. However, greater catalytic activity was achieved by TiO2 with the extended lattice parameter than standard TiO2. Increasing in the length of the lattice parameter ‘’a’’ caused the bottom of the TiO2 conduction band to move higher, thus improving its photocatalytic activity.Moreover, Habib et al. showed that the structural morphology of the powdered BaTiO3 was temperature dependent (Habib et al., 2008). According to their result, the hydrothermal BaTiO3 obtained at low temperature (90 °C) had less pores compared to those attained at 120 and 150 °C. The study involving the relationship between photocatalytic activity of BaTiO3 thin film with porosity and surface area was conducted by Augurio et al.(Augurio et al., 2022). The porous BaTiO3 thin films exhibited higher photocurrent response than non-porous BaTiO3 thin film, indicating that porosity is beneficial in photocatalysis. This suggested that porous BaTiO3 can enhance interfacial charge transfer whilst lowering the charge carrier recombination rates, thus improving the photocatalytic activity. In another study, Zhan et al. controlled the hydrothermal reaction time (from 15 min to 480 h) to obtain BaTiO3 nanoparticles (Zhan et al., 2012). The XRD results showed no diffraction peaks after 15 min of hydrothermal reaction, demonstrating that the material lacked crystalline phases. However, longer hydrothermal reaction times (20 min to 48 h) led to the appearance of diffraction peaks in the XRD patterns that were attributed to the cubic BaTiO3. A continuous increase in the intensity of the diffraction peaks was observed with an increase in reaction time, demonstrating a persistent rise in the crystallinity and size of the crystals. Surmenev et al. produced BaTiO3 nano and micro rods via the hydrothermal method. The BaTiO3 nano- and micro rods were obtained at a temperature of 160–210 °C, using 0.02 and 0.15 M (NaOH) concentration and within 45–90 min (Surmenev et al., 2021). The XRD results showed that BaTiO3 purity drastically increased as NaOH concentration increased from 0.025 to 0.15 M. Furthermore, BaTiO3 tetragonal phase was clearly visible after 6 hrs of hydrothermal synthesis at 210 °C and varied alkalinities (from 0.025 to 0.15 M), whereas 45 and 90 min produced a combination of cubic or tetragonal phases. The results showed that the hydrothermal reaction conditions such as temperature, alkalinity and time, have a great impact on the formation of BaTiO3 structures with different morphologies. Wei et al. controlled the size of BaTiO3 nanoparticles via hydrothermal approach with Fe doping and ethylenediamine (en) addition (Wei et al., 2008). The crystal size of the synthesized BaTiO3 were investigated by X-ray powder diffraction (XRD), transmission electron microscopy (TEM), scanning electron microscope (SEM) and high-resolution transmission electron microscopy (HR-TEM). The results showed that BaTiO3 crystal size decreased as it was doped with Fe, indicating that Fe-doping suppress the crystal growth. It was further noticed that as Fe doping concentration increases, the average particle size also decreases (Fig. 6 ). Additionally, the addition of en, which served as both a solvent and a capping agent, may inhibit particle growth and cause a contained effect that changed the shape of the particles from spherical to cubic. It has been reported that semiconductors with smaller particle sizes have excellent photocatalytic activity as compared to those with large particles.BaTiO3 heterostructures are easily fabricated using the hydrothermal method. Li et al. prepared BaTiO3/TiO2 heterostructure nanotube arraysusing a straight forward hydrothermal process, the hydrothermal reaction was carried out at different reaction times, temperature and concentration of Ba(NO3)2 (R. Li et al., 2013). Zhao et al. and Kappadan et al. demonstrated the preparation of Ag2O/BaTiO3 and BaTiO3/ZnO heterostructures, respectively, using hydrothermal method (Zhao et al., 2020)(Kappadan et al., 2020a). Based on their experimental results, BaTiO3 nanoparticleswere anchored on hexagonal rod-shaped ZnO (Kappadan et al., 2020a). The combination of hydrothermal and microwave method could be used to fabricate BaTiO3 (Sun et al., 2006). For example, Amaechi et al. prepared Fe-doped BaTiO3 via ultrafast microwave-assisted hydrothermal method (Amaechi et al., 2021). Furthermore, the hydrothermal approach could be used with the electrospinning method. By combining an electrospinning and a hydrothermal technique, Ren et al. developed ZnO/BaTiO3 nanofiber heterostructures (Ren et al., 2012). As seen in Fig. 7 (a), BaTiO3 nanofibers had a rather smooth surface and formed a network topology. BaTiO3 nanofibers ranged from 300 to 400 nm in diameter and up to several micrometers in length. The ZnO nanoparticles were uniformly dispersed on the rough surface of BaTiO3 nanofibers (Fig. 7(b)). The elemental composition of the pure BaTiO3 nanofibers and ZnO/BaTiO3 nanofiber heterostructures showed the presence of Ba, Ti, O and Zn. The detected Al element was from aluminium foil.The sol–gel method is another simple method that is used to prepare barium titanates such as BaTiO3, BaTi4O9, Ba2TiO4 and BaTi2O5. Normally, barium acetate (Ba(OAc)2 and titanium (VI) isopropoxide (C12H28O4Ti) are used as barium and titanium precursors, respectively. The mixture of titanium (VI) isopropoxide and barium acetate solution tends to form BaTiO3-gel which is further dried and calcined at high temperatures (400–1200 °C) (Kavian and Saidi, 2009). The synergic sol gel and template method was used to fabricate BaTiO3 nanotubes (Cao et al., 2006). The formation of BaTiO3-gel was attained via mixing (Ba(OAc)2 and titanium isopropoxide, thereafter the nanostructured BaTiO3-gel grew on the porous alumina membrane (pore size 200 nm). The resultant alumina template covered with BaTiO3-gel was calcined at 700 °C to form BaTiO3 nanotubes with 50 µm length. The calcination temperature had a significant impact on the BET surface area of the BaTiO3. Pffaf (Pfaff, 1992) indicated that the BET surface area of BaTiO3 prepared via sol–gel method decreased as the calcination temperature increased (Table 2 ). The high specific surface was obtained when BaTiO3-gel was calcined at low temperature (200 °C). It has been reported that at elevated temperatures, nanoparticles tend to agglomerate extensively thus resulting in a significant reduction in BET surface area and pore diameter (Zhang et al., 2015).However, XRD diffraction showed highly pure crystalline BaTiO3 obtained at higher temperature. At high calcination temperatures above 800 °C, the crystal structure of BaTiO3 transformed from cubic to tetragonal structure. This crystal structure transformation (cubic to tetragonal) was depicted by XRD peak splitting at 2θ value approximately 45° (Fig. 8 ) [70]. From this study, it can be noted that crystallinity did have an influence on the optical properties of the semiconductors. According to literature, the optical band energy gap of the semiconductors decreases with an increase in crystallinity. Nishioka and Maeda (Nishioka and Maeda, 2015) studied the influence of the post heating of the hydrothermally synthesized Rhodium-doped barium titanate (BaTiO3:Rh) which could increase crystallinity and further improve photocatalytic activity. The XRD patterns were stronger and narrower after post heating 900 °C, thus confirming crystallization. However, the specific surface area was reduced from 8 to 4 m2 g−1. UV–vis diffuse reflectance spectroscopy (DRS) was employed to investigate the optical properties of BaTiO3:Rh. Upon post heating, DRS exhibited a series of changes as the temperature increased (with exception of the sample at 1150 °C). At higher temperatures, Rh4+ species induced greater absorption at longer wavelengths. This would make sense because at a high-temperature heat treatment increased the oxidation of Rh3+ to Rh4+ in BaTiO3:Rh. In terms of photocatalytic activity, unheated samples tend to show low activity, while on the other hand the activity increased with an increase in post heating temperature until 1000 °C. At elevated temperature above 1000 °C, the photocatalytic activity of BaTiO3:Rh decreased drastically.Sol gel was combined with low temperature hydrothermal reaction procedure to prepare BaTiO3 nanopowder (Wang et al., 2013). In the study reported by Wang et al., it was found that experimental conditions such as potassium hydroxide concentration (KOH), reaction temperature and time had a significant role on the crystallinity and morphology of BaTiO3 powder (Wang et al., 2013). A highly crystalline pure BaTiO3 with a cubic structure was obtained at 120 °C (after 2 h of reaction time) with KOH concentration over 1.0 M. The hydrothermal and reaction time showed less effect on the crystallinity and morphology, whereas KOH concentration showed a significant impact on the crystallinity and morphology. It was observed that, when the KOH concentration rised from 1.0 M to 8.0 M, the average size of the BaTiO3 particles decreased from 370 nm to 100 nm.Solid state synthesis is a common method which is usually employed to produce polycrystalline materials such as barium titanate (BaTiO3). This method requires a very high temperature, however its benefits include simplicity and high yield production. The main factors which affect solid state reaction include reaction temperature, pressure, chemical and morphological properties of the starting reagents/materials. The solid state synthesis of BaTiO3 nanoparticles was reported by Qi et al. (Qi et al., 2020). In their studies, different molar ratios of Barium nitrate (Ba(NO3)2) and Ti powder (Ba/Ti) as starting materials were varied and calcined at different temperatures (500, 550 and 600 °C). The calcination temperature played a crucial role in the formation of BaTiO3. This was confirmed by the XRD pattern which showed that there was no formation of BaTiO3 at temperatures below 500 °C, since only XRD peaks for starting materials (Ba(NO3)2 and Ti) were revealed (Fig. 9 ). At high temperature (600 °C), the peaks were almost indexed to BaTiO3 material, thus now confirming the effect of calcination temperature on these materials. Other studies reported the thermal decomposition reaction of barium carbonate (BaCO3) and titanium dioxide (TiO2) for the formation of BaTiO3 (Pithan et al., 2005). Since the rate of reaction is controlled by the diffusion rate of Ba ions into Titanium dioxide (TiO2) lattice, the shape and size of the BaTiO3 produced was more influenced by the TiO2 morphology. The formation of titanates were explained in detail in the literature (Beauger et al., 1983). Trzebiatowski et al. reported that the formation of barium titanate (BaTiO3) and barium orthotitanate (Ba2TiO4) occurs simultaneously via the below chemical reaction (Brdi et al., 1950); (1) BaCO3 + TiO2 → BaTiO3 + CO2 (2) 2BaCO3 + TiO2 → Ba2TiO4 + 2CO2 In other studies they have reported that Ba2TiO4 forms when BaTiO3 reacts with TiO2 as shown in equation (4), thereafter the formed Ba2TiO4 reacts with the remaining TiO2 to produce meta titanate as shown in equation (5) (Beauger et al., 1983). (4) BaTiO3 + BaCO3 → Ba2TiO4 O3 Ba2TiO4 + TiO2 → 2BaTi Solid state reaction method could be combined with sol–gel method. Mi et al. prepared nano BaTiO3 ceramics using TiO2 precursor gel and BaCO3 as starting raw materials (Mi et al., 2020). The XRD results showed the initial formation of BaTiO3 at calcination temperatures of 600 °C. A cubic BaTiO3 structure was formed when the calcination temperature reached 800 °C. At 900 °C calcination temperature, the diffraction peak of (200) separated into peaks of (002) and (003), thus suggesting phase transition from cubic to tetragonal phase. In another experiment, Ren et al. used a solid state method to fabricate Bi2O3/BaTiO3 heterostructure (Ren et al., 2013). Firstly, BaTiO3 was prepared from Ba(CH3COO)2 and TiCl4 via the hydrothermal treatment process. Thereafter, Bi2O3/BaTiO3 heterostructure were prepared through ball milling and calcination process using the prepared BaTiO3 and commercial Bi2O3 (mass ratio BaTiO3: Bi2O3 = 4:1). After the calcination procedure, it was discovered that Bi3+ had dissolved in the BaTiO3 lattice and that a chemical connection had been created at the interface between Bi2O3 and BaTiO3.Recently sound energy has been utilized to prepare different metal oxide semiconductors such as BaTiO3 for different applications. In contrast to basic reactions, ultrasound-assisted reactions actually have a lot of advantages. High pressure, low pressure, and localized boiling zones are all produced by ultrasound in the reaction mixture. This shortens the reaction period and makes room-temperature synthesis possible. It has been noted that ultrasonography facilitates the uniform dispersion of reactants in a reaction mixture. Dang et al. reported sonochemically synthesized BaTiO3 nanoparticles (Dang et al., 2011). In their study, mixtures of ethanol and distilled water were prepared with different volume ratios. Thereafter, BaCl2 and TiCl4 (molar ratio Ba:Ti = 1:1 were added to the above solution mixture, followed by addition of NaOH. The solution suspension was exposed to ultrasonic irradiation for 40 min at low temperature (50 °C). The applied ultrasonic energy was 150 W/cm2. Following synthesis, the precipitate was centrifugally separated, twice washed with deionized water, and then dried for two hours in a vacuum at 100 °C. In another study, BaTiO3 submicronic particles were prepared following multiple procedures such ultrasonication, microwave drying and thermal treatment (Rotaru et al., 2017). Mixture of BaCO3 and TiO2 as raw materials were ultrasonicated (ultrasonic frequency: 20 kHz, 750 W nominal electric power) in milli-Q ultrapure water. After 30 and 60 min of ultrasonication, the prepared samples were dried in the microwave furnace for 10 min. The last procedure was thermal treatment of the samples at different temperatures (780–1300 °C) for 3 hrs. Ashiri et al. reported similar approach to obtain BaTiO3 nanocrystals via rapid ultrasound-assisted wet chemical method (Ashiri et al., 2015). Utara and Hunpratub synthesized cubic structure of BaTiO3 nanoparticles using ultrasonic method at room temperature without thermal treatment step (Utara and Hunpratub, 2018a). The starting precursors were barium hydroxide (BaOH)2 and diisopropoxytitanium bis(acetylacetonate) (C12H28O6Ti). The effect of ultrasonic reaction time on the morphology of BaTiO3 nanoparticles (NPs) was investigated using TEM micrographs. It was found that the particle sizes of the BaTiO3 NPs decreased with increase in ultrasonic reaction time. The average particle size reduced from 56.69 ± 30.14 nm (30 min of ultrasonic irradiation) to 32.72 ± 11.83 nm (4 hr of ultrasonic irradiation). Similar observations were reported by Moghtada and Ashiri (Moghtada and Ashiri, 2016). It was concluded that smaller particles were produced as a result of ultrasonic irradiation at 50 °C.Co-precipitation method is the most frequently utilized synthesis approach for metal oxides (Rao et al., 2017). This method involves dissolving of metals salts in an appropriate solvent, followed by the addition of a precipitating agent. The most widely used precipitating agents include sodium hydroxide (NaOH), ammonium hydroxide (NH4OH) and potassium hydroxide (KOH). In case of preparing BaTiO3 using oxalate co-precipitation, it is challenging to obtain optimal conditions where both Barium (Ba) and Titanium (Ti) precipitates at the same time. Since Titanium (Ti) precipitates as titanly oxalate in the presence of alcohol at pH ≤ 2 whereas Barium (Ba) precipitates as BaC2O4 at pH ≥ 4. Titanium generates soluble anionic species such as TiO(C2O4)2 2– in the pH between 2 and 4, thus influencing the stoichiometry ratio of Ba/Ti simultaneously (Geetha et al., 2016). It has been reported that through manipulation of several chemical conditions such as pH, reactants, and reaction medium, it is possible to make Ba and Ti to precipitate at the same time. Prasadarao et al. investigated the influence of pH (range 2–10) on the synthesis of BaTiO3 from barium chloride (BaCl2) and potassium titanyl oxalate (KTO) (Prasadarao et al., 2001). The formation of barium titanyl oxalate was obtained at pH 2.5 and an increase in pH to 5 led to the formation of barium titanyl hydroxy oxalate. At higher pH values (7–9), precipitation reactions yielded a mixture of titanium dioxide (TiO2) and barium oxalate (BaC2O4). He et al. also prepared BaTiO3 powder via the co-precipiation of BaCl2 and TiOCl2 in an highly-alkaline environment (He et al., 2014). The pH solution and concentration of the starting precursors (BaCl2 and TiOCl2) had an effect on the particle grain size and homogeneity of the BaTiO3 powder. An average particle size of approximately 80 nm was obtained at pH 14 and reaction temperature of 80 °C. In another study, Zhang et al. used BaCl2, TiCl4 as starting raw materials and tartaric acid as a precipitant agent for the preparation of tetragonal BaTiO3 nano-powder (X. Zhang et al., 2021). The white precipitated were formed by adding slowly a solution of ammonium hydroxide solution. Followed by thermal treatment at different calcination temperatures (750–1050 °C) for 4 hrs. The microwave assisted co-precipitation was reported to produce BaTiO3@rGO nanocomposite (Khan et al., 2021a). BaTiO3 and GO as starting materials were prepared separately via the sol–gel method and modified Hummers method, respectively. Thereafter, a certain amount of BaTiO3 and rGO were added to 50 ml of deionised water and stirred for 1 hr at room temperature. Then, NaOH solution was slowly added to the above mixture solution, and heated for 1 hr in a microwave oven. The reduction of GO into rGO was confirmed by color change of the solution from brown to black. The co-precipitated nanocomposite was washed with mixture of ethanol/water and dried at 60 °C in an oven for 12 hr. The TEM images of pure BaTiO3 and BaTiO3@rGO are shown in Fig. 10 (a-b). As shown in Fig. 10(a), pure BaTiO3 exhibits spherical nanoparticles with a size distribution of 10–30 nm, whereas Fig. 10(b) shows spherical BaTiO3 nanoparticles with an average particle size range of 15–34 nm, which are uniformly distributed on the surface of rGO sheet. Table 3 highlights the summary of some of the synthetic methods for BaTiO3 powder.These techniques are mostly applied to prepare powdered BaTiO3, however powdered piezo-photocatalyst are difficult to be recycled in practical applications. For instance, after the degradation process, some parts of the powdered catalyst may persist in the aqueous solution, thus leading to secondary pollution. Therefore, recently piezo-photocatalyst based thin films are being developed for better recoverability, thus in the next section some common techniques that are used to produce BaTiO3 based thin films will be highlighted.There are various methods implemented for the preparation of BaTiO3 thin films, these include physical and chemical techniques. The physical methods include sputtering deposition, pulsed laser deposition (PLD), spin coating, dip coating and the Dr Blade method (Asadzadeh et al., 2021; Cernea, 2004). Chemical methods include chemical vapour deposition (CVD), sol–gel method and hydrothermal method. All of these have their own advantages and disadvantages. Cernea et al. explained most of these methods basic principle and their own benefits (Cernea, 2004). In this review, a few common physical and chemical methods are discussed below.Dip coating is one of the most popular liquid-phase deposition methods for the fabrication of thin-films. This method involves dipping a substrate in the solution containing a starting material/ceramic powder, binder, solvent and dispersant. Once the material of interest has been deposited, the substrate is removed slowly from the solution and dried at ambient temperature. Several parameters such as immersion period, withdrawal rate, number of immersion cycles, solution composition, concentration and temperature tends to affect the film characteristics, smoothness and thickness (Schneller et al., 2013). This method has been used for the production of numerous piezoelectric thin-films including Pb (Zr, Ti)O3, CaBi4Ti4O15, ZnO, PVDF and BaTiO3. Ashiri et al. reported a crack-free nanostructured BaTiO3 produced from a modified sol–gel dip coating method (Ashiri et al., 2014). The silica substrate was immersed into a sol prepared from barium acetate, glacial acetic acid, titanium tetraisopropyl alkoxide and 2-propanol. After deposition, the substrate with coated BaTiO3 was taken out from the sol–gel solution with a withdrawal rate of 1 cm/min and dried at 100 °C. The resultant substrate coated with BaTiO3 was further calcined at 800 °C for 1 hr (heating rate 5 °C/min) to produce a thin film with a thickness of approximately 2 nm.In the spin coating process, the coating material is firstly dissolved in an appropriate solvent and the solution is dropped at the centre of the solid substrate surface. The solid substrate is then spun at controlled high speed. During this process, the solid substrate is rotated around an axis which is perpendicular to the coated region. The thickness and other properties of the final thin film depends greatly on the spinning rate of the substrate, viscosity of the solution, solvent evaporation rate, spinning time and surface wettability. This method is suitable and can be used for fabrication of several ceramics, including barium titanates such as BaTiO3 (Aminirastabi et al., 2020).Chemical vapour deposition is a widely used method to produce 2D nanomaterials and thin films. In this process, a solid material is deposited from the vapour by a chemical reaction occurring on or in the vicinity of a typically heated substrate (Mittal et al., 2021). The thin film nanostructures and thickness can be tuned by controlling the deposition conditions and the CVD system key factors. These include the substrate material and precursors, composition of reaction gas mixture, total pressure gas flows and temperature. Suzuki and Kijima (Suzuki and Kijima, 2006) prepared nanostructured BaTiO3 thin film from bis-dipivaloylmethanate barium (Ba(DPM)2) and titanium (IV) isopropoxide (Ti(OiPr)4 deposited on platinum/alumina/silica/silicon (Pt/Al2O3/SiO2/Si) substrate using the CVD technique assisted with Inductively Coupled Plasma (ICP). The size of the resultant nanostructured BaTiO3 thin film was greatly influenced by substrate temperature. The single phase BaTiO3 structure and particles sizes of approximately 30 nm were successfully obtained at temperature of 500 °C (Fig. 11 (a)). The deposited spherical BaTiO3 nanoparticles on the surface of the substrate were more agglomerated with less pores. Fig. 11(b) shows a cross section image of the deposited dense nanoparticles on the substrate surface, however the thickness of the thin film was not determined. At substrate temperatures above 600 °C, the deposited BaTiO3 nanoparticles fused into a columnar form as shown in Fig. 11(c)-(d). The bottom of the thin films formed, exhibited a columnar structure, whereas the structure surrounding the surface was made of nanoparticles.Tape casting known as the Dr.Blade method has been widely used for the production of ceramic thin films. This technique is usually used to obtain thin films with a thickness ranging from 10 to 1000 μm. In this method, the powdered starting materials are mixed with appropriate solvents and binders to form a homogeneous mixture which is tape casted on the solid substrate (Asadzadeh et al., 2021). Thereafter, the tape casted substrate is dried at certain temperatures. The drying rate and temperature tend to be the most important factors which control the crack free of the thin film. Other factors which can affect the thin film properties and thickness include relative content of ceramic powder (starting materials), solvent and binder. Lilge et al. hydrothermally synthesized BaTiO3 powder from BaCl2.8H₂O and Ti [OCH(CH3)₂]₄ (Lilge et al., 2020). The hydrothermal reaction place was performed in a microwave for 120 min at 140 °C. The resultant BaTiO3 powder was further used to prepare a photoanode electrode using the Dr Blade method. For the preparation of the photoanode electrode, the powdered BaTiO3 was mixed with ethylene glycol Triton X-100 and ethanol. The slurry mixture was then taped casted on the FTO substrate (area of 1 cm2) to form a thin film. The BaTiO3 pasted on the surface of FTO appeared to be spherical in shape and agglomerated (Fig. 12 ).Despite the fact that BaTiO3 as a semiconductor has received a lot of attention for piezo-photocatalytic applications due to its incredible ferroelectric/piezoelectric properties and accessibility in a wide assortment of sizes and morphologies, it has significant limitations, most which are linked to its photocatalytic activity (X. Liu et al., 2020; Ray et al., 2021). Owing to its wide energy band gap of approximately 3.2 and 3.4 eV, it is associated with rapid recombination of photogenerated electron-holes which reduces its photocatalytic activity. Recently, various strategies have been explored including tailoring the morphology and particle sizes, doping and fabrication of heterojunction/composite photocatalyst to prevent some of these limitations (Scheme 1 ).The surface morphology and particles size of the semiconductor photocatalyst/piezocatalyst plays an important role in the catalytic degradation of organic waste pollutants. It has been reported that BaTiO3 with different morphological structures including nanowires, nanofibers, nanorods, nanotubes, nanocubes and nanoparticles exhibits different piezoelectricity. For example, 1-D fiber/wire piezoelectric materials show a superior piezocatalytic response as compared to spherical particles. Whereas, thin sheet-like 2-D structures also generate more piezoelectricity under mechanical vibration (Mondal et al., 2022). As piezocatalyst, Liu et al. explored different nanostructures (nanocubes (NCs), nanoparticles (NPs) and nanofibers (NFs)) of BaTiO3 for piezocatalytic degradation of Rhodamine B (Rh B) (D. Liu et al., 2020). BaTiO3 nanofibres showed greater piezocatalytic performance as compared to nanocubes (NCs) and nanoparticles (NPs) due to a higher surface area, easy deformation structure and fine crystal size. Moreover, Jiao et al. prepared different BaTiO3 nanostructures via the hydrothermal route at different reaction times (starting from 4 to 16 hr) (Jiao et al., 2017). Spherical BaTiO3 nanoparticles formed at 4–8 hrs were more effective for photocatalytic degradation of Rh B dye than other morphological nanostructures such as bowl like and agglomerated spherical particles. To further understand the enhancement in photocatalytic activity due to morphological-tuning of the semiconductor photocatalyst, different analyses including BET surface area, photon energy, electrochemical Impedance Spectroscopy (EIS), photoluminescence (PL) and photocurrent also need to be conducted. Since the method of preparation has an impact on the final morphological product, various authors have also synthesized these piezomaterials using varying methods Xiong et al. fabricated BaTiO3 nanocubes, since cubic structures are known to have the best ability to reduce crystal defects and increase the surface-to-volume ratio (Xiong et al., 2015). The effect of reaction time (24, 48, 74 hr) using the hydrothermal method was employed to produce the cubic like BaTiO3 structure (Fig. 13 (a)-(f)). The particle size increased with an increase in hydrothermal synthesis duration. Furthermore, the edges of the cube got sharper as the reaction time increased, showing an increase in the cubic phase's crystallinity. The BaTiO3 nanocubes formed over period of 48 hrs exhibited impressive photocatalytic performance under light irradiation. This better performance was due to more uniform morphological distribution, higher crystallinity, small particle size and higher surface area which lead to more active sites, reduction in migration path of charge carriers, narrowing the energy band gap and reducing the rate of charge carrier’s recombination.The optical properties of the BaTiO3 nanocubes calcined at different temperature were investigated by photoluminescence (PL) and UV vis spectrophotometer. The calcination temperature had an effect on the optical properties of the hydrothermally synthesized BaTiO3 reported by Hasbullah et al.(Hasbullah et al., 2019). The energy band gap calculated from tauc’s plot ( Fig. 14 ) for BaTiO3 calcined at 500, 600, 700 and 1000 °C were 3.18, 2.87, 2.83 and 2.74 eV, respectively. Upon increasing the calcination temperature, the energy band gaps of BaTiO3 were expected to increase due to their high crystallinity. However, BaTiO3 resulted in lower energy band gap than expected. This could be due to inadequate oxygen delivery during the calcination process in ambient air resulted in oxygen deficiency in BaTiO3 structures (Orhan et al., 2005). As a result, the calcined BaTiO3 had a greater density of oxygen vacancy or non-bridging oxygen. The oxygen vacancy has the potential to change the BaTiO3 structure and cause localized electronic states. Therefore, resulting in reduction of energy band gaps for extremely crystalline BaTiO3 structures.Moreover, photoluminescence (PL) was employed to study the rate of photogenerated electrons and holes recombination. As seen in Fig. 15 (a)-(f), the PL intensity was expected to decrease with increase in crystallinity. However, in this study the PL intensity reached its highest peak as the calcination temperature was elevated to 1000 °C. It was hypothesized that the increase in photoluminescence intensities is due to the presence of a localized state within BaTiO3 structures. With sufficient stimulation, the localized state effectively lowers the band gap of BaTiO3 structures, hence resulting in strong photoluminescence intensity.Metal or non-metal doping is one of the most popular methods used to modify semiconductor photocatalysts to improve their optical properties such as a reduction of band gap and photogenerated charge carriers (electron-holes), increase in photocurrent response and interfacial charge carries. The improvement of these properties tends to enhance the photocatalytic and piezo-photocatalytic degradation of organic wastewater pollutants. It has been reported that the modification of photocatalyst semiconductors such as TiO2 (Khairy and Zakaria, 2014), ZnO (Kaur and Singhal, 2014), WO3 (Peleyeju and Viljoen, 2021), BiVO4 (Orimolade and Arotiba, 2020)and BaTiO3 (Ray et al., 2021) by metal ion doping can successfully shift their optical absorption to the visible light region, thus narrowing their band gaps. Recently, a lot of research has shifted towards metal doping rather than no-metal doping since metal doping synthesis is easily achievable. To date, many transition metals including copper (Cu), Iron (Fe), Manganese (Mn), Tungsten (W) and Cerium (Ce) to mention a few have been explored as BaTiO3 photocatalyst dopants for their improved break down of several organic contaminates like methylene blue (MB), tetraclycline (TC), methyl orange (MO), and atrazine. Among these transition metal dopants, Cu has been shown to be the most efficient BaTiO3 dopants due to the fact it has shown greater improvement in degradation of organic pollutants as compared to Mn-, Fe-, Ce-, W- and Cr doped BaTiO3 (I. C. Amaechi et al., 2019; Ifeanyichukwu C. Amaechi et al., 2019; Basaleh and Mohamed, 2020; Nageri and Kumar, 2018; Senthilkumar et al., 2019). Basaleh and Mohamed (Basaleh and Mohamed, 2020) investigated the degradation activity of undoped and cu-doped BaTiO3 for the removal atrazine from wastewater. According to their outcomes, 5 wt% Cu/BaTiO3 showed the highest degradation efficiency of 100 % after 60 min, which was 33 times better compared to the undoped BaTiO3. The addition of Cu to the BaTiO3 surface reduced the band gap of undopoed BaTiO3 sample from 3.28 to 2.77 eV, thus improving the photocatalytic activity of the Cu-doped BaTiO3 sample.Noble metals including gold (Au), silver (Ag), platinum (Pt) and palladium (Pd) have been shown to improve the BaTiO3 piezo-photocatalyst sensitivity either under visible light or ultrasonic vibration. These noble metals are receiving more attention from researchers because of their superb utilisation of the solar spectrum, from visible to infrared, through the SPR effect (Chao et al., 2020; Cui et al., 2013). The BaTiO3 plasmonic photocatalyst have been fabricated from doping these noble metals with pure BaTiO3 sample. Under solar irradiation, the plasmonic photocatalyst generates an internal electric field which causes the photogenerated charge carriers to move in opposite directions. According to the charge transfer process in plasmonic photocatalysts, electrons from noble metal NPs can travel to the photocatalyst's CB and vice versa. Therefore, resulting in improved separation of charge carriers of plasmonic photocatalyst for better photocatalytic performance. In a study conducted by Xu et al., plasmonic piezo-photocatalyst (Ag/BaTiO3) compared to pure BaTiO3 showed an improved absorption under simulated solar irradiation (Xu et al., 2019). Due to this improvement, they discovered that Ag/BaTiO3 had a greater photocatalytic effectiveness than pure BaTiO3. The SPR of silver (Ag) nanoparticles resulting from internal band transitions from the 5d band to and within the 6sp band of the noble metal resulted in an increase in the piezo-photocatalytic activity of the modified BaTiO3.Another way of adjusting the band gap and improving the photocatalytic performance of the semiconductor is via non-metal doping. Unlike noble metals which are very expensive, non-metal materials are less expensive and can be applied as dopants for several photocatalyst to be used in wastewater treatment. Carbon based materials have been widely used as non-metal dopants to improve piezo-photocatalytic performance of BaTiO3 since they can improve the rate of electron transfer and also reduce the electron-hole recombination rate. Some of the widely used carbon-based materials include carbon nanotubes (CNTs), activated carbon nanofibres (ACFs), graphene oxide (GO), Biochar, and Carbon nanodots (CNDs) to mention a few (Orimolade et al., 2021a). These distinct carbonaceous materials have different morphologies (surface area and pore size) and surface chemical characteristics (functional groups, hydrophobicity, and hydrophilicity) which all have a significant role in photocatalytic degradation of waste pollutants. Over past years, several few types of these carbonaceous materials have been employed to modify BaTiO3 structure. However, the utilization of graphene oxide (GO) has been shown to be the most effective strategy. Unlike other carbons, GO offers a variety of benefits including high UV–visible light transmittance, quick electrical and thermal conductivity, superior mechanical and tribological characteristics, and corrosion resistance. In addition, the delocalization of pi (π) network of the layers effectively suppresses electron-hole recombination thus resulting in improved photocatalytic performance (Zou et al., 2019). For instance, Zhao et al. showed an improved photocatalytic performance of BaTiO3 after loading it with different mass ratios of graphene oxide (Zhao et al., 2018). Firstly, the graphene oxide was prepared from the oxidation of graphite powder using the Hummer’s method and, later the freeze drying method was employed for the preparation of graphene oxide-BaTiO3 hybrid photocatalyst (Fig. 16 (a)). Under light exposure, the hybrid material had superior photocatalytic performance than the unmodified BaTiO3, as shown in Fig. 16 (b-c). Similar observations were reported by Rastogi et al. and Wang et al., whereby the introduction of graphene oxide (GO) into BaTiO3 lattice structure resulted in a higher photoresponse under the ultraviolet region or in the visible region than BaTiO3 pristine (Rastogi et al., 2016b) (Wang et al., 2015).The combination of BaTiO3 with other several semiconductors to form BaTiO3 based heterojunction photocatalyst is another approach of improving piezo-photocatalytic efficiency of the BaTiO3 pristine. Mostly metal oxide semiconductors such as TiO2, ZnO, SnO2, Bi2O3, Bi2WO6 and Cu2O which have an unequal band gap as BaTiO3 are used to form heterojunctions (Mengying et al., 2017; Ray et al., 2021; Sharma et al., 2016; Wang et al., 2021). The formation of a heterojunction results in a band alignment which promotes the extension lifetime of the photoexcited holes and electrons within the heterostructured catalyst, thus reducing the rate of electron and holes recombination. Heterojunctions such as p-n (between a p-type semiconductor and an n-type semiconductor), n-n (between two n-type semiconductors), and p-p (between two p-type semiconductors) can be created depending on the kind of semiconductors that are combined. Furthermore, the band alignment of the heterostructured catalyst can be classified as Type I (straddling), Type II (staggered), and Type III (broken). In semiconductors, type II (including Z scheme) have been reported to efficiently improve electrons and holes separation (Orimolade and Arotiba, 2020). The charge transfer mechanism of type II was explained more in detail by Orimolade et al., Zhang et al. and Peleyeju et al. (Orimolade and Arotiba, 2020)(Zhang and Jaroniec, 2018)(Peleyeju and Arotiba, 2018). As shown in Fig. 17 , electron transfer within a heterojunction interface commonly follows a two-step pathway depending on the Femi energy level of the coupled semiconductors. On the first pathway mechanisms (Fig. 17(a)), when the Fermi energy level of SC-1 (p-type semiconductor) is smaller than that of SC-2 (n-type semiconductor), electrons (e-) migrate from SC-1 conduction band (CB) to SC-2 conduction band (CB), while holes (h+) migrate from SC-2′s valence band (VB) to SC-1′s valence band (VB). However, when the Fermi energy level of SC-1 is greater than that of SC-2, electrons (e-) from SC-2 merge with the holes (h+) from SC-1 following band alignment in the heterojunction, thus resulting in electrons and holes separation from SC-1 and SC-2 in Fig. 17(b). The accessible separated holes (h+) in SC-2 and electrons (e-) in SC-1 are responsible for piezo-photocatalytic breakdown of organic waste pollutants. This type mechanism pathway of electrons and holes separation is known also as Z-scheme (Fig. 17(b)).Several BaTiO3 based heterojunctions have been fabricated using different synthetic methods such as hydrothermal, sol–gel, solid state method and co-precipitation method for various applications, including wastewater treatment. In water and wastewater treatment, BaTiO3 have been coupled with several metal oxide semiconductors including ZnO, SnO2, TiO2, Bi2O3, Fe2O3 and MnO2 for better piezocatalytic/photocatalytic removal performance. Other non-metal oxides including g-C3N4, Ag3PO4 and AgBr have also been coupled with BaTiO3 for improved photocatalytic/piezocatalytic activity (Mengying et al., 2017; Ray et al., 2021). Feng et al. synthesized BaTiO3/SnO2 hybrid heterostructured catalyst using the hydrothermal method for piezocatalytic degradation of organic contaminates (Feng et al., 2020). The effect of SnO2 loading on BaTiO3 had a huge impact of piezo-current response, as shown in Fig. 18 (a), BaTiO3 loaded with SnO2 and SnO2-Sb generated greater piezoelectrochemical current response than pure BaTiO3. Furthermore, electrochemical impedance measurements showed the evidence of improved electron mobility via reduction in charge transfer resistance (Rct) of the composites (BaTiO3/SnO2 and BaTiO3/SnO-Sb) (Fig. 18(b)).Over the past decades, several wastewater technologies including non-destructive and destructive methods as shown in Fig. 19 , have been employed to remove toxic organic contaminates and pathogenic bacteria. Among them, advanced oxidation methods have been extensively applied as the most effective methods that accelerates the oxidation and degradation of a wide range of organic and inorganic chemicals that are resistant to traditional treatment methods. Piezocatalaysis is one of the emerging AOPs which uses energy harvesting materials called piezoelectric materials to convert mechanical energy into electrical energy. Recently, pieozocatalysis has gained much attention in several electrochemical applications including bacterial disinfection (Kumar et al., 2019a), hydrogen production (Hong et al., 2010), wastewater treatment and degradation of water pollutants (Mengying et al., 2017). In bacterial disinfection and degradation of pollutants, the piezoelectric materials (piezocatalyst) generate negative and positive electric charges under the influence of mechanical vibration at opposite surfaces. These free electric charges are responsible for redox reactions resulting in highly reactive species (ROS) such as •O2– and •OH. These strong reactive oxygen species (ROS) are capable of breaking down toxic organic compounds into less toxic compounds, hydroxyl radicals (•OH) have been recognized as secondary oxidants (after the strongest fluorine) due to their high standard reduction potential (Eo •OH/H2O) of roughly 2.8 V versus SHE. Coupling piezocatalysis with photocatalysis can enhance the degradation performance of piezo-photocatalyst and supress the rate of electron and holes recombination in photocatalytic degradation processes. Therefore, this article intends to review enhanced piezo-photocatalytic degradation of organic dyes (section 6.1), pharmaceuticals (section 6.2) and bacteria (section 6.3) using BaTiO3 based catalysts.In the past, BaTiO3 based catalysts have been extensively investigated for their piezocatalytic and piezo-photocatalytic removal ability of several wastewater pollutants including organic dyes. For example, Wu et al. synthesized BaTiO3 nanoparticles and nanowires using a two-step hydrothermal method for piezocatalytic removal of methyl orange (MO) from wastewater (Wu et al., 2018b). Under ultrasonic vibration, BaTiO3 nanowires (NWs) were easily deformed therefore showed better piezocatalytic performance as compared to BaTiO3 nanoparticles (NPs) with poor deformability. The highest piezocatalytic efficiency obtained by BaTiO3 NWs under ultrasonic vibration (power 80 W) was about 92 % within 160 min, with reaction processes following pseudo-first order kinetics model (Fig. 20 (a)). Scavenger studies were conducted to investigate the reactive oxygen species that were more effective in breaking down of MO into CO2 and H2O. Various trapping agents such as tert-butyl alcohol (TBA), benzoquinone (BQ) and disodium ethylene diamine tetra-acetate dehydrates (EDTA-2Na) were used to supress hydroxyl radicals (•OH), superoxide (•O2–) and holes (h+), respectively. As shown in Fig. 20(b), upon the addition of TBA, the degradation efficiency reduced dramatically thus confirming that hydroxyl radicals (•OH) were the most effective ROS species for MO break down, followed by superoxide radicals ((•O2–). Another study was conducted by Hong et al., were BaTiO3 as a piezocatalyst generated more hydroxyl radicals (•OH) and superoxide radicals ((•O2 –) to break down Acid orange 7 (AO7) dye into CO2 and H2O (Hong et al., 2012). Under piezoelectric effect, the strained BaTiO3 dendrites decomposed about 80 % of the AO7 dye after 90 min. Several factors including influence of pH, catalyst dose and initial concentration which can affect the piezocatalytic process were investigated. In case of catalyst loading, the piezocatalytic efficiency increased with an increase in catalyst dosage until reaching a plateau region with 0.025 g of BaTiO3 catalyst. This was due to more available strained induced charges on the surface of BaTiO3 as its total surface-active sites increased with an increase in the amount of catalyst dosage. Therefore, resulting in enhancement of piezocatalytic degradation efficiency. The pH solution and initial AO7 concentration significantly influenced piezocatalytic processes, the piezocatalytic efficiency decreased with an increase in intial AO7 concentration. It was speculated that as initial concentration increases, more AO7 molecules increases which cover less active surface sites of the catalyst thus leading to a decrease in piezocatalytic efficiency. The highest piezocatalytic efficiency was slightly reduced in alkaline media and enhanced in acidic media. In acidic conditions, the surface of the BaTiO3 dendrites is protonated (positively charged) thus enlarges electrostatic interaction between anionic AO7 dye (negatively charged) and positively charged BaTiO3 surface, and resulting in higher piezocatalytic removal of AO7.To improve the photocatalytic activity of BaTiO3, Li et al. fabricated new hybrid composites (Ag2O-BaTiO3) by combining BaTiO3 ferroelectric with Ag2O semiconductor. Under ultrasonic vibration, an internal electric field was generated by ferroelectric BaTiO3 nanocrystal to reduce the rate of electrons and holes recombination thus enhancing the photocatalytic performance of the hybrid composite (Ag2O-BaTiO3) (Li et al., 2015). Fig. 21 shows the effect of ultrasonic vibration on the photocatalytic degradation of Rh B dye using hybrid Ag2O-BaTiO3 photocatalyst. Four photocatalyst materials such as commercial P25 nanoparticles, Ag2O, BaTiO3, mixture of BaTiO3 and Ag2O were used for photocatalytic degradation comparison study. As depicted in Fig. 21(a), P25, BaTiO3 nanocubes, or Ag2O nanoparticles were not effective in the degradation of Rh B under ultrasonic irradiation only. However, the physical mixture of Ag2O and BaTiO3 as well as the Ag2O-BaTiO3 hybrid composite showed a slight deterioration of Rh B. These results shows that the combination of ferrolectric BaTiO3 nanoctrystal and Ag2O semiconductor can improve piezocatalysis/sonocatalysis performance of the Ag2O-BaTiO3 hybrid piezocatalyst. Fig. 21(b) illustrates the photocatalytic degradation of Rh B with all four samples in the absence of an ultrasonic irradiation. The synthesized BaTiO3 showed no photocatalytic degradation towards Rh B, whereas Ag2O, Ag2O-BaTiO3 and their physical mixtures showed higher photocatalytic degradation performance towards the removal of Rh B. Under both UV light and ultrasonic irradiation, Ag2O-BaTiO3 hybrid photocatalyst completely degraded all Rh B within a short space of time (1.5 h) (Fig. 21 (c-d)). Their piezo-photocatalytic or sono-photocatalytic mechanisms were explained as follows: under UV light irradiation, the Ag2O surface generates electrons (e-) and holes (h+), these charge carriers (e-, h+) are required to be separated in order to produce reactive oxygen species (ROS) such as hydroxyl and superoxide radicals for deterioration of Rh B. Under ultrasonic vibration, the ferroelectric BaTiO3 nanocrystal in the Ag2O-BaTiO3 hybrid composite generated an internal piezo-electric field which acted as a driving force for the separation of charge carrier’s (electrons and holes) (Fig. 21 (e-f)). The suppression of rapid electrons and holes recombination separation led to an improved photocatalytic activity of the hybrid composite (Ag2O-BaTiO3).In addition, BaTiO3 photocatalytic performance was further improved by modifying it with several co-catalysts such as ZnO, SnO2, Fe2O3, TiO2 and Bi2O3. Heterostructured BaTiO3/ZnO composites were prepared using the sol–gel and hydrothermal methods for the degradation of MB, MO, and RhB (Kappadan et al., 2020b; Karunakaran et al., 2014; L. Wang et al., 2019). In the study conducted by Kappadan et al., hydrothermal method was employed to prepare n-n heterojunction photocatalyst of BaTiO3/ZnO using Barium acetate, titanium tetra isopropoxide and Zinc nitrate hexahydrate as precursors (Kappadan et al., 2020b). The scanning electron microscopic images of the composite (BaTiO3/ZnO) clearly showed spherical BaTiO3 nanoparticles anchored on hexagonal rod-shaped ZnO. The formation of n-n heterojunction resulted in improved photocatalytic performance, as evidenced by the reduction of band gap energy from 3.1 to 2.97 eV and charge transfer resistance from 871 to 745 Ω. The mechanisms, as shown in Fig. 22 (a), revealed that an improved charge carriers separation was achieved by forming a typical type II band alignment within the BaTiO3/ZnO heterojunction interface, which allowed BaTiO3 photogenerated electrons to migrate into the conduction band (CB) of ZnO, whereas ZnO photogenerated holes were transported into the valence band (VB) of BaTiO3. The prepared heterostructured BaTiO3/ZnO (BTZ) photocatalyst showed good stability since even after the 3rd cycle the degradation efficiency was above 91 % for methylene blue (MB) (Fig. 22(b)). However, after the 4th cycle the photocatalytic degradation efficiency slightly reduced from 91 to 86 %. The total mineralisation of MB dye was calculated using TOC, under UV light irradiation, the BTZ heterostructure recorded a TOC removal of 87.1 % after 60 min. A recent report by Liu et al. modified BaTiO3 with TiO2 via hydrothermal process to form BaTiO3-TiO2 core–shell heterostructures (Liu et al., 2019). The heterostructured photocatalyst was applied for photocatalytic deterioration of Rh B dye from wastewater. The ratio of BaTiO3:TiO2 played a significant role in the photocatalytic degradation of Rh B. All BaTiO3-TiO2 core–shell heterostructures with different molar ratios exhibited better photocatalytic removal towards Rh B as compared to pure BaTiO3. The BaTiO3-TiO2 core–shell heterostructures (1.2:1) showed the greatest photocatalytic performance compared to other samples, its performance was 1.8 times greater than pure TiO2. The improved photodegradation activity was due to the fact that BaTiO3-TiO2 core–shell heterostructures (1.2:1) exhibited the lowest photoluminiscent (PL) intensity thus lowering the rate of electrons and holes recombination. Wu et al. boosted the photocatalytic activity of heterostructured BaTiO3/TiO2 nanocomposites with piezotronic effect under ultrasonic vibration (J. Wu et al., 2020). Under ultrasonic activation, the built-in electric field exhibited by ferroelectric BaTiO3 facilitated the charge transfer and separation within BaTiO3/TiO2, thus improving its photocatalytic activity. Comparing with other BaTiO3 based metal oxides (Alex et al., 2019; Cui et al., 2017; Fan et al., 2012; Karunakaran et al., 2014; Lin et al., 2007; Liu et al., 2019; Selvarajan et al., 2017; Zhou et al., 2019), BaTiO3/TiO2 composite outperformed them in terms of photocatalytic performance due to excellent suppression of electrons and holes recombination.Plasmonic photocatalysts have also attracted a lot of attention in the photocatalytic removal of organic pollutants in wastewater. Several noble metals including platinum (Pt), gold (Au), rhodium (Rh) and silver (Ag) have been doped with BaTiO3 to promote chemical redox reaction under UV light and ultrasonic irradiation. For example, chao et al. fabricated a heterostructured Au@BaTiO3 photocatalyst using the hydrothermal method for the breakdown of Rh B under UV light exposure (Chao et al., 2020). The plasmonic heterostructured photocatalyst showed an enhanced photocatalytic activity towards the removal of Rh B. The heterostructured composite (Au@BaTiO3) nearly degraded about 100 % of Rh B after 36 min and the reaction process followed Langmuir-Hinshelwood model. The degradation rate constant (k) obtained from Langmuir-Hinshelwood model were 0.05446 and 0.01118 min−1 for Au@BaTiO3 and pure BaTiO3, respectively. The apparent rate constant (k) for Au@BaTiO3 was 4.9 times than that of pristine BaTiO3. These results confirmed an enhancement in the photocatalytic properties of heterostructured Au@BaTiO3 photocatalyst. Several studies investigated the photocatalytic performance of silver (Ag) doped BaTiO3 photocatalyst for the catalytic degradation of hazardous organic dyes (Rh B and MO) (Cui et al., 2013; Lin et al., 2021; Nithya and Devi, 2019; Xu et al., 2019). For instance, Khan et al. reported Ag-doped BaTiO3 prepared via sol–gel method for photodegradation of Rh B (Khan et al., 2021b). The XRD and HR-TEM results confirmed their tetragonal phase and crystallite size range of 46–54 nm, respectively. The incorporation of silver (Ag) dopants into BaTiO3 reduced its band gap from 3.87 to 3.47 eV. Furthermore, they observed a reduction in photoluminescent peak intensity upon addition of silver ions into tetragonal BaTiO3, confirming restrain in electrons and holes recombination. The BET surface of Ag (5 %)-doped BaTiO3 had a higher surface area of 20.1 m2.g−1 as compared to pure BaTiO3 (15.4 m2.g−1). Therefore, due to Ag-doped BaTiO3 having a higher surface area and lower band gap, their photocatalytic performance was improved. Under light irradiation (400 W sodium lamp), 1 %Ag@ BaTiO3, 3 %Ag@BaTiO3 and 5 %Ag@BaTiO3 showed better photocatalytic performance than pure BaTiO3. The improved photocatalytic activity of Ag-doped BaTiO3 could be due to the reduced rate of electrons and holes recombination, surface area increase and reduction in band gap energy after Ag-deposition. The highest photocatalytic removal for Rh B was 79, 58, 46 and 51 % when 5 %Ag@BaTiO3, 3 %Ag@ BaTiO3, 1 %Ag@ BaTiO3 and BaTiO3 were applied as photocatalysts, respectively. Niu and Xu observed the same photocatalytic enhancement when Ag-BaTiO3 was further co-doped with other metal dopants such as Ni, Pd and Pd–Sn–Ni through a one-step ball milling process (Niu and Xu, 2019). The co-doped Ag@BaTiO3 rate constant was 4.5 times greater than undoped BaTiO3. Table 4 summarizes an overview application of BaTiO3-based piezo/photocatalyst for catalytic degradation of organic dyes in water and wastewater.In recent decades, antibiotics have been identified as emerging contaminants owing to their endurance in aquatic ecosystems. They are widely used for treating numerous bacterial infections. However, they find their way into several waterbodies through incorrect disposal of unused or expired medications and human excretion. In ground and surface water, antibiotics are found in lower concentrations ranging from μg/L to mg/L (Cabeza et al., 2012; Orimolade et al., 2021b; Wang and Wang, 2016). Unfortunately, the detection of these pharmaceuticals in aquatic environment can lead to several harmful biological and economic impacts. For example, continuous ingestion of drinking water contaminated with pharmaceuticals can result in the formation of drug-resistant bacterium strains to humans and animals. The elimination of several emerging pharmaceuticals by BaTiO3 based ferroelectric/piezo-photocatalyst system have attracted so much attention recently. Kurniawan et al. investigated the photocatalytic performance of BaTiO3 and BaTiO3/TiO2 composites for the removal of acetaminophen (Ace) from distilled water under UV–vis irradiation (Kurniawan et al., 2018). The XRD patterns confirmed the cubic phase structure of BaTiO3 nanoparticles. The BaTiO3/TiO2 composites as compared to individual pristine (BaTiO3 and TiO2) performed better, thus confirming that the synergetic effect of the two semiconductors can improve visible light absorption and efficient charge separation. Furthermore, it was found that the photocatalytic performance of the composites (BaTiO3/TiO2) can be enhanced by varying the molar weight ratios (w/w) of BaTiO3 and TiO2. Under optimal conditions, dosage of 1 g/L, pH 7, 4 hrs reaction time and initial Ace concentration of 5 mg/L, BaTiO3, TiO2 and BaTiO3/TiO2 photocatalytically degraded about 18, 33 and 95 % of Ace respectively. As shown in Fig. 23 , initial degradation pathways for Ace were through hydroxylation and photolysis. The intermediates formed during this route included hydroquinone and 1,4-benzoquinone, which were consistent with the results obtained by Zhang et al. and Aguilar et al. (Zhang et al., 2008) (Aguilar et al., 2011). The hydroxyl radicals (•OH) further attacked these intermediates to form hydroxylation products. The first detected intermediate was quinoneimine which was effortlessly hydrolized to 1,4-benzoquinone. The intermediates were then oxidized further into carboxylate acid and carbon dioxide by destroying their aromatic structures. The total mineralization via photocatalysis process generally does not occur quickly, however after some few hours these organics can be totally mineralized. Demircivi and Simsek reported tungsten-doped BaTiO3 (W-BaTiO3) for enchanted photocatalytic removal of tetracycline under and visible light-driven and UV-A light irradiation (Demircivi and Simsek, 2019). The composite was prepared through a simple hydrothermal method using a Teflon-lined stainless-steel autoclave at 200 °C, followed by washing, drying and calcination for 2 hrs at 700 °C. The effect of tungsten (W) loading on BaTiO3 was investigated for tetracycline degradation. It was found that BaTiO3 doped with low amounts of tungsten (W) exhibited higher photocatalytic activity than pure BaTiO3 and BaTiO3 doped with higher amount of tungsten (W). According to this study, pH played a significant role in the photocatalytic degradation of tetracycline. The photocatalytic degradation removal percentage increased with an increase in pH solution. The degradation efficiency was recorded to be 3, 80 and 90 % at pH 3, 5.60 and 10, respectively. In addition, tungsten-doped BaTiO3 (W-BaTiO3) was used to treat spiked real water samples.(tap and drinking water). Under light irradiation (after 3 hrs), the composite achieved a degradation efficiency of 74 and 76 % in drinking and tap water, respectively. The decrease in the degradation efficiency from tap and drinking water was due to other various pollutants or ions detected in the water matrices, which has some negative effect on the photocatalytic process (Bilgin Simsek, 2017). In a study conducted by Demircivi et al., decorated BaTiO3 with carbon fibers (CFs) were synthesized for enhanced photocatalytic degradation of tetracycline (Demircivi et al., 2020). The incorporation of CFs to BaTiO3 reduced the band gap of the composite and showed enhancement in photocatalytic activity of BaTiO3/CF. Under UV and visible light irradiation, BaTiO3/CF showed the highest degradation efficiency of 96 % as compared to BaTiO3 (W-BaTiO3). According to re-usability studies, the composite showed the highest stability within 5 cycles. However, after 6 cycles, a sharp decline in degradation efficiency was observed from 96 to 77 %. This could be due to the loss of tiny photocatalyst powder with an increase in recyclability. Almost a 100 % degradation efficiency was reported when Ti32-oxo-cluster/BaTiO3/CuS p-n heterojunction was employed for wastewater treatment under both visible light irradiation and mechanical vibration (Piezo-photocatalysis) (Zhou et al., 2021).The piezoelectric/ferroelectric materials including ZnSnO3, MoS2, NaNbO3, BiFeO3, KNbO3, BiOCl, Bi4 Ti3O12 and BaTiO3 nanoparticles in powder forms have been utilized as promising piezocatalysts for several applications. However their applications are limited in cleaning water due to the inability of being recovered from aqueous solution (Lin et al., 2020). Sharma et al. investigated the piezocatalytic activity of cement-based BaTiO3 composites for the removal of several pollutants in water such as pharmaceutical (paracetamol) and dyes ((Rh B), (MO) and (MB)) (Sharma et al., 2020). Sharma and co-workers, combined powdered BaTiO3 nanoparticles with cement to form cement-ferroelectric composites which can be easily recovered from aqueous solution after wastewater treatment. Under ultrasonic vibration, the poled BaTiO3 cement composites showed significant piezocatalytic removal of all organic pollutants. The poled composite exhibited the highest piezocatalytic degradation of approximately > 90, 86, 85, and 79 % for Rh B, MB, MO and paracetamol, respectively. The composites could be reused up to the 5th cycle of piezocatalysis under ultrasonic vibration. Another antibiotic drug which is in the class of fluoroquinolone antibiotics is Norfloxacin (NFX). This antibiotic drug is not easily degradable and can further contribute to antibiotic resistance when used. Therefore, the double Z-scheme of BiFeO3/CuBi2O4/BaTiO3 was fabricated by Zhang et al., and employed for photocatalytic degradation of Norfloxacin (NFX) under solar-light irradiation (Zhang et al., 2020). The combination of BiFeO3, CuBi2O4 and BaTiO3 to form the composites, extended the light absorption to UV–visible and near-infrared (NIR) light to allow efficient use of the whole solar light spectrum. As shown in Fig. 24 (a), the nanocomposites Z-scheme of BiFeO3/CuBi2O4/BaTiO3 strongly absorbed in the 200–800 nm range. The calculated optical band energy gap from the tauc’s plot were approximately 3.30, 1.76, 2.29 and 2.20 eV for BaTiO3, CuBi2O4, BiFeO3 and BiFeO3/CuBi2O4/BaTiO3, respectively (Fig. 24(b)). The effect of irradiation time, catalyst amount and initial NFX concentration were investigated on the photocatalytic performance of the Z-scheme composites. It was noticed that as irradiation time and catalyst dosage increased, the degradation efficiency also increased (Fig. 24 (c-d)). The degradation rate increased with an increase in initial NFX concentration within a particular concentration range (Fig. 24(e)), this was due to the fact that more NFX molecules remained in aqueous solution at higher concentrations. Under the same optimal parameters (dosage of 1.0 g/L, within 60 min, and NFX concentration of 2.5 mg/L), BiFeO3/CuBi2O4/BaTiO3 nanocomposites had a better photocatalytic activity than individual samples of BaTiO3, CuBi2O4, BiFeO3. The composite reached its highest degradation of 93.5 % which could be attributed to a better separation of electrons (e-) and holes (h+) to improve the redox ability thus enhancing the photocatalytic activity of the catalyst. The electrons and holes separation were confirmed by photoluminescence (PL) spectrum (Fig. 24(f)). Generally, a low PL intensity reveals greater electrons (e-) and holes (h+) separation efficiency. As shown in Fig. 24(f), BiFeO3/CuBi2O4/BaTiO3 had the lowest PL intensity than BaTiO3, CuBi2O4 and BiFeO3, thus indicating better enhancement in photocatalytic activity of the composite. The extent of mineralization of NFX (10.0 mg/L) obtained from TOC was 3.9, 6.8, 40.9 and 60.3 % for BaTiO3, CuBi2O4, BiFeO3 and BiFeO3/CuBi2O4/BaTiO3, respectively. However, the extent of mineralization for NFX (2.5 mg/L) reached up to 93.5 % within 1 hr. According to scavenger experiments conducted, it was found that hydroxyl radicals (•OH) and holes (h+) played an important role in the deterioration of NFX than superoxide radicals (•O2 –).Pharmaceuticals used in aquaculture are also a source of pollution that is delivered directly into surface water. For example, atrazine is a well-known herbicide that is used to control broadleaf and grassy weeds in water. It was banned in most countries because of its negative impact on aquaculture and humans (Cavas, 2011). Due to that, Basaleh and Mohamed (Basaleh and Mohamed, n.d.) developed Copper (Cu)-doped BaTiO3 photocatalyst for the removal of this toxic herbicide (atrazine) from wastewater so as to provide clean water to the environment. The photocatalyst (Cu-BaTiO3) was prepared through the hydrothermal and photo-assisted deposition method. The prepared photocatalyst was added into 300 ml atrazine solution (50 ppm) and irradiated with Xenon lamp. The photocatalytic removal for atrazine using 0.5 Cu/BaTiO3, 1.0 Cu/BaTiO3, 3.0 Cu/BaTiO3 and 5.0 Cu/BaTiO3 were recorded to be 45, 65, 100 and 100 %, respectively. As for pure BaTiO3, the removal efficiency was low as 3 % due to the high electron and holes recombination. These results confirmed that doping BaTiO3 with Cu can suppress the rate of electrons and holes recombination, thus increasing photocatalytic performance of Cu-BaTiO3. The suppression of electrons and holes recombination was confirmed by photocurrent response and photoluminescence (PL) spectrum. The Cu/BaTiO3 had a greater photocurrent response of 12.8 mA cm−2 than pure BaTiO3 (2.8 mA cm−2). In another study, BaTiO3 was co-loaded with two electrocatalyst (Pt and RuO2) to promote sufficient redox ability for piezocatalytic degradation of tricyclazole under mechanical vibration (Feng et al., 2019). The platinum (Pt) was selected since it has been considered as the best catalyst for oxygen reduction, whereas RuO2 can produce large amounts of hydroxyl radicals and facilitate protons transport during electrocatalytic reaction. The loading of co-catalysts to BaTiO3 resulted in surface area increment from 25.5 to 28.8 m2g−1. Specific surface area of the materials is another factor which can influence the photocatalytic performance of the photocatalyst. Under ultrasonic vibration (40 kHz, 110 W), the composite achieved the highest removal percentage of 86 % which was higher than the values obtained using RuO2/t-BaTiO3 (51.0 %) and Pt/t-BaTiO3 (75.9 %). According to apparent rate constant values (k) obtained from pseudo-first order kinetics, the piezocatalytic reaction rate for the composites (k = 0.0320 min−1) was 3.11 times greater than pure BaTiO3 (k = 0.0103 min−1) and the sum of k values of Pt/t-BaTiO3 (0.0125 min−1) and RuO2/t-BaTiO3 (0.0124 min−1). The outstanding performance of the composite confirmed that coupling ferroelectric materials with a good catalyst can yield a synergistic enhancement effect on the piezocatalytic degradation. Another BaTiO3 based heterostructured catalyst which has been employed for piezocatalytic/photocatalytic and piezo-photocatalytic removal of pharmaceutical pollutants is BaTiO3/La2Ti2O7 heterojunction. Li et al. prepared BaTiO3/La2Ti2O7 composites via a two-step hydrothermal and microwave hydrothermal synthesis for piezo-photocatalytic degradation of ciprofloxacin (Y. Li et al., 2021). After 90 min of photocatalytic degradation, BaTiO3/La2Ti2O7 recorded a degradation efficiency of 16.5 and 22.7 % greater than that of BaTiO3 and La2Ti2O7, respectively, thus indicating that the formation of a heterojunction improved the photocatalytic degradation process. Under ultrasonic vibration (piezocatalysis), the degrading efficiency of CIP over BaTiO3/La2Ti2O7 (37.7 %) was roughly 19 % greater than that of BaTiO3 and La2Ti2O7. The degradation rate constant (k value) obtained from pseudo first-order kinetics model was 0.00593 min−1 for BaTiO3/La2Ti2O7, which is roughly 2.8 and 2.3 times greater than BaTiO3 and La2Ti2O7, respectively. The higher degradation efficiency of 50.2 % was achieved when piezocatalysis and photocatalysis were merged. For all samples, the performance of piezo-photocatalysis was superior to that of photocatalysis or piezocatalysis. This might be due to photocatalysis's low visible-light absorption and low carrier separation efficiency. The catalytic performance was likewise low during the piezocatalytic procedure due to the restricted amount of free charge carriers created. Photogenerated charge carriers were efficiently separated under the combined action of visible light and ultrasound, and CIP degradation efficiency was greatly increased when compared to sole-ultrasound and sole-visible-light irradiation. The composites showed higher catalytic activity than individual samples, thus indicating that the BaTiO3/La2Ti2O7 heterojunctions boost the catalytic process. The photocatalytic degradation of tetracycline and Rh B was reported by Zheng et al., using a Z-type BaTiO3/γ-Bi2O3 heterojunction which was prepared via hydrothermal, co-precipitation, and calcination (Zheng et al., 2022). The morphological structure of the prepared samples appeared to be of a tetrahedron shape, irregular nano-particles and a mixture of both shapes (tetrahedron and nanoparticles) for γ-Bi2O3, BaTiO3 and BaTiO3/γ-Bi2O3 heterojunction, respectively (Fig. 25 (a-c)). It was found that the calcination temperature had no effect on the morphology, the obtained average particles for γ-Bi2O3, BaTiO3 and BaTiO3/γ-Bi2O3 (HS3) were roughly around 5.9 μm, 425.9 nm, and 1.9 μm, respectively. The effect of catalyst dose, pH of the solution and different water bodies on photodegradation of tetracycline was explored, however these parameters had a little impact on the photocatalytic degradation process (Fig. 25 (d-g)). The degradation efficiency for both pristine (γ-Bi2O3 and BaTiO3) were below 67 %, for γ-Bi2O3 and BaTiO3 were found to be 59.65 and 66.28 %, respectively. However, for all Z-type BaTiO3/γ-Bi2O3 (HS) heterojunctions with different molar ratio, reaction time and calcination temperature, the degradation removal percentages were above 93 %, with HS3 exhibiting the highest degradation efficiency of 97.95 % for tetracycline. Even for Rh B dye, the degradation efficiency for single γ-Bi2O3 and BaTiO3 were below 67 %. It was found that γ-Bi2O3 and BaTiO3 degraded about 63.34 and 45.35 %, respectively. The photocatalytic degradation efficiencies for all HS heterojunction samples were above 73 %, with HS3, HS4, and HS5 breaking down the Rh B molecule entirely. The enhanced photocatalytic activity was due to electrons transferred via Z-type from Bi2O3 conductor band (CB) to BaTiO3 valence band (VB) by work function and charge density difference which resulted in charge separation.Another type of hydrothermally synthesized Z scheme heterojunction of La(OH)3@BaTiO3 (LB) composite was investigated for deterioration of an A-ring of tetracycline (Zheng et al., 2022). The reactive oxygen species (ROS) such as hydroxyl radicals (OH), superoxide (O2 –), holes and electrons completely degraded 100 % of tetracycline. The scavenger studies confirmed that four active species played a role in their photocatalytic degradation, as follows: h+ = •O2 − >•OH > e- (before 20 min reaction) and h+ > •OH >•O2 – > e- (after 20 min reaction). According to these results, it means that holes (h+) played a major role in the photocatalytic degradation whereas electrons (e-) showed the least contribution during the photocatalytic degradation of tetracycline. The application of BaTiO3-based catalyst for removal of pharmaceuticals is summarized in Table 5 .Despite using BaTiO3-based catalyst for piezo-photocatalytic degradation of organic dyes and pharmaceuticals, there are some factors which affects the piezo-photocatalytic degradation process negatively. Some of these factors include nature of the semiconductor, amount of the catalyst, solution pH, reaction time, light intensity, dissolved reactive oxygen species and temperature. For examples, some organic pollutants showed maximum adsorption removal and piezo-photodegradation at lower (acidic media) or higher pH (basic media) due their complex structure. Therefore, limiting their applications in real wastewater treatment because it means prior to degradation processes, the pH of the real wastewater samples needs to be adjusted. Another critical factor is the surface area of the piezo-photocatalyst, since the piezo-photodegradation efficiency increases with an increase in surface area of the catalyst. It is very important to select piezo-photocatalyst with very high surface area because of more active sites, which assist in the enhancement of piezo-photodegradation. Furthermore, this processes requires reactive high amount of reactive oxygen species (ROS) such as hydroxyl radicals and superoxide to completely oxidize organic dyes and pharmaceutical into less harmful by-products such as carbon dioxide (CO2) and water (H2O). Therefore, lower levels of reactive oxygen species would result in incomplete oxidation of organic pollutants.Piezocatalytic, photocatalytic and piezo-photocatalytic disinfection has attracted more attention in elimination of pathogenic bacteria. These processes usually use reactive oxygen species (ROS) such as hydroxyl radicals, superoxide, hydrogen peroxide and h+, e- generated when piezo-photocatalyst is exposed under light irradiation or mechanical vibration to kill bacteria (J. He et al., 2021). Generally, the idea of photocatalytic disinfection is to first remove each bacterium's cell wall, removing its protection, and then to damage its cytoplasmic membrane, causing the cellular material inside the newly torn cell envelope to degrade (Fig. 26 ). Ferroelectric based catalysts such as barium titanate (BaTiO3) have been used as piezocatalyst/photocatalysts for bacterial disinfection due to its exceptional optical and piezoelectric/ferroelectric properties. For example, Zhao et al. fabricated a novel p-n type Cu2MgSnS4/BaTiO3 (CMTS@BaTiO3) heterojunction for the degradation of organic dyes and bacterial disinfection (Ali et al., 2021). In terms of bacterial disinfection, CMTS@BaTiO3 obtained 72–76 % and 84–90 % inhabitation for E.coli and S.aureus, respectively, which was three times greater than pure CMTS and BaTiO3. Furthermore, CMTS@BaTiO3 heterojunction was tested for inactivation of E.coli and S.aureus in the presence of wastewater containing dyes (MG and MB). When compared to as-grown bacteria, inactivated levels of bacterial percentages employing CMTS@BaTiO3 composites were around 94.22–101.24 % with MB and MG, respectively, against E.coli and 97.59 to 87.96 % for S.aureus. In comparison to CMTS@BaTiO3, Kumar et al. demonstrated piezocatalytic, photocatalytic and piezo-photocatalytic disinfection of E.coli using unpoled and poled BaTiO3 ceramic under UV light and ultrasonic vibration (Kumar et al., 2019a). During the piezocatalytic process (under ultrasonic vibration), unpoled BaTiO3 showed a 56 % of bacterial disinfection within 30 min. Under light exposure and ultrasonic vibration, piezo-photocatalysis improved the bacterial degradation rate and the piezo-photocatalytic degradation of E.coli increased from 56 to 70 % within the same reaction time. When the poled BaTiO3 ceramic were employed for piezocatalytic degradation (under ultrasonic vibration), it was found that about 97 % of bacteria were killed. Under the piezo-photocatalysis process (ultrasonic vibration and light exposure), poled BaTiO3 ceramic showed some catalytic enhancement towards catalytic degradation of bacteria (99.99 % recorded within 20 min). From these results, it can be concluded that poled BaTiO3 ceramic have better photocatalytic activity than unpoled BaTiO3 ceramic and coupling piezocatalysis with photocatalysis further enhanced the catalytic degradation activity of BaTiO3.Kumar et al. investigated the photocatalytic and antibacterial activity of poled BaTiO3 prepared via a solid state method (Kumar et al., 2019b). From the FESEM analysis, it was found that BaTiO3 consisted of large dense grains with clear grain boundaries (Fig. 27 (a)). The calculated average grain size of BaTiO3 sample ranged from 40 to 60 μm (Fig. 27(b)). Both unpoled and poled BaTiO3 were shown to respond under UV irradiation. The poled BaTiO3 exhibited the highest photocurrent response of about 0.006 μA, which was>100 times greater than that of unpoled BaTiO3. Thus, indicating that poled BaTiO3 had a higher photocatalytic activity (Fig. 27 (c-d)). This was due to its remnant polarization which inhibited recombination of photogenerated charges. However, under dark conditions both samples (poled and unpoled BaTiO3) showed no response to UV irradiation.The colony forming unit (CFU) method was employed to study the effect of poled BaTiO3 samples on absolute bacterial mortality. As shown in Fig. 27(e), the results display that there was no substantial antibacterial activity with unpoled BaTiO3 and negative side of poled samples in the dark. However, the positive side of poled BaTiO3 exhibited strong antibacterial activity with about 90 % bacterial destruction after 60 min when UV light was irradiated (Fig. 27(f)). In the absence of a catalyst (BaTiO3 poled or unpoled), the UV light alone showed antibacterial activity since it can prevent bacterial growth by destroying their structural DNA. Kushwana and co-workers reported (Kushwaha et al., 2015) Li-Doped Bi0.5Na0.45 K0.5TiO3–BaTiO3 (BNKLBT) for antibacterial activity against E.coli and A.flavus using standard disc diffusion method. This method involves bacterial growth using Luria–Bertani (LB) broth–agar in a petridish disc. Three different concentrations (10, 50 and 100 μg) of BNKLBT were tested against bacteria and compared with commercial antibiotics (kanamycin (k30)). Based on their results, the zone of inhibition increased as BNKLBT concentration increased on the disc. The reason for this trend was not explained in this work. In another report, BaTiO3 nanoparticles were tested for antibacterial efficacy against S.aureus and P. aeruginosa by Shah et al.(Shah et al., 2018). The diffusion method was employed to investigate the antimicrobial activity of BaTiO3 nanoparticles. The optimal concentration of 100 μg/ml (BaTiO3) achieved bacterial inhibition of 85 ± 3.5 % and 80 ± 3 % against S.aureus and P.aeruginosa biofilms, respectively.Shuai et al. prepared PVDF/xAg-pBT composites via laser sintering method for bacterial disinfection (Shuai et al., 2020a). According to their studis, PVDF/xAg-pBT composites generated piezoelectric potential/voltage. Furthmore, it was found that with an increase in Ag concentration in the PVDF/pBT composites, the piezoelectrical current and voltage firstly rose then dropped. Babu et al. and Parl et al. observed comparable enhancement after adding conductive fillers to polymer-ceramic composites (Babu and de With, 2014) (Park et al., 2012). Their assumptions on the output voltage and current improvement were based on the conductivity enhancement. The antibacterial activity of the prepared scaffolds (PVDF/4Ag-pBT and PVDF/pBT) were evaluated using zone of inhibition method. Fig. 28 (a-b) shows the bacterial inhibition zones of the scaffolds. PVDF/4Ag-pBT was shown to be more successful in inhibiting the growth of E.coli as compared to PVDF/pBT scaffold. Further test including Turbidimetric test were conducted to verify antibacterial activity of the scaffolds. The transparent vials including and excluding E.coli suspensions were labelled as control and blank group, respectively. The turbidity was the same for vials containing PVDF/4Ag-pBT and control. Surprisingly, the vial containing E.coli solution incubated with PVDF/4Ag-pBT was clear as a blank group, thus indicating that the scaffold inhibited bacterial growth of E.coli (Fig. 28(c)). The SEM images also confirmed that PVDF/4Ag-pBT scaffold raptured and destroyed the whole rod-shape structure of E.coli, whereas PVDF/pBT scaffold had minimal impact on the smooth rod-shape of E.coli (Fig. 28 (d-e)). Fig. 28 (f-g) displays bacterial inhibition rate of scaffolds and cumulative or non-cumulative of silver ion concentration released by PVDF/4Ag-pBT. The PVDF/4Ag-pBT scaffold inhibited bacteria at a rate of over 81 %, while the PVDF/pBT scaffold had no antibacterial activity.Overall, BaTiO3-based catalyst have been shown to be appropriate for applications involving the combination of photocatalysis, piezocatalysis and other catalysis processes. An overview of the previous studies addressing the use of BaTiO3-based catalyst for water disinfection is summarized in Table 6 . The use of piezo-photocatalyst based materials for bacterial disinfection is still in its infancy. The process of piezo-photocatalysis has been widely applied including hydrogen production (H2), degradation of organic pollutants in wastewater, carbon dioxide (CO2) conversion and hydrogen peroxide (H2O2) production. In piezo-photocatalytic disinfection, it is challenging to develop a general mechanism for bacteria inactivation because of a wide range of pathogens and their cell complexity. Therefore, deeper understanding of the mechanisms the piezo-photocatalytic process is still required since suggested mechanisms highly rely on the generated reactive oxygen species during redox reactions. It is still debatable whether polarized charge carriers participate in the redox catalytic process. According to certain studies, photogenerated electrons and holes play a crucial part in redox processes, and the polarization potential of piezoelectric materials merely helps to separate photogenerated charge carriers (Fu et al., 2022).Barium titanate (BaTiO3) offers a wide range of applications in the energy and environmental fields. In comparison to certain applications like water hydrogen production, its usage in the piezo-photocalaytic wastewater treatment and bacterial disinfection is limited and recent. This review article has presented a complete summary of current developments in the use of barium titanate-based catalysts in photocatalytic, piezocatalytic and piezo-photocatalytic decomposition of organics and bacterial disinfection in water and wastewater. The selection of BaTiO3 as a suitable piezo-photocatalyst was due to its outstanding dielectric/ferroelectric/piezoelectric characteristics, low toxicity, low cost, environmental friendliness, existence in broad range of sizes and morphologies, multiple crystal structures and good stability. These characteristics are undoubtedly being used in the piezo-photocatalytic elimination of a variety of organic contaminants. Since, BaTiO3 as a photocatalyst suffers from poor conductivity and rapid recombination of photogenerated charge carriers. In this review, we have covered several strategies to circumvent these restrictions including morphology control, doping, metal loading, and heterojunction construction with appropriate semiconductors. Another way of improving electrons and holes separation is through coupling photocatalysis with other catalysis processes such as piezocatalysis and electrocatalysis. These strategies have achieved excellent results while conserving energy. The effectiveness of BaTiO3-based catalyst for piezo-photocatalytic degradation of organic pollutants depends mainly on several factors such as pH, reaction time, amount of catalyst, ultrasonic and light power. The fundamental benefit of heterogeneous photocatalysis is its capacity to use solar energy in the form of solar photons, which gives the degradation process a large boost in environmental value. Particularly for large-scale aqueous-phase applications, solar light-assisted photodegradation of wastewater contaminants can make it an economically viable method.Even though several reports have shown some strategic ways to improve photocatalytic activity of barium titanate, however some few other limitations and challenges needs to be highlighted for its success in wastewater treatment. 1. There is limited research on the structure of the barium titanate after being exposed to ultrasonic vibration and light irradiation. It is very important to investigate the stability of the barium titanate structure using TEM, SEM and XRD before and after wastewater treatment. 2. Even though doping barium titanate with plasmonic metals have shown significant improvement in photocatalytic activity, the practical applicability of plasmonic BaTiO3 piezo-photocatalyst is limited due to the high cost and photo-corrosion associated with noble metal nanoparticles. 3. Since the size and morphology of the doped noble metals can affect the photocatalytic activity of the BaTiO3, there are no reports in the literature on the size and morphology of the noble metals nanoparticles doped on the surface of BaTiO3 piezo-photocatalyst. To establish the ideal condition of noble metal nanoparticles needed for photocatalytic enhancement, it is therefore necessary to assess the surfaceplasmonresonance (SPR) impact from different sizes and shapes. 4. From our analysis, there’s limited understanding of piezo-photocatalytic mechanism for BaTiO3-heterojunction piezo-photocatalyst. Therefore, for future studies it is very important to study reaction mechanisms for piezo-photocatalytic degradation of organic pollutants in detail. 5. Furthermore, many reports in literature have given limited information regarding how synthetic methods and structural properties of BaTiO3 piezo-photocatalyst affects the efficiency of the process. As a result, future research should pay close attention to how these factors impact piezo-photocatalyst effectiveness. 6. The amount of organic pollutant degradation is determined by the photocatalysts' mineralization capacity. However, the majority of earlier studies on BaTiO3-based materials did not identify the total organic carbon (TOC) in the mineralization process, which should be taken into account for an accurate assessment of the whole degradation of organic contaminants. There is limited research on the structure of the barium titanate after being exposed to ultrasonic vibration and light irradiation. It is very important to investigate the stability of the barium titanate structure using TEM, SEM and XRD before and after wastewater treatment.Even though doping barium titanate with plasmonic metals have shown significant improvement in photocatalytic activity, the practical applicability of plasmonic BaTiO3 piezo-photocatalyst is limited due to the high cost and photo-corrosion associated with noble metal nanoparticles.Since the size and morphology of the doped noble metals can affect the photocatalytic activity of the BaTiO3, there are no reports in the literature on the size and morphology of the noble metals nanoparticles doped on the surface of BaTiO3 piezo-photocatalyst. To establish the ideal condition of noble metal nanoparticles needed for photocatalytic enhancement, it is therefore necessary to assess the surfaceplasmonresonance (SPR) impact from different sizes and shapes.From our analysis, there’s limited understanding of piezo-photocatalytic mechanism for BaTiO3-heterojunction piezo-photocatalyst. Therefore, for future studies it is very important to study reaction mechanisms for piezo-photocatalytic degradation of organic pollutants in detail.Furthermore, many reports in literature have given limited information regarding how synthetic methods and structural properties of BaTiO3 piezo-photocatalyst affects the efficiency of the process. As a result, future research should pay close attention to how these factors impact piezo-photocatalyst effectiveness.The amount of organic pollutant degradation is determined by the photocatalysts' mineralization capacity. However, the majority of earlier studies on BaTiO3-based materials did not identify the total organic carbon (TOC) in the mineralization process, which should be taken into account for an accurate assessment of the whole degradation of organic contaminants.Overall, the use of BaTiO3-based piezo-photocatalysts for real-world remediation of pharmaceutical, organic dyes and microbes from wastewater has a bright future. Therefore, it is highly recommended that advanced oxidation processes such as photocatalysis and piezocatalysis should be used for wastewater treatment in the future instead of traditional techniques.This work was supported by GES 4.0 (Global Excellence Structure 4.0), the Centre for Nanomaterials Science Research (CNSR) University of Johannesburg), National Research Foundation of South Africa (SRUG200326510622), National Research Foundation of South Africa (NRFTTK117999) and the Faculty of Science University of Johannesburg (UJ), South Africa for financial support.

The coupling of piezocatalysis and photocatalysis known as piezo-photocatalysis has attracted a lot of attention as one of the most effective advanced oxidation process (AOPs) for wastewater treatment, especially for the degradation of organic pollutants and disinfection of microbes. To advance this technology, there’s a need to develop lead free piezoelectric materials to drive both piezocatalytic and photocatalytic process to prevent secondary pollution due to lead toxicity. Hence, barium titanate (BaTiO3) has been widely used as lead free piezoelectric material for several applications including water splitting, bacterial disinfection, and wastewater treatment due to its exceptional optical and piezoelectric properties. This work presents a comprehensive review on the application of BaTiO3 as a promising lead-free piezo-photocatalyst for the catalytic degradation of organic pollutants and bacterial disinfection from aqueous solution. This review article details the optical and piezoelectric properties, modification strategies, and synthetic methods of BaTiO3. Furthermore, the application of BaTiO3 as a preferred piezo-photocatalyst for wastewater treatment and a future perspective is presented.

Fuel cells have attracted a great deal of attention as clean energy conversion devices because of their high efficiency and low emissions [1,2]. Proton exchange membrane fuel cells (PEMFCs) have already reached the level of practical use but remain costly, because they require catalysts that include acid-resistant platinum group metals (PGM). Currently, anion exchange membrane fuel cells (AEMFCs) have been investigated as low-cost alternatives to PEMFCs due to their potential use of non-PGM catalysts and their enhanced oxygen reduction kinetics under alkaline conditions [3–10].The main technical challenges for the practical application of AEMFCs involve the achievement of both high performance and high durability. Improvements in the chemical/mechanical stability and anion conductivity of the anion exchange membranes (AEMs), crucial components of AEMFCs, have been continuing as part of the effort to meet these challenges [11–17]. Pan et al. reported that a self-cross-linked quaternary ammonia polysulfone (QAPS) developed a hydrophilic/hydrophobic phase-separated structure, and the conductivity exceeded 100 mS cm−1 at 80 °C, similar to that of Nafion [11]. Hassan et al. reported a current density close to 10 A cm-2, which they achieved due to their use of a high IEC membrane, in which they suppressed the swelling by use of cross-linking [12]. Ponce-Gonzalez et al. reported that lengthening the spacers of the side chains improved alkaline stability [13]. In our previous reports, we described the characteristics of an in-house-developed AEM called quaternized poly(arylene perfluoroalkylene), i.e., QPAF-4, which has high anionic conductivity similar to that of QAPS, high mechanical strength, due to the introduction of perfluoroalkylene groups, and high alkali stability, due to the introduction alkylene spacers [14].The development of effective non-PGM catalysts is also crucial. For the anode, primarily non-PGM catalysts based on Ni have been reported [5,7,18–21]. For the cathode, non-PGM catalysts based on Fe, Co, etc. have been reported [7,22–24]. Among these, Hossen et al. reported the remarkable result that an Fe–N–C catalyst had the same performance as that of Pt/C, due to the combination of the N–C material used to synthesize the catalyst and the optimization of the ionomer content of the cathode catalyst layer (CL) [24].In AEMFCs, water management is also extremely important, for both the anode and cathode, because water is produced at the anode via the hydrogen oxidation reaction (HOR) and is consumed at the cathode via the oxygen reduction reaction (ORR). The water moves from the cathode to the anode, due to electro-osmotic drag associated with the movement of OH−, and also moves from the anode to the cathode, i.e., back-diffusion of generated water. AEMFCs must provide enough water to maintain hydration of the AEM and electrodes without flooding or drying of the CLs [25–29]. Kasepar et al. reported improvement of flooding and drying for CLs by controlling the use of microporous layers (MPLs) associated with the gas diffusion layers (GDLs) and the humidification of the feed gases [29]. In another recent study, Mustain et al. reviewed the water management of AEMFCs in detail [30]. Peng et al. reported suppression of flooding by lowering the humidification temperature to the extent that ionomer decomposition did not occur and increasing the hydrophobicity of the GDL and CL [31]. Dekel et al. also reported, by means of simulation, that the water required for the cathodic reaction increases with increasing current density and that the lack of water shortens the AEMFC life [32,33]. Recently, Omasta et al. reported that back-diffusing water is the main source for maintaining the hydrated state of AEMs during cell operation, in addition to being an important water source for the cathodic reaction [34,35]. A limitation of these studies on water management has been that the cells were evaluated under unrealistic conditions involving the use of high gas flow rates (e.g., 1 L min−1 for 4 cm2 cell) in order to achieve maximum power density [35–40]. These evaluation conditions result in low gas utilization rates (3% or less at 1 A cm−2 for hydrogen), far from the specifications of practical AEMFC systems. In the present work, we make use of practical gas flow conditions and focus on these water management challenges, reporting improvements of cell performance for membrane-electrode assemblies (MEAs) using a commercial non-PGM catalyst (Fe–N–C) for the cathode and a novel anion exchange ionomer (QPAF-4, see Fig. 1 ) for both the membrane and the CL binder.QPAF-4 was synthesized based on the synthetic procedure of Ono et al. [14]. Further details of the synthesis and characterization can be found in the Supporting Information (Scheme S1, Fig. S8 and Table S2).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 use of a planetary ball mill for 30 min. Subsequently, 5 wt% QPAF-4-MeOH (ion exchange capacity (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 adjusted to 0.8. In the same way, the catalyst inks for the cathodes were prepared with the Fe–N–C catalyst (XPMF2000E, 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 adjusted to 0.43. These catalyst inks were directly sprayed onto the microporous layers (MPL) of the gas diffusion layer (GDL) as the anode (Carbon cloth GDL, GDL with MPL formed after water repellent treatment of PANEX30 PW03 from Zoltek) and cathode (29BC, SGL Carbon Group Co., Ltd.) by the pulse-swirl-spray (PSS, Nordson Co. Ltd.) technique in order to prepare the gas diffusion electrodes (GDEs). The electrode areas were 4.41 cm2, the Pt loading of the catalyst layers (CLs) was 0.50 ± 0.02 mg cm−2, and the Fe–N–C loading of the CLs was 0.50 ± 0.05 mg cm−2. The prepared GDEs were immersed in 1 M KOH 80 °C for 2 days before measurement in order to ion-exchange to the OH⁻ form. Similarly, the QPAF-4 electrolyte membranes (IEC = 2.0 meq g−1, average thickness = 30 μm) were also immersed in 1 M KOH at 80 °C for 2 days before measurement. Excess aqueous KOH and water were removed from the GDEs and membranes with a laboratory cloth prior to assembly. 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 to 10 kgf cm−2 with four springs.The cell voltages (V) as a function of current density (I) were measured with hydrogen and oxygen at 60 °C at various pressures. The back-pressure (BP) was controlled at 0–100 kPa (gauge: kPag). 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 80–100% 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. In the hydrogen pump test, hydrogen was flowed through both electrodes at 60 °C 100% RH and 100 mL min−1, and the anode overpotential was measured by use of an Automatic Polarization System (HZ-5000, Hokuto Denko Co.) at the same current density positions and stabilization times as those used in the I–V measurements.For a more detailed comparison of the CLs using Fe–N–C and Pt/CB, they were investigated by use of various analytical methods, as follows. The cross-sections of the CLs on GDEs were observed by FIB-SIM (FB2200 and SU3500, HITACHI High-Tech Corp.). The wettability of the CL surfaces was 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 the CL surfaces, and the contact angles were measured. The above reagent was pipetted on each CL formed on 29BC GDL, and the contact angle between the reagent and the CLs was measured with analysis software (FAMAS, Kyowa Interface Science Co., Ltd.).We applied N2 adsorption in order 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 of 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. In the case of the catalyst powders, the powders were directly filled in the spherical cells. The N2 adsorption values of the GDLs were larger than those of the CLs. In order 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 (5 cm × 5 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 in order 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 of 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. The catalyst powders were directly filled in the spherical cells. 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 Fig. 2 , the polarization curves and ohmic resistance changes of the cell using Fe–N–C CL and Pt/CB CL (as reference) as the cathode CL are shown. Fig. 2a shows time courses of voltage and current density with increasing and decreasing current for each cathode CL. The polarization curves and Tafel plots shown in Fig. 2b and d were drawn using the quasi-steady-state voltage data, i.e., the final values observed during each period of current density in Fig. 2a. The cell using the Fe–N–C CL exhibited large hysteresis in the I–V curve, i.e., a large difference in potential between the increasing and decreasing current curves. In the case of the Pt/CB CL, the hysteresis was very small. The ohmic resistance of the cell using the Fe–N–C CL increased with increasing current density; however, in the low current density region, the change with decreasing current density decreased, and hysteresis was observed in this region. The Tafel slopes of the cell using Fe–N–C CL were very different for increasing and decreasing current density. The change of the slope with increasing current density was larger than that with decreasing current density.The Tafel slopes were analyzed with a component analysis technique developed in our laboratory (Fig. 2d and e) [41]. This technique involves fitting the I–V curves with a primary Tafel slope, typically corresponding to a transfer coefficient α of 1.0, which would be 66 mV dec−1 at 60 °C, typically together with doubled (132 mV) and quadrupled (264) slopes. Perry et al. have shown that either gas mass transport or ionic transport limitations can lead to Tafel slope doubling, and the combination of the two can lead to quadrupling [42]. That analysis is based on the hydrogen anode being essentially nonpolarized under acidic conditions. For the AEMFC, it is well known that the hydrogen anode is significantly polarized [5,30,35]. As shown in the Supporting Information, an MEA operated under hydrogen pump conditions exhibited a rather small polarization at low current densities, which would not perturb the low current density region of the H2–O2 cell. However, at high current density, the Tafel slope was 476 mV, corresponding to quadrupling, approximately half of which (238 mV) can be assigned to the hydrogen anode. Thus, an approximate I–V curve for the hydrogen anode can be generated and used to correct the observed cell voltages (see Fig. S1-S6 for further details). The corrected I–V curves for the Pt/CB (Fig. 2d) and Fe–N–C (Fig. 2e) CLs are shown for increasing and decreasing current density. As shown in Fig. S2, the apparent (uncorrected) slope of 532 mV for the Pt/CB H2–O2 cell was decreased to approximately 294 mV (532–238 = 294), i.e., more consistent with a quadrupled slope (264 mV). The additional 30 mV polarization is small but might possibly be due to the additional coupling of water mass transport with gas and ionic transport. The precisely quadrupled slopes, although not observed, are also shown in Fig. 2d and e for reference. For both Pt/CB and Fe–N–C catalysts, the low current density region, with an initial Tafel slope of 56.0 mV, can be assigned to pure kinetic control. Even though there is curvature, the curve-fitting allows us to clearly determine the slopes precisely. For Pt/CB, with both increasing and decreasing current density, the behavior transitioned directly to the quadrupled slope, with additional polarization, bypassing the doubled slope. For Fe–N–C during increasing current density, the I–V curve increased to a significantly higher value, 448 mV, corresponding precisely to slope octupling, most likely due to a strong effect of limited water transport. At high current density, the behavior became unstable, with the potential increasing chaotically, presumably due to the influx of generated water from the anode, giving rise to a deviation from the slope of 448 mV (Fig. 2e); this behavior was time-dependent, as seen from Fig. 2a. During the decreasing-current portion, the I–V curve for the Fe–N–C CL became less steeply sloped (269 mV), i.e., consistent with slope quadrupling, which is most likely due to the relaxation of one of the three types of transport limitation, which we propose would be principally water transport. The precise assignments for these components will be taken up in ongoing work. Further details of the analysis can be found in the Supporting Information, including a summary of the calculated Tafel slopes in Table S1.In order to investigate the factors controlling the I–V hysteresis, the effect of the BPs of the supply gases on performances were evaluated (Fig. 3 a–d). In the case of the cell using Fe–N–C CL, the degree of I–V hysteresis decreased as the BPs on both electrodes increased from 0 to 50, and 100 kPag (Fig. 3a). Under 100 kPag BP on both electrodes, the cell performances for the Fe–N–C CL and Pt/CB CL were comparable, and there was negligible I–V hysteresis (Fig. 3b). Fig. 3c and d show I–V curves with BP applied to only one side, i.e., anode and cathode, respectively. Despite the presence or absence of BP at the anode, I–V hysteresis was observed but was not observed when only the cathode was pressurized to 100 kPag. These results indicate that the I–V hysteresis occurs only in the cell using Fe–N–C CL as the cathode, and the degree of the hysteresis is reduced by applying BP to the cathode.With increasing BP, the amount of water vapor in the gas decreases, the oxygen partial pressure increases, and the amount of liquid water also increases in the CL. The effects of oxygen partial pressure and water vapor pressure are shown in Fig. 3e and f, respectively. The performance of the cell under air was lower than that under O2, but the I–V hysteresis was hardly observed (Fig. 3e). The I–V hysteresis increased with lowering the relative humidity of the gas supplied to the cathode, namely, lowering the water vapor pressure (Fig. 3f). Fig. 4 shows Tafel plots and ohmic resistances with 0 kPag at both electrodes, 100 kPag and 100% RH at the cathode, 100 kPag and 80%RH at the cathode. In the part deviating from the Tafel slope in Fig. 4a, the voltages for both increasing and decreasing current decreased in the order of 100 kPag + 100% RH at the cathode >100 kPag + 80% RH at the cathode >0 kPag, and the I–V hysteresis decreased in that order. On the other hand, for the ohmic resistance (Fig. 4b), the values decreased in the order of 100 kPag + 80% RH at the cathode > 100 kPag + 100% RH at the cathode > 0 kPag. The results of Fig. 4 show that the I–V hysteresis occurred in the mass transport region at high current density, but little ohmic resistance hysteresis was observed in this region. In the cathode, a decreasing amount of liquid water due to a decrease of BP and relative humidity increased the ohmic resistance because of decreasing water content in the cathode ionomer and membrane. These results suggest that the I–V hysteresis is caused by a deficiency of liquid water at cathode reaction sites with increasing current density.In the reaction in the cathode of the AEMFC, water is also essential to the reaction. These results indicate that the key factor controlling the hysteresis is water concentration in the cathode, not oxygen concentration. The lack of water supply in the cathode led to the hysteresis and also the increased resistance of the electrolyte membrane. The results of the BP changes also suggest that the increased amount of liquid water associated with increasing BP suppressed the hysteresis. During increasing current density, the void volume of the CL is large but decreased at high current density due to the voids becoming occupied with liquid water, which is supplied by both the cathode gas and back-diffusing water generated in the anode. These hysteresis phenomena were highly significant in the Fe–N–C CL but were hardly observed in the Pt/CB CL.Therefore, these results suggest that the difference of these hysteresis phenomena might arise from a difference in the absorption capacity of liquid water for both CLs, thus affecting the supply of water at reaction sites in the cathode. In order to suppress the hysteresis in the cell using Fe–N–C at the cathode, it is necessary to find ways to better manage the water supply so as to optimize the trade-off between the number of effective reaction sites and the void volume, which accompany the appropriate types of mass transport at the various current densities.In the contact angle changes for various reagents with different surface tensions, using both CLs (Fig. 5 ), the values for the Fe–N–C CL were approximately 10° lower than those for the Pt/CB CL at each measurement point. This result indicates that the Fe–N–C CL was more hydrophilic than the Pt/CB CL. In a comparision between the Fe–N–C catalyst and the Fe–N–C CL coated with QPAF-4 (Fig. 6 a), the volume for pores below 20 nm decreased 84% as a result of QPAF-4 addition. In a comparison between thePt/CB catalyst and the Pt/CB CL coated with QPAF-4 (Fig. 6b), the volume for pores below 100 nm decreased 85% as a result of QPAF-4 addition. These results show that the Fe–N–C CL pores in the range of 20–100 nm were relatively empty compared to those in the Pt/CB CL, despite the addition of QPAF-4. In comparing the water vapor adsorption isotherms of Fe–N–C and Fe–N–C CL (Fig. 7 a) and Pt/CB and Pt/CB CL (Fig. 7b), the total volume of water vapor adsorption for Fe–N–C was approximately a factor of two larger than that for the Pt/CB, irrespective of the presence or absence of QPAF-4. These results regarding the microstructure and hydrophilicity indicate that the Fe–N–C CL had many hydrophilic voids in the range of 20–100 nm and that the capacity for water absorption was much higher than that for Pt/CB. Both MEAs used the same fabrication technique, binder, and membrane; however, the catalysts differed. Thus, the differences in the porosity characteristics in the CL are due to the differences in the carbon structures of Fe–N–C and Pt/CB.Consequently, we can suggest as a major difference between Fe–N–C and Pt/CB from the morphology analysis that the void volume of the Fe–N–C catalyst absorbed the water generated during increasing current density, and this led to the lack the reactant water at the cathode reaction sites, thus contributing to the I–V hysteresis.The comparison of the cross-sectional structure images in Fig. 8 a and b shows that the Fe–N–C CL was about 15 μm in thickness and had larger pores than the Pt/CB CL, and the Fe–N–C CL contained micrometer-order and submicrometer-order pores. On the other hand, the thickness of the Pt/CB CL was approximately 5 μm and contained fine pores of 1 μm or less. Fig. 8c and d schematically depicts the behavior of the water distribution in the Fe–N–C CL during increasing current density, together with the corresponding Tafel plots. We divided the regions acording to the values of current density in the Tafel plots. Region a' corresponds to reaction kinetics control (56 mV), and the cell voltage undergoes abrupt changes in the degree of decrease with increasing current density in regions a to c. In region a (low current density, Fig. 8c), the potential drops as the supply of water starts to become a limiting factor. The reactant water supplied from Fe–N–C CL pores, originating from water supplied by back-diffusion from the anode, is relatively balanced with the reaction. At that time, the ohmic resistace of cell is maintained at low values, as shown in Fig. 4b (100% RH, 0 kPag). In region b (mid-range current density, Fig. 8d), the cell voltage enters the high-slope regime due to the coupling of all three types of mass transport, i.e., gas, ions, and water, with the latter being the most important. We consider that the reactant water supply from Fe–N–C CL pores becomes a limiting factor for the reaction, due to the insufficiency of water generated at the anode. At that time, the ohmic resistace of the cell also increases due to decreasing membrane water content, as shown in Fig. 4b (100% RH, 0 kPag). In region c (high current density, Fig. 8e), the slope deviates from 448 mV and decreases markedly at current densities over 0.2 A cm−2. The reactant water supplied from Fe–N–C CL pores becomes sufficient, so that water mass transport becomes less of a limiting factor for the cathode reaction, due to the large amount of water generated at the anode. At that time, the increase of the ohmic resistance is reversed due to the rehydration of the membrane from the water generated in anode, as shown in Fig. 4b (100% RH, 0 kPag). In the case of the decreasing current density, going back from region c to a, there was no sudden slope increase of the cell voltage. We consider that the water transport, although still a factor, was less limiting, due to a sufficient supply at reaction sites, because the balance between the water supply from back-diffusion and the reaction demand was maintained as a result of the large water volume in the Fe–N–C CL. These results also show that the cell potentials in region a' for both increasing and decreasing current density did not change. These indicate that there were negligible changes in the number of reaction sites, and basically, the behavior in regions a, b and c can be explained by combinations of various types mass transport, with water playing a major role. In the case of the Pt/CB cathode CL, the ohmic resistance in the current density range over 0.2 A cm−2 hardly changed with increasing current density, as shown in Fig. 2c. In the case of the Fe–N–C cathode CL, however, the values of the ohmic resistances were larger than those of the Pt/CB cathode CL and increased with increasing current density. These results also suggest that the water content of the membrane with the Fe–N–C cathode CL was lower than that of the Pt/CB cathode CL and continued to decrease, and thus, the water permeability in the membrane was also insufficient. More specifically, in the case of Fe–N–C, the total water consumption of the cathode used for the water absorption and reaction of the CL is larger than that of the water back-diffusing from the anode. The large void volume of Fe–N–C is concluded to lead to a decreased number of contact points at the interface between the membrane and the CL, and also a decreased number of supply pathways for back-diffusing water from the anode. These conditions lead to an increased cathode-ionomer interfacial resistance in the high current density region. These results indicate the importance of providing the cathode reactant water with the back-diffusing water from the anode [28,35], while maintaining liquid water at a level that does not cause flooding. We also found that lowering the humidity of the anode increased the hysteresis, not only in Fe–N–C but also in Pt/CB (Fig. S7). This proves that the supply of water to the reaction sites of the cathode via the back-diffusing water from the anode is essential. In order to reduce the hysteresis, it is necessary to enhance the membrane water permeability, thus increasing the amount of back-diffusing water and forming a homogeneous interface between the membrane and the CL.In order to suppress the hysteresis in the cell using Fe–N–C and the higher ohmic resistance of the cell, we must find ways to control the water management conditions. We should optimize the trade-off between the number of effective reaction sites and the void volume, improve the interface between the membrane and the CL, and improve the membrane water diffusivity, which accompany the appropriate types of mass transport in the various current density ranges. By addressing these challenges, the development of the AEMFC will progress.We investigated the cell performance for the AEMFC using a non-PGM catalyst (Fe–N–C) for the cathode and an in-house-developed anion exchange ionomer (QPAF-4) for both the membrane and the CL binder under practical gas flow conditions, i.e., high utilization. The cell using the Fe–N–C CL exhibited large hysteresis in the I–V curve under ambient pressures for both electrodes. Irrespective of the presence or absence of BP in the anode, the I–V hysteresis occurred but was not observed when only the cathode was pressurized to 100 kPag. In the cathode, decreasing amounts of liquid water due to a decrease of BP and relative humidity increased the ohmic resistance because of decreasing water content in the cathode ionomer and membrane. These results suggest that the difference of these hysteresis phenomena arises from the difference in the absorption capacity of liquid water for both CLs, affecting the supply of water at reaction sites in the cathode. We were able to clarify differences between Fe–N–C and Pt/CB also based on morphology analysis. The void volume of the Fe–N–C absorbed the generated water during increasing current density and led to an insufficiency of reactant water at the cathode reaction sites, so that water mass transport became a major limiting factor, causing the I–V hysteresis. The Tafel slope component analysis revealed that the I–V behavior in this region can be characterized as a direct transition from kinetic control (56 mV slope) to combined gas-ion-water transport control, with a unique slope octupling behavior (448 mV slope). These results also support the importance of back-diffusing water from the anode for the rate of the cathode reaction. Kanji Otsuji: Formal analysis, Investigation, Writing - original draft. Naoki Yokota: Formal analysis, Investigation. Donald A. Tryk: Tafel slope component analysis, Writing - review & editing. Katsuyoshi Kakinuma: Validation, Investigation. Kenji Miyatake: Conceptualization, Validation, Resources. Makoto Uchida: Conceptualization, Methodology, Validation, Resources, 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.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).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.jpowsour.2020.229407.

We focus on the water management challenges and report on the improvements of cell performance for anion exchange membrane fuel cells (AEMFCs) using a non-PGM catalyst (Fe–N–C) for the cathode and an in-house-developed anion exchange ionomer (quaternized poly(arylene perfluoroalkylene), QPAF-4) for both the membrane and the catalyst layers (CLs) binder under practical gas flow rates conditions. The cell using the Fe–N–C cathode exhibited similar current-voltage (I–V) performance compared with those using Pt catalyst supported on carbon black. The cell using the Fe–N–C catalyst showed I–V hysteresis between increasing and decreasing current. The hysteresis decreased with increasing back-pressure. Based on the results of various I–V measurements, we conclude that the hysteresis is related to water supplied to the cathode using the Fe–N–C catalyst. Tafel slope component analysis revealed that a severe polarization occurred, amounting to slope octupling, with increasing current density, most likely due to the addition of water transport to the usual combination of gas and ionic transport. This severe polarization was alleviated after the cathode layer became sufficiently hydrated. We found from these results that water management is essential, due to the role of water as a reactant in the cathode reaction, for high-performance AEMFCs.

The harsh reduction in fossil fuel reserves and the increasing concern about the environmental pollution have boosted the development of new strategies for the valorization of alternative sustainable sources in order to mitigate the problems associated with CO2 emissions and global warming [1–3]. In this regard, biomass has been considered one of the most suitable renewable alternatives for future energy and fuels due to its availability and carbon neutral emissions. Thus, in recent years, there is an increasing interest in the development of new strategies for the production of value-added and sustainable biofuels, chemicals, and bioproducts [4–7], in which biomass is used as a raw material. Amongst them, thermochemical processes have deserved a remarkable attention in the literature [1,5,6], particularly biomass steam gasification [8–11], biomass fast pyrolysis [12–15], and the steam reforming of the bio-oil produced in the pyrolysis process [16,17].Biomass steam gasification is one of the most studied and developed technologies for the production of H2 rich syngas. However, the excessive tar content in the syngas is currently a challenge to be overcome [8,18].Alternatively, the bio-oil produced in the pyrolysis reaction has attracted increasing attention for the production of other high value-added products, such as H2, automotive fuels and chemicals, by means of several catalytic and thermochemical routes [19–21]. The bio-oil is a complex mixture of oxygenated compounds and water, and is mainly composed of small carboxyl and carbonyl molecules (acids, ketones, aldehydes), sugar-derived compounds (furans, anhydrosugars), and lignin-derived compounds (phenols, aromatic oligomers) [22,23]. Nevertheless, its direct application involves several drawbacks associated with its properties (low heating value, low volatility, thermal instability and strong corrosiveness [24,25]), along with the difficulties involving its feeding (due to incomplete vaporization and re-polymerization of unstable compounds). Accordingly, bio-oil stabilization (for its further utilization as fuel or raw material in other catalytic processes) has deserved a remarkable attention in the literature. Thus, a wide range of strategies (physical, thermal or catalytic treatments) have been extensively analyzed for bio-oil conditioning, as are: esterification, aldol condensation, ketonization, in situ cracking, and mild hydrodeoxygenation [19,26–29].In spite of the aforementioned drawbacks, the bio-oil can be used as fuel by mixing with diesel (with the bio-oil content being of up to 75 wt%), or for the production of several feedstocks based on bio-oil compounds, i.e., synthesis of wood adhesives or resins from phenolic compounds [30]. Besides, the bio-oil can be catalytically transformed downstream by the following routes: i) deep hydrodeoxygenation (HDO) in order to produce fuels, ii) ex-situ catalytic cracking for the production of olefins and BTX aromatics or for vapour upgrading by carrying out a second step, and iii) steam reforming aimed at the production of H2 [6,19,20,31].In all these processes, the selection of suitable catalytic materials plays a key role for their industrial scale viability. Thus, a wide range of catalysts have been used in order to attain the desired purity of the products obtained, decrease the severity of reaction conditions, attenuate catalyst deactivation and/or reduce catalyst costs [32]. In the hydrodeoxygenation (HDO) process, bifunctional catalysts are considered the most promising ones, since they strike a suitable balance between catalysts activity (provided by the acid sites) and catalyst deactivation by coke deposition [33]. In the catalytic cracking process, catalysts with acidity and shape selectivity, such as zeolite based ones (HZSM-5, Y-type zeolite, H-mordenite and so on), are preferred in order to produce olefins and BTX aromatics [34]. Concerning the steam reforming process, metal supported catalysts (particularly Ni and noble metals based catalysts supported on Al2O3) are commonly used, and extensive research has been made in order to improve their activity and stability [6,16,35].Nevertheless, the fast catalyst deactivation by coke deposition in the aforementioned routes is still the main challenge to overcome [16,34,36]. It is well-established that the mechanisms of catalyst deactivation and coke formation are greatly influenced by the feed composition [37]. Thus, certain bio-oil compounds, namely, aldehydes, saccharides (mainly levoglucosan) and phenolic compounds, such as the guaiacols produced from the thermal degradation of lignin, are considered the main responsible for coke formation in the catalytic pyrolysis process [38–40]. Within this scenario, the upgrading of the bio-oil by catalytic cracking prior to its valorization in a second step may help to overcome the fast catalyst deactivation. Accordingly, different catalytic materials have been widely investigated, as are acid metal oxides (mainly Al2O3), basic materials (such as MgO and CaO), or other transition metal oxides (such as ZrO2, ZnO, TiO2, Fe2O3) [27,41]. Besides, the use of inexpensive catalysts, waste products or natural minerals is gaining increasing attention due to its low cost and availability [24,38,42]. Accordingly, Ro et al. [38] analyzed the product selectivity of lignin pyrolysis in a fixed bed reactor by using low-cost additives (bentonite, olivine, and spent FCC catalyst) as in-situ catalysts, and HZSM-5 catalysts placed downstream in a fixed bed reactor. They reported higher catalytic activity and lower coke deposition when bentonite was tested. Valle et al. [32] approached the modification of the raw bio-oil by its continuous catalytic upgrading over dolomite in a low-cost reaction system. They concluded that the composition of the upgraded bio-oil is suitable for downstream valorization processes, such as the production of H2 by steam reforming or aromatic hydrocarbons by a two-step hydrogenation-cracking process. However, no studies have been reported in the literature on the joint process of continuous biomass pyrolysis and in-line catalytic modification of the volatile stream; that is, there are no detailed studies aimed to ascertaining the main bio-oil compounds responsible for the catalyst activity decay in the steam reforming reactions. The aim of this paper is therefore to analyze the feasibility of continuous bio-oil upgrading for its further transformation into high value-added products. Accordingly, continuous pinewood sawdust pyrolysis and in-line catalytic conditioning by means of different low cost materials (inert sand, olivine, spent FCC catalyst and γ-Al2O3) has been analyzed by paying special attention to product yields and compositions. Thus, a detailed knowledge of the modified stream will allow ascertaining its suitability for further valorization in other catalytic routes (this study focuses on steam reforming). Furthermore, it will also allow understanding catalyst deactivation in order to improve the catalyst performance. The main mechanisms of bio-oil transformation on these catalysts will also be analyzed in this study.The biomass used is forest pine wood (pinus insignis), which has been crushed, ground and sieved to a particle size in the 1–2 mm range. This particle size eases continuous feeding operation. Table 1 summarizes the most important properties (ultimate analysis, proximate analysis and the higher heating value) of the biomass used in this study, whose empirical formula is CH1.47O0.67. The ultimate analysis has been determined in LECO CHN-932 and VTF-900 elemental analyzers. An ultra-microbalance SARTORIOUS M2P is on-line with a computer for the processing of the data provided by the analyzers. The proximate analysis (volatile matter, fixed carbon and ashes) has been determined in a thermogravimetric analyzer (TA Instrument TGA Q5000IR). The higher heating value (HHV) has been measured in a Parr 1356 isoperibolic bomb calorimeter.Inert silica sand, olivine, γ-Al2O3 and spent FCC catalysts have been tested in order to ascertain their capacity for improving the composition of the biomass pyrolysis volatile stream for its subsequent reforming for H2 production. Silica sand and olivine have been supplied by Minerals Sibelco, γ-Al2O3 by Alfa Aesar and the FCC spent catalyst is the one used in the FCC unit at Petronor Refinery in Muskiz, Spain. Thus, materials with different features have been selected: i) inert silica sand, ii) olivine with basic character and activity for reforming oxygenate compounds [43,44], iii) γ- Al2O3 and spent FCC catalyst of moderate acidity and adequate for promoting oxygenate cracking [45–47]. Apart from their suitable catalytic activity, the selection of the materials was based on their low cost, and bearing in mind their possible application as guard beds prior to the stream valorization in a second catalytic step of steam reforming. Furthermore, the use of a spent FCC catalyst involves reusing and therefore valuing a refinery waste material.Prior to use, the spent FCC catalyst was agglomerated with bentonite (50%) in order to increase mechanical strength as well as provide meso and macropores to the catalyst to avoid the blockage of the zeolite external pores by coke deposition [48,49]. Firstly, the spent FCC catalyst was regenerated by calcination with air at 575 °C for 1 h for burning all the coke deposited in the refinery unit. It was then agglomerated by wet extrusion with bentonite, and dried overnight. Finally, the catalyst was calcined at 575 °C for 2 h. All the catalysts were ground and sieved to a particle size in the 0.8–1.6 mm range.The physical properties of the catalysts were determined by N2 adsorption–desorption (Micromeritics ASAP 2010). The chemical composition was measured by X-ray Fluorescence (XRF) spectrometry. This analysis was carried out under vacuum using a sequential wavelength dispersion X-ray fluorescence spectrometer (WDXRF), PANalytical AXIOS, equipped with Rh tube and three detectors. The samples were prepared mixing flux Spectromelt A12 from Merck (ref. No. 11802) with powder catalyst in a ratio of approximately 20:1. Before the chemical analysis, the samples were melted in an induction micro-furnace.The total surface acidity of all materials was analyzed by NH3-TPD in an AutoChem II 2920 Micromeritics equipment. Thus, the procedure was as follows: i) Removal of the possible impurities adsorbed on the sample with a He stream following a ramp of 15 °C min−1 to 550 °C, ii) NH3 adsorption (150 μL min−1) until reaching sample saturation; (iii) desorption of the physisorbed NH3 with a He stream at 150 °C, and (iv) continuous signal recording by TCD of the chemisorbed NH3 following temperature programmed desorption from 150 to 550 °C.The experimental equipment used in this study is shown in Fig. 1 , which is composed of a conical spouted bed reactor (CSBR) and an in-line fixed bed reactor. Continuous biomass pyrolysis (500 °C) was carried out in a conical spouted bed reactor, whose suitable performance for biomass pyrolysis and gasification has already been proven [22,50–53]. The main dimensions of the CSBR are as follow: height of the conical section, 73 mm; diameter of the cylindrical section, 60.3 mm; angle of the conical section, 30°; diameter of the bed bottom, 12.5 mm, and diameter of the gas inlet, 7.6 mm. These dimensions were selected based on previous hydrodynamic studies, and ensure a stable operation in a wide range of gas flow rates [54–57]. 50 g of silica sand (0.3–0.35 mm) were used as bed material in the CSBR. In addition, the unit was provided with a lateral outlet pipe located above the bed surface to continuously remove the char particles from the CSBR.The biomass was continuously fed (0.75 g min−1) into the CSBR by means of a solid feeding system that allowed continuous feeding in the range from 0.5 g min−1 to 5 g min−1. The feeding system consisted of a vessel equipped with a vertical shaft connected to a piston placed below the bed material. As the piston rised, the biomass was fed into the reactor helped by a vibration system.The gas feeding system is provided with three mass flow meters, which allow feeding N2 (used as fluidizing agent during the heating process), H2 (for the reduction of metal-based catalysts in further reforming studies), and air (used for coke combustion). Besides, the water to generate the steam used as fluidizing agent in the pyrolysis step was fed by a high precision Gilson 307 pump. A water flow rate of 3 mL min−1 was used in all runs, with the steam/biomass weight ratio being 4. Before entering the reactor, the water was vaporized and the steam preheated to 500 °C. The CSBR and the preheater were placed inside a radiant oven of 1250 W.The biomass pyrolysis volatiles formed in the CSBR circulate through a fixed bed reactor (600 °C) connected in-line, where the inert or catalyst (silica sand, olivine, γ-Al2O3 or spent FCC) were placed. It is to note that, given the low activity of these low-cost materials, the masses of the catalysts used correspond to the same bed length in all the runs. Thus, the significant differences in the densities of the materials were considered (sand, 2600 kg m−3; olivine, 3300 kg m−3; FCC, 1246 kg m−3, and γ-Al2O3, 1666 kg m−3), and so the bed masses were 44.2 g of silica sand, 46.2 g of olivine, 17.3 g of spent FCC catalyst and 19.9 g of γ-Al2O3. Accordingly, all runs were carried out with a gas hourly space velocity (GHSV) of 3100 h−1. The fixed bed reactor was placed inside a radiant oven of 550 W.Both reactors were placed inside a convection oven kept at 270 °C in order to avoid the condensation of the volatiles formed in the pyrolysis step, which were fed into the catalytic step. The outlet stream was fed into the product condensation device prior to its analysis.The product stream leaving the fixed bed reactor was analyzed in-line by a GC Agilent 6890 provided with a HP-Pona column and a flame ionization detector (FID). The sample was injected to the GC by means of a thermostated line kept at 280 °C to avoid the condensation of heavy oxygenated compounds. Cyclohexane (not formed in the process) was used as an internal standard to validate the mass balance closure, which was fed into the product stream at the outlet of the catalytic reactor. Furthermore, the non-condensable gases were analyzed by means of a micro GC Varian 4900, which allowed detailed quantification of the product stream. The liquid compounds (dissolved in acetone to avoid the clogging of the GC–MS injector) were identified by means of a GC–MS spectrometer (Shimadzu 2010-QP2010S) provided with a BPX-5 (50 m × 0.22 mm × 0.25 µm). The temperature sequence of the oven was as follows: steady heating from 45 °C to 290 °C following a ramp of 3 °C min−1 for separating the volatile products, with this temperature being kept for 5 min in order to ensure total removal of all products from the column. The column was connected to a mass spectrometer, which operated under the following conditions: ion source and interface temperatures 200 °C and 300 °C, respectively, operating in the 40–400 m/z range.The physical and chemical properties as well as the acidity of the materials used are shown in Table 2 . It can be seen that the materials selected for biomass pyrolysis volatile stream modification have several differences in their main properties. Regarding physical properties, it can be observed that silica sand and olivine have a very low surface area, whereas FCC and γ-Al2O3 have BET surface areas of 81 and 100 m2 g−1, respectively. Accordingly, sand and olivine are not porous materials, and spent FCC and γ-Al2O3 catalysts are mesoporous materials with an average pore diameter of 168–169 Å. It should be noted that the spent FCC catalyst is based on HY zeolite; however, it has been agglomerated with bentonite to increase its mesoporous structure for facilitating the diffusion of bulky molecules and so avoid the blockage of zeolite external pores by coke deposition [58]. Thus, mesoporous materials with a uniform pore size promote the interaction of large organic molecules with the active sites [27]. The spent FCC catalyst has a microporous surface area of 57 m2 g−1, which is evidence of the presence of a zeolite on its structure.Concerning the chemical composition of each material, sand and γ-Al2O3 contain small amounts of impurities. The spent FCC catalyst is mainly composed of SiO2 and Al2O3, as well as various metal oxides, which are accumulated in the catalyst in the consecutive reaction-regeneration cycles in the refinery unit. The high amount of Fe2O3 (7.68 wt%) in the olivine is noteworthy, as it plays an important role in its catalytic activity by promoting the reforming of oxygen compounds [59]. Olivine has been widely used in biomass gasification for tar reduction due to its activity for cracking and reforming reactions and its low cost compared to metal catalysts [43,60]. Several researches stated that the catalytic activity of olivine depends on the amount of Fe present on its composition, as well as on its oxidation state, with Fe being more active as its reduction state is increased [43,61].Moreover, the total acidity of γ-Al2O3 is higher than the one of the spent FCC catalyst, 106 and 47 µmolNH3 gcat −1, respectively. The low acidity of the spent FCC catalyst is attributed to the fact that it has been agglomerated with bentonite, which decreased the amount of HY zeolite to 8 wt%. Acid catalysts enhance dehydration and decarbonylation of oxygen components to form carbon monoxide and water as primary products in the deoxygenation reaction [62], as well as cracking, oligomerization, alkylation, isomerization, cyclization and aromatization via a carbonium ion mechanism [27]. It is to note that the moderate acidity of the catalysts used in this case lead to lower deoxygenation activity, but also to lower coke formation by secondary cyclization and condensation reactions [63,64]. Conversely, basic catalysts, such as olivine, enhance ketonization and aldol condensation reactions, leading mainly to the formation of carbon dioxide and water [62].Continuous biomass fast pyrolysis has been carried out in a CSBR at 500 °C with the aim of maximizing the bio-oil yield. Previous studies have shown the good performance of this reactor for biomass pyrolysis [22,51,65]. Thus, this reactor provides several advantages compared with other reactor configurations, namely: i) short residence time of the volatiles in the reactor (of around 20 ms due to the high velocity of the gas, thus minimizing volatile transformation by secondary reactions, and so maximizing the bio-oil yield in the biomass pyrolysis), ii) high heat and mass transfer rates, i.e., the high velocity of both gas and solid phases and their countercurrent contact improve heat and mass transfer rates, and iii) rapid removal of the char from the reactor by the segregation of char from sand in the fountain, which allows continuous operation. Besides, its simple design eases the scalability of the pyrolysis process.According to the previous biomass pyrolysis studies, a moderate temperature of 500 °C minimizes secondary reactions, which is a promising fact to decrease the gas yield from bio oil cracking [22,65]. In fact, the most important parameters for maximizing bio oil production in the biomass pyrolysis are [66]: i) very high heating rates, ii) high heat and mass transfer rates; iii) moderate temperatures (of around 500 °C), iv) very short residence times, and, v) rapid char removal from the reactor. Table 3 shows the yields of the main products obtained in the biomass steam pyrolysis at 500 °C in a CSBR. The inert nature of steam in the biomass pyrolysis has been previously verified, i.e., product distribution is the same as when N2 is used as fluidizing agent [67]. Under the conditions studied, the char yield is 17.3 wt%, and is continuously removed from the pyrolysis reactor by means of a lateral outlet. This char is suitable for the production of diverse products, such as adsorbents, fertilizers, catalyst supports and soil amenders [68–70].As observed, the gas fraction is mainly composed of carbon monoxide (2.3 wt%) and carbon dioxide (4.7 wt%). The low yield of methane (0.2 wt%) and light hydrocarbons (almost negligible) is indicative of the low extent of secondary cracking reactions in the volatile stream [71]. Thus, a high yield of bio-oil is obtained (75.4 wt%), and so the overall yield of volatile compounds fed into the next step is 82.7 wt%.Regarding the bio-oil, more than 100 compounds have been identified in this fraction, and therefore the yields of only the functional groups and main compounds have been included in Table 3. It can be observed that bio-oil is composed of acids (3.1 wt%), aldehydes (2.5 wt%), alcohols (1.8 wt%), ketones (7.3 wt%), phenols (16.6 wt%), furans (2.3 wt%) and saccharides (4.5 wt%). Among the different functional groups, phenols are the most abundant ones, which are formed from the degradation of lignin [72]. The high yield of saccharides is noteworthy, mainly levoglucosan (4.5 wt%), which is the major individual compound in the bio oil, and is obtained as a primary product by cellulose depolymerization.The bio-oil contains a significant amount of water coming from the raw biomass moisture (10 wt%) and also from dehydratation reactions involving cellulose and hemicellulose [73,74]. The high water and oxygen content in the bio-oil, as well as the low pH and low heating value, make bio-oil upgrading to be essential for subsequent use [75,76].The catalysts used for the steam cracking of the biomass pyrolysis volatile stream are γ-Al2O3, spent FCC and olivine. The pyrolysis step has been carried out at 500 °C and the cracking step at 600 °C. Inert silica sand has also been used in order to assess the effect of thermal cracking on the volatile stream. The overall product distribution obtained in the two-step pyrolysis-cracking process is a complex mixture of many compounds, and so the products have been grouped into three fractions: i) the gas fraction composed mainly of CO and CO2, as well as low amounts of H2 and C1-C4 hydrocarbons; ii) the bio-oil, which is a complex mixture of oxygenated compounds and water; iii) the char fraction, which is the non-volatilized biomass fraction. Fig. 2 shows the effect of each catalyst on the product fraction yields. The yields of each fraction obtained in the pyrolysis step at 500 °C have also been included. The char fraction is continuously removed from the pyrolysis reactor, and is not therefore fed into the second catalytic step. Accordingly, the yield of char remains constant (17.3 wt%) in all the runs, independently of the catalyst used.As observed, all the catalysts are active for cracking, as they increase the yield of the gas fraction in detriment of that of the bio-oil. Furthermore, bio-oil cracking is more severe as the acidity of the catalyst is higher. Thus, when the spent FCC catalyst is used, the gas yield increases from 7.3 wt% to 26.1 wt%, and when γ-Al2O3 is used to 32.5 wt%. However, a basic catalyst, such as olivine, has lower cracking activity, as it only increases the gas yield to 18.9 wt%. Apart from their different character, the catalysts have significant differences in their physical properties, with γ-Al2O3 and spent FCC catalyst being mesoporous materials and olivine a non-porous material. Thus, the limited porous structure hinders the diffusion of bulky oxygenated compounds into the bed material, leading to a lower extension of cracking and deoxygenation reactions [77].As observed in Fig. 2, thermal cracking is significant when inert silica sand is used. Thus, the bio-oil yield decreases from 75.4 wt% to 68.3 wt% and the gas yield increases from 7.3 wt% to 14.4 wt%. Therefore, apart from the effect of the catalyst acidity/basicity, the fact that the cracking step is performed 100 °C above that of pyrolysis leads to bio-oil thermal cracking reactions in parallel to catalytic ones. Fig. 3 a and 3b show the effect of the catalyst on the gas fraction composition and on the yields of the individual components, respectively. As observed in Fig. 3a, CO2 is the main compound in the gas fraction at the inlet of the cracking reactor (obtained by pyrolysis at 500 °C). However, when inert sand is used, a sharp increase in CO and CH4 concentrations (45 and 12 vol%, respectively) is observed, at the expense of decreasing that of CO2 (28 vol%), which is due to the thermal cracking reactions.Moreover, the particular features of FCC and γ-Al2O3 catalysts, especially the total acidity, modified significantly the gaseous product composition, leading to the highest concentrations of CO and HCs, which is evidence of the higher extension of the cracking reactions as when compared with the use of olivine or inert sand. Accordingly, when γ-Al2O3 and FCC catalysts are used, the CO concentration accounts for almost 50 vol% of the gas fraction when any one of these catalysts is used, with that of CO2 being 20.0% and 29.1 vol% respectively. Moreover, the use of γ-Al2O3 and FCC catalysts also leads to an increase in CH4 and light hydrocarbon concentrations, particularly that of the olefin fraction, which stem from the decarbonylation of oxygenated intermediates or alkyl aromatics [78]. The higher concentration of CO than CO2, as well as the increase in the yields of aromatics and olefins, was also observed by Ro et al. [38], who analyzed the upgrading of the lignin-derived bio-oil using different catalysts (bentonite, olivine, spent FCC catalyst and HZSM-5). The promotion of decarbonylation over decarboxylation reactions was also reported by Wang et al. [79] in the catalytic pyrolysis of hybrid poplar wood.Significant features are also observed in the composition of the gas fraction obtained using olivine: 33.7 vol% CO, 29.2 vol% CO2, 26.4 vol% H2, 7.1 vol% CH4 and 3.5 vol% C2-C4 hydrocarbons. Thus, apart from the deoxygenation reactions of dehydratation, decarbonylation and decarboxylation, olivine also enhances oxygenate compound reforming and the water gas shift reaction, which lead to the formation of CO, CO2 and H2. This is mainly due to the chemical composition of olivine, with contains Fe0 on its surface, promoting reforming and WGS reactions [80,81].Given the differences observed in the overall gas yield obtained depending on the catalytic material (Fig. 2), the yields of individual gaseous products have been displayed in Fig. 3b. As observed, γ-Al2O3 and spent FCC catalyst, who account for a gas yield of 32.5 and 26.1 wt%, respectively, showed the highest yields of CO, CH4 and light hydrocarbons. Thus, the yields of these compounds increase with catalyst acidity, which is clear evidence that acidity promotes cracking reactions. Besides, these acid catalysts enhance decarbonylation reactions rather than decarboxylation ones, with CO being the main compound in the gaseous stream.Moreover, it is noteworthy that CO2 yield was higher than that of CO and remaining gaseous compounds only when olivine was used. The basic nature of olivine enhances ketonization and aldol condensation reactions, which involve the formation of CO2 and water [62]. Table 4 shows the bio-oil composition once the pyrolysis volatiles have passed through each catalyst bed. As aforementioned, the products identified have been grouped based on their functional groups, and the composition of the main compound families is shown in Table 4.As observed in Table 4, all the catalysts significantly modify the composition of the bio-oil. It is to note that similar trends were observed for both the yields and the concentrations of individual bio-oil compounds. Furthermore, temperature has also a significant influence when these catalysts are used. As aforementioned, the first step of biomass pyrolysis is conducted at 500 °C, whereas the second catalytic step is carried out at 600 °C. Accordingly, the amount of phenols, which are formed from the depolymerisation of lignin macromolecules [82], was substantially reduced with all the catalysts used, with this decrease being especially noteworthy in the fraction of catechols and guaicols.When the inert sand was used, a decrease in the phenolic concentration was observed due to the sharp reduction in the catechol fraction (from 11.0 to 5.6 wt%), with alkyl-phenols and guaicols being hardly affected by thermal cracking. Moreover, the saccharide fraction, which is formed from the depolymerisation of cellulose and hemicellulose and is mainly composed of levoglucosan [74,83], decreased from 6.0 to 4.1 wt% due to the poor thermal stability of the other saccharide compounds [22]. The fraction of acids, ketones and furans, which are formed from the decomposition of cellulose and hemicellulose in the biomass [82,83], did not undergo a substantially modification, whereas the aldehyde concentration increased from 3.3 to 5.4 wt%, mainly by enhancing the formation of benzaldehyde instead of lighter species, such as formaldehyde and acetaldehyde [82]. A considerable reduction in the alcohol fraction was observed as opposed to that of polycyclic aromatic alcohols, which increased from 0.3 to 2.5 wt%. In fact, olefinic alcohols may undergo aromatization reactions leading to heavier polycyclic aromatic alcohols.As mentioned before, the use of spent FCC and γ-Al2O3 catalysts led to low bio-oil yields (56.6 and 50.2 wt% respectively) due to the features of these materials, especially the total acidity (see Table 2), which promoted cracking reactions [27]. The bio-oil composition obtained with these materials is also a consequence of secondary reactions leading to a substantial increase in the hydrocarbon concentration (6.1 and 8.5 wt%, over FCC and Al2O3, respectively). Accordingly, the higher acidity of these catalysts compared to olivine promotes hydrocarbon formation [38].Concerning the phenol functional group, although similar concentrations were obtained (17.9 and 15.9 wt% for FCC and γ-Al2O3 catalysts, respectively), significant differences were observed in the distribution of catechols, guaiacols and alkyl-phenols. Thus, while catechols are the main phenolic compounds in the bio-oil obtained with the FCC catalyst (11.6 wt%), followed by alkyl-phenols (5.2 wt%), the phenolic fraction obtained with Al2O3 catalyst was only composed of alkyl-phenols, which is due to the secondary recombination and cyclization reactions via Aldol condensation [22,74]. The higher selectivity of Al2O3 catalyst to alkyl-phenols revealed the effective dealkoxylation of guaiacols and cathecols [84]. Thus, the guaicol fraction in the volatiles derived from biomass pyrolysis at 500 °C may undergo oxygen-aromatic carbon bond cleavage to form phenol/aromatic hydrocarbons or undergo oxygen-alkyl carbon bond cleavage to form benzenediols or benzenetriols (catechols). This catechol fraction may then be converted into alkyl-phenols by deoxygenation reactions. Guaiacol cracking can be initiated by homolytic cleavages of CH3–O or O–H bonds leading to the formation of methane, dihydroxybenzene (catechols), o-cresol (alkyl-phenol), and 2-hydroxybenzaldehyde, among others [27,85].It is noteworthy that FCC and γ-Al2O3 catalysts led to full disappearance of acids, light alcohols, and saccharides, and to a significant reduction in the concentration of aldehydes and ketones. Thus, the small oxygenate and olefin molecules in the volatile stream formed in the biomass fast pyrolysis may be converted into aromatics via aromatization, with oxygen being released as CO, CO2, and H2O [27,86]. As aforementioned, the FCC catalyst has a microporous structure due to the presence of HY zeolite. The shape-selectivity of this zeolite promotes the diffusion of the mentioned compounds (acids, aldehydes, alcohols, ketones, and furans) into the zeolite channels and the reactions of deoxygenation ending up in the formation of aromatic hydrocarbons [87]. Moreover, the basic properties of the bentonite, which is the binder to agglomerate the spent FCC catalyst, promote ketonization reactions involving carboxylic acids and carbonyl compounds [27]. In the case of Al2O3, its better textural properties, as well as its higher acidity, enhance further decomposition on the acid sites of the catalyst, and therefore increase the hydrocarbon fraction. Furthermore, there is a higher concentration of water in the bio-oil stream treated with FCC and γ-Al2O3 catalysts (45.7 and 50.1 wt%, respectively) due to secondary cracking-dehydration reactions.Regarding olivine, a decrease in the amount of acids, aldehydes, and furans was observed compared to inert sand. The ketone fraction remained almost constant (at around 9 wt%), although longer chain ketones were formed when olivine was used. Therefore, basic catalysts promote, on the one hand, ketonization of acids and, on the other hand, aldol condensation of small ketone and aldehyde molecules to larger chain ketones by carbon–carbon coupling reactions [27,62]. A more detailed analysis is hindered by the complexity of the reactions occurring when the pyrolysis volatiles cross the catalyst bed and the fact that several reactions may occur simultaneously and lead to opposite effects. The concentration of phenols also decreased from 17.5 to 13.1 wt% (mainly guaicol compounds), and the yield of hydrocarbons (mainly naphthalene compounds) increased to 1.1 wt% as a result of secondary cracking reactions. Besides, the presence of Fe0 metal in the olivine chemical composition promotes deoxygenation reactions, leading to an increase in the production of these aromatic hydrocarbons [38].The results shown in Table 4 are evidence of a less oxygenated nature of the bio-oil obtained when acid catalysts were used. Accordingly, the yields of the oxygenated compounds, i.e., all the functional groups shown in Table 4, except the one of hydrocarbons, decreased as follows: pyrolysis 500 °C (67.5 wt%) > inert sand (54.5 wt%) > olivine (46.9 wt%) > spent FCC catalyst (37.4 wt%) > γ-Al2O3 (28.8 wt%).The treatment described in this study pursues the production of an upgraded volatile stream for its further in-line catalytic valorization in a third step for the production of H2 in a steam reforming process or the production of fuels, chemicals and aromatic hydrocarbons by other catalytic routes.As aforementioned, the main challenge to overcome in these processes is the fast catalyst deactivation by coke deposition. Several researches have reported that certain bio-oil compounds, such as phenolic ones, are the main precursors of coke formation [38,39]. However, amongst the different compound lumps contained in the phenolic fraction (alkyl-phenols, catechols and guaicols), it is not clear which is the main responsible for catalyst deactivation. Moreover, other authors have emphasized the relevance of removing the acids from the pyrolysis volatile stream in order to avoid operational problems in further catalytic valorization processes [32]. Thus, Gayubo et al. [40] attributed the formation of deactivating coke to mainly phenols and aldehydes, whereas Remón et al. [88] reported that, apart from guaicol phenolic compounds, furfural (aldehyde) and levoglucosan (saccharide) have high tendency to produce coke in steam reforming reactions.In this study, several low-cost materials have been used downstream the pyrolysis process in order to condition the volatile stream by removing and/or reducing undesirable compounds. The significant differences in the catalysts used led to volatile streams of considerable compositional diversity, which allowed delving into the understanding of the relationship between the composition of the feed into the reforming step and catalyst deactivation.Accordingly, the use of olivine led to a significant removal of acids and phenols, with the latter due mainly to the reduction in the guaicol fraction. Besides, the chemical composition of olivine, with Fe0 on its surface, plays a positive role in the bio-oil oxygenate decomposition and reforming reactions.The use of acid catalysts (FCC and γ-Al2O3) results in a bio-oil composition with a considerable reduction in the aldehyde fraction, and free of acids, alcohols and saccharides at the expense of hydrocarbon formation. The phenolic fraction was considerably reduced compared to the pyrolysis conducted at 500 °C (from 22.0 wt% to 17.9 and 15.9 wt% for FCC and Al2O3 catalysts, respectively) as a consequence of thermal and catalytic cracking. Moreover, the higher acidity of γ-Al2O3 catalyst promoted the conversion of heavy oxygenated compounds into alkyl-phenols, whereas catechols were the major fraction when the FCC catalyst was used.Future studies will be conducted using these low-cost materials (inert sand, Al2O3, spent FCC catalysts and olivine) as guard beds in order to attenuate the fast catalyst deactivation in the in-line biomass pyrolysis-steam reforming process. Thus, the stability of reforming catalysts and their deactivation will be analyzed with the aim of understanding the role played by the volatile composition, and knowledge will be acquired about the main species responsible for the deactivation of the reforming catalyst.The conical spouted bed reactor allows attaining a reproducible volatile stream for its in-line catalytic pyrolysis. The modification of the pyrolysis volatile stream composition by catalytic cracking was analyzed by using different low cost catalysts and inert sand placed downstream in a fixed bed reactor. The features characterizing each material (physical properties, chemical composition and acidity) play a key role in the transformation of the volatile stream, leading to remarkable differences in the distribution and composition of the gaseous stream.The biomass pyrolysis conducted at 500 °C in a CSBR led to a gas yield of 7.3 wt% (with CO and CO2 being the main products), and a bio-oil yield of 75.4 wt%, which was composed of mainly phenols, ketones, and saccharides. At 600 °C, thermal cracking was evidenced when inert sand was used, increasing the yield of the gas to 14.4 wt%, and so reducing that of the bio-oil to 68.3 wt%. Thus, thermal cracking reactions occurred in parallel to the catalytic ones with all the catalysts tested.Bio-oil cracking was more severe as catalyst acidity was increased, i.e., olivine < spent FCC catalyst < Al2O3. Besides, acid catalysts enhanced decarbonylation over decarboxylation reactions, with CO being the main compound in the catalytic cracking. The chemical composition of olivine, with Fe phase on its structure, also promoted reforming and water gas shift reactions, leading to the formation of CO, CO2 and H2.The bio-oil composition was affected when either inert sand or any catalyst was used. Alcohols, saccharides, and especially the phenolic fraction were substantially reduced due to thermal cracking when the inert sand was used. This significance of this drop depended on the catalyst used. The basic character of olivine promoted ketonization of acids, and aldol condensation of ketones and aldehydes, leading to the formation of CO2 and water. Concerning FCC and γ-Al2O3 catalysts, both led to a substantial increase in the hydrocarbon fraction (6.1 and 8.5 wt%, respectively). Accordingly, the acidity of these catalysts played a key role in the cracking of pyrolysis volatile oxygenates, since the acid sites promoted deoxygenation reactions, as well as cracking, oligomerization, alkylation, isomerization, cyclization and aromatization, which greatly increased the hydrocarbons fraction. The phenolic fraction was influenced by the type of catalyst employed by promoting the formation of catechols and alkyl-phenols when FCC and Al2O3 catalyst, respectively, were used.The results provided in this study are of special relevance for further studies wherein the production of H2 will be approached by feeding the bio-oil stream leaving the catalytic process into the two step biomass pyrolysis-steam reforming strategy. Enara Fernandez: Investigation, Visualization, Writing - review & editing. Laura Santamaria: Writing - original draft, Visualization, Writing - review & editing. Maite Artetxe: Writing - original draft, Conceptualization, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition. Maider Amutio: Conceptualization, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition. Aitor Arregi: Validation, Visualization, Writing - review & editing. Gartzen Lopez: Conceptualization, Validation, Writing - review & editing, Visualization, Supervision, Project administration. Javier Bilbao: Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition. Martin Olazar: 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 carried out with the financial support from Spain’s ministries of Science, Innovation and Universities (RTI2018-101678-B-I00 (MCIU/AEI/FEDER, UE)) and Science and Innovation (PID2019-107357RB-I00 (MCI/AEI/FEDER, UE)), the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 823745, and the Basque Government (IT1218-19 and KK-2020/00107).

Biomass pyrolysis and the in-line catalytic cracking of the pyrolysis volatile stream has been approached in this study. The pyrolysis step was carried out in a conical spouted bed reactor at 500 °C, whereas the inert sand or the cracking catalysts (γ-Al2O3, spent FCC and olivine) were placed in a fixed bed reactor at 600 °C. Product analysis was carried out on-line by means of chromatographic methods, and the distribution and composition of the main products obtained have been related to the features characterizing each catalyst (physical properties, chemical composition and acidity). Decarbonylation reactions were favoured over decarboxylation ones when acid catalysts (spent FCC and γ-Al2O3) were used, whereas olivine promoted ketonization and aldol condensation reactions. The Fe species in the olivine structure enhanced reforming and WGS reactions. Bio-oil cracking was more severe as catalyst acidity was increased, leading to an increase in the hydrocarbon fraction. The Al2O3 derived bio-oil was substantially deoxygenated, with a considerable reduction in the phenolic fraction, which accounted mainly for alkyl-phenols. The three materials tested led to a significant decrease in acid and phenolic compounds in the volatile stream, making it suitable for further catalytic valorization for the production of H2, fuels and chemicals.

The structure and nature of the active sites of hydrogenation catalysts based on copper has not been the focus of scientific research since the early 2000′s, despite previous debate regarding the catalytic mechanism. This subject has recently become relevant again due to the demand for stable, Cr-free industrial catalysts in light of the sunset date for CrVI in 2017, rendering the established copper chromite Adkins catalyst obsolete [1–4]. This catalyst consists of a copper oxide phase distributed over a copper chromite spinel phase (CuO·CuCr2O4) and exhibits excellent stability and activity [3]. Investigated as early as 1991, copper aluminate spinel based catalysts (CuO·CuAl2O4) offer an analogous structure to the Adkins catalyst, where Cr is replaced by Al in the spinel phase [5–7]. These alternative catalysts often contain other metals as additional components for industrial hydrogenation processes [5,8]. Several examples of manganese as component in copper aluminate catalysts are available in the patent literature, though an explanation of the role that manganese plays in these catalysts is lacking [6,9,10]. Manganese possibly increases the stability of the catalysts (e.g. against acidic impurities), and it is claimed that manganese is a necessary part of the catalyst. Recently, also (non-spinel type) Cu/Zn oxides were reported as interesting alternative to Adkins catalysts for hydrogenolysis reactions [11].The normal spinel structure, found in copper chromite (CuCr2O4), contains tetrahedrally coordinated CuII cations in A-sites and octahedrally coordinated CrIII cations in B-sites, following the general formula: A ( t e t ) II B 2 ( o c t ) III O 4 Copper aluminate spinel (CuAl2O4) is partially inverse, with CuII and AlIII each found in both tetrahedral and octahedral sites [12]. In manganese aluminate spinel (MnAl2O4), the degree of inversion is dependent on the method of preparation and oxidative transfer between sites can occur, with MnII in tetrahedral sites converting to MnIII in octahedral sites and the formation of an Al2O3 phase with the displaced AlIII [13,14]. Fast redox processes within manganese oxides are considered key for their catalytic activity and Mn is therefore often used as an oxophilic redox promotor and electron scavenger to improve selectivity or activity of metal oxide catalysts [15].The oxidic precursors require an activation step to become hydrogenation catalysts, involving reduction of the copper species under hydrogen flow at elevated temperature. Hydrogen is thought to penetrate the mixed metal oxide bulk to react with CuII ions, yielding H+ and Cu0 [16,17]. Copper atoms migrate to the surface of the catalyst particle and form hemispherical copper nanoparticles in close contact with the residual bulk spinel [18,19]. The protons remain sequestered in the resulting cation deficient lattice in tetrahedral sites previously occupied by Cu2+, covalently bonded to one lattice oxygen and stabilising the structure in the active state [18,20]. XPS measurements combined with XRD showed CuII in tetrahedral sites are reduced to CuI (at 150 °C), which migrate to octahedral sites, and to Cu0 (at 250 °C), whereas CuII in octahedral sites are reduced at higher temperature (300 °C) to Cu0 nanoparticles and CuI, which remain stabilised in the octahedral sites of the spinel [18,21]. In-situ XANES investigations of CuAl2O4 (formed by impregnation of Al2O3 with 5 wt% Cu) showed the final copper oxidation states as 70% Cu0 and 30% CuI, in a spinel-like environment [22].The role of the reduced spinel and the copper nanoparticles in the catalytic mechanism of hydrogenation reactions of CC and CO double bonds is the subject of some debate. Three possible mechanisms are described in previous research, in which the different copper species are assigned different roles. The first possible mechanism was championed by Bechara et al. in 1985, where they found the catalytic activity of isoprene and 1,3-pentadiene hydrogenation to correlate with the amount of CuI in octahedral sites of a copper chromite spinel, as well as with H species occluded in the spinel [23]. Bechara et al. concluded that the oxidisable part of the reduced spinel (surface) was therefore more important than the surface area of the metallic copper and identified the active site as a CuI-H pair [23]. Hubaut et al. extended this research to the selective 1,2-hydrogenation of α,β-unsaturated aldehyde or ketone to the allylic alcohol [24]. This mechanism requires the catalytic activity to be controlled by the amount of CuI present in the activated spinel.The second proposed mechanism identified Cu0 nanoparticles as the site for hydrogen activation under reaction conditions, similar to Group VIII (Pt, Pd or Ni) metal catalysts, but with higher activation energy [25]. Gudkov et al. reported the dependency of the rate equation for the hydrogenation of butyraldehyde on atomic hydrogen. Dissociatively adsorbed hydrogen was confirmed to participate in the reaction mechanism by isotope studies of irreversible butyraldehyde hydrogenation using adsorbed deuterium, where deuterium/hydrogen exchange rates increased with increasing copper content [25]. The capability of metallic copper to dissociatively adsorb hydrogen was reported to be dependent on particle size and the presence of high index faces (211), (311) and (755) [26,27]. It is expected that a mechanism where the copper nanoparticles supply atomic hydrogen and are the active site for hydrogenation would result in catalytic activity correlating with larger metal surface area, which is affected by both particle size and shape.The third mechanism, described by Yurieva et al. for the hydrogenation of acetone by a reduced copper chromite spinel catalyst, depends on both the surface of Cu0 nanoparticles and the bulk spinel lattice [28]. According to the authors, acetone is adsorbed on the surface of the Cu0, which supplies two electrons to the π* orbital of the carbonyl group, giving the carbon a negative charge. Simultaneously, a proton from the spinel lattice transfers to the oxygen to form an alcohol group. This mechanism then describes the migration of the resulting oxidised Cu2+ back into the spinel lattice to occupy a previously vacated cation site, whilst a second proton migrates in the reverse direction to the anionic carbon, allowing the alcohol to desorb from the nanoparticle surface. In this way, the reduced spinel lattice behaves as a Brønsted acid, supplying protons, and the nanoparticle acts as a CuII ↔ e - Cu0 switch, supplying electrons. However, the migration of copper between the spinel and nanoparticle is likely mass transport limiting to the catalytic rate of reaction, making this aspect of the mechanism debatable.In order to determine which, if any, of these three mechanisms proposed for the traditional chromate system is most likely to be correct in the contemporary copper aluminate spinel-based catalyst (CuO·CuAl2O4) the structure of the activated copper aluminate was investigated. Detailed insights were obtained by characterising the bulk catalysts using XRD, TPR, XANES and EXAFS after synthesis by co-precipitation and calcination to give the spinel structure, and during and after activation in hydrogen forming the final catalyst. The catalytic performance in the reduction of model substrate butyraldehyde by copper aluminate catalysts with varying copper metal surface area and a pure copper spinel model CuAl2O3 was studied to deduce structure–activity relationships. In particular, the effect of manganese on the structure and redox properties of copper aluminate spinel-based catalysts, which has not been previously investigated, was used as a tool to probe the active structure. Using a combination of techniques, this work attempts to investigate the role of manganese in structure formation, activation and catalytic behaviour of these industrially relevant catalysts and contribute to the discussion on the reaction mechanism.Copper aluminate (CuO·CuAl2O4) catalysts were synthesised via co-precipitation of the metal nitrates with sodium carbonate. A metal nitrate feed solution (0.6 M Cu(NO3)2·3H2O, 0.6 M Al(NO3)3·9H2O and 0.1 M Mn(NO3)2·4H2O in Mn including catalyst synthesis) and Na2CO3 (2 M) were co-fed into a precipitation vessel containing warm water (50 °C, stirring at 400 rpm) at a rate of 5 mL/min (0.2 M final Cu concentration). The pH was kept at 6.5 by small adjustments to the rate of addition and the precipitates were aged for 1 h (50 °C). The precipitates were subsequently filtered and washed by re-suspending in deionised water until the spent wash fluid had a conductivity ≤ 0.5 mS. The obtained solids were dried at 120 °C overnight and subsequently calcined at 750 °C for 2 h (2 K/min). Catalysts were then activated under H2 flow, heating at 1 K/min to 300 °C. The final temperature was then held for 1 h before flushing with argon and allowing to cool.Pure copper aluminate spinel (CuAl2O4) was prepared in a similar co-precipitation method as described above, using a modified molar ratio of Cu:Al = 1:2. The calcination step was also modified to 800 °C for 2 h (5 K/min). Residual CuO was removed using saturated (NH4)2CO3 solution in an ultrasonic bath for 2 h. Thereafter, the leached spinel was stirred at 50 °C for 30 min, filtered, washed with deionised water and dried at 120 °C overnight. Elemental Analysis (copper and manganese). The samples were digested by treatment with concentrated acids and the metal contents were analysed by photometry. Catalytic tests. Catalytic hydrogenation was carried out on butyraldehyde as a model compound in a 300 mL stainless steel autoclave (Parr) equipped with a heater and overhead stirrer. The reactor was loaded with butyraldehyde (8 g), hexane (100 mL), and n-dodecane (1.2 g) as a GC internal standard. Before addition of the activated catalyst, the liquid phase was de-gassed using argon (Westfalen, 5.0) for 5 min. The autoclave was pressurized with H2 (60 bar, Westfalen, 5.0) and subsequently heated to 120 °C. After reaching the desired temperature, stirring was commenced (750 rpm) and the reaction time started. After 1 h, the reaction was quenched by stopping the stirring and cooling with an ice bath to 17 °C, upon which the pressure was released. The liquid-phase composition was determined by gas chromatography.Liquid-phase composition of the batch reaction products was determined by gas chromatography (HP6890 gas chromatograph) equipped with a HP-1 methyl siloxane capillary column (60 m × 0.25 mm × 0.25 mm) and a flame ionisation detector (FID). The oven program started at 40 °C with a hold time of four minutes which was then heated to 280 °C with 30 K/min. Helium was used as a carrier gas with 1.2 mL/min and the chromatograph was set to operate at constant pressure. Characteristic retention times and response factors (fx ) were determined using calibration standards. The relationship between fx , mass (m) and GC peak integral (A) referenced to the internal standard n-dodecane for component identification and quantification.Catalysts were characterised by X-ray diffraction (XRD) to determine crystallite size and phase composition, temperature programmed reduction (TPR) for reduction behaviour, N2O-chemisorption for copper surface areas, and finally X-ray absorption spectroscopy for oxidation state (XANES) and local structure (EXAFS). XRD. X-ray diffractograms were measured on an X’Pert Pro (PANalytic) with a Bragg-Brentano geometry or a Rigaku desktop X-ray Diffractometer with a Miniflex2 counter detector using Cu Kα radiation in the 2θ range 20-70° and step size of 0.01° 2θ. In-situ XRD was performed on an X'Pert Pro PW 3040/60 by PANalytical with Bragg-Brentano geometry and Cu Kα radiation (λ = 1.54 Å, 45 kV, 40 mA). The instrument was equipped with a Ni-Kβ filter and a solid state detector (X'Celerator). Scanning range was 5–70° 2θ with a step size of 0.017°. The in-situ measurements were conducted in a HTK 1200 sample chamber by Anton Parr on a special sample holder equipped with heating. Reduction gas (5% H2 in N2) was mixed with Bronkhorst MFCs from the pure gases to obtain a flow of 10 mL/min. Heating rates were set to 2 K/min and samples were allowed to equilibrate after reaching the desired temperatures for 15 min. Temperature programmed reduction experiments were conducted on a Micromeritics AutoChem 2910. 20 mg of the sample to be analyzed were fixed in the quartz reactor (Ø= 9 mm) with quartz wool. To account for effects such as varying storage time or air humidity, samples were pre-dried and flushed with helium prior to reduction at 120 °C (10 K/min) for 30 min in helium (15 mL/min). After cooling back to room temperature, the cooling trap was equipped with a LN2/isopropanol cooling bath before the TPR experiments were started. Heating rate was set to 10 K/min with 75 mL/min 2.5% H2 in argon. TPR experiments were concluded usually at 600 °C and never exceeded 800 °C to avoid damage on the equipment. Removal of hydrogen from the reduction gas was observed with a thermal conductivity detector (TCD) detector. By calibrating the TCD integral of the TPR measurement with hydrogen consumption of a known quantity of pure CuO (see SI Fig. 1), the degree of reduction was calculated by relating hydrogen consumption stoichiometrically to copper content in the catalyst. For this calculation it was assumed that CuII is completely reduced to Cu0, as shown in Eq. (2): (2) CuIIO + H2 → Cu0 + H2O N2O chemisorption. Surface area of metallic copper was determined with the N2O pulse chemisorption method (N2O gas purity 99.5%, Westfalen) on a Micromeritics AutoChem 2910. The loop volume was calibrated with nitrogen gas as well as the pulse size for complete conversion with the aid of manual injections. Prior to chemisorption, 200 mg of the sample were fixed in the U-tube quartz reactor and subsequently flushed with helium, followed by the activation procedure described above for the catalysts with pre-drying (120 °C, 30 min, 10 K/min). Flows were adjusted to 20 mL/min to account for the reduced amount of catalyst. After activation, the cool trap was supplied with liquid nitrogen. lt was verified that flowing nitrogen, product of copper oxidation with nitrous oxide, was not affected and only leftover N2O was removed by the cooling trap. Via the TCD-integrals calibrated with pure CuO, a stoichiometric factor of two and a surface density of copper of 1.47 · 10−19 m−2 were determined. The relative error of the copper surface area determination was estimated to be ±2%. In situ-IR with CO chemisorption. For the infrared spectroscopic measurements, a FTS-175 by BioRad was used. A self-made gas cell with a tablet holder as well as heating unit were used. A 2416 by Eurotherm was used as a controller for heating. Gas supplies were connected to the cell with Swagelok tubing. Helium, 1% CO in He and hydrogen gas for the reduction of catalyst tablets were connected and steered by Bronkhorst mass flow controllers. Catalysts were pressed into self-supporting tablets for measurements. After closing of the cell, it was put into the IR spectrometer and connected to the gas supply. The IR was then flushed with N2 for at least 20 min prior to the first measurement. Activation/treatment procedures were set at 5 mL/min gas flows with heating rates of 1 K/min. After activation, the catalyst tablets were cooled to room temperature in helium before CO-adsorption was begun with varying measurement intervals. XAS. X-ray absorption spectroscopy experiments were carried out at several beamlines at Diamond Light Source (UK). Measurements Mn K-edge (6539 eV) were carried out at the B18 beamline, using a Si(111) monochromator with Pt coated optics and harmonic rejection and argon filled ion chambers. A Mn0 foil spectrum was simultaneously obtained with each measurement for energy calibration. However, due to severe oxidation a MnO2 spectrum was used to determine the amplitude correction. Samples were prepared by diluting the oxidic and reduced catalysts in boron nitride and mixing to homogeneity before pressing into pellets (catalyst depth of approximately 1 unit edge step). The reduced catalyst pellets were sealed in an airtight cell with polyimide (Kapton) film windows before placing in the beam. XANES and EXAFS spectra were collected up to k = 12 in transmission and fluorescence mode.In-operando measurements at the Cu K-edge (8979 eV) during catalyst reduction were carried out at the I20 beamline, using a Si(111) monochromator with Rh coated optics and harmonic rejection mirrors. A Cu0 foil spectrum was obtained before measurements were carried out for energy calibration and to determine the amplitude correction. Samples were prepared by diluting the oxidic catalyst in boron nitride and mixing to homogeneity. The material was pressed into pellets and pushed through a metal sieve. Material of a 125–250 μm diameter sieve fraction was transferred to a quartz capillary flow cell (OD 3 mm, wall thickness 0.05 mm) and held in place with quartz wool (catalyst depth of approximately 1 unit edge step). A thermocouple was placed inside the quartz capillary, which was mounted above a hot air blower and attached to a gas flow controller. Hydrogen gas was then flowed through the capillary (15 mL/min) and temperature was ramped to 300 °C (1 K/min). XANES and EXAFS spectra were collected continuously up to k = 16 in transmission mode, with a 26 °C temperature change during each spectrum. The final temperature was then held for 1 h before allowing the system to cool rapidly under He flow (15 mL/min). XANES. X-ray absorption near edge structure measurements were processed in Athena (Demeter 0.9.25, using Ifeffit 1.2.12) for background subtraction and normalisation [29]. A Linear Combination Fit (LCF) of the experimental data was carried out with standard reference spectra (Cu0 foil, Cu2O, CuO, CuAl2O4). The weighting of the fitting standards was forced to sum to 1 and the fits were evaluated with an R-factor (SI Equation 1) and reduced χ2. These values and the accompanying linear combination fits are shown in the supporting information (SI Table 1 and SI Fig. 3) EXAFS. Extended x-ray absorption fine structure measurements were processed in Artemis (Demeter 0.9.25, using Ifeffit 1.2.12) for summation of FEFF computed scattering paths and evaluation of the EXAFS equation (SI Equation (2)) for each path. Scattering paths were generated by FEFF calculations using CIF input files of Cu0, MnO2 and MnAl2O4 (modified to remove positions of multiple occupancy) [30–32]. EXAFS fitting of the experimental data (see Supporting Information, SI Table 3 and 4 for fit details) allowed estimation of the average coordination number (n) of the first Cu-Cu scattering shell. This is possible because the amplitude of oscillations in k-space that result in the first peak after Fourier transformation to R-space is proportional to the average number of neighbouring atoms in the first coordination shell. Nanoparticle diameters (D) and numbers of atoms (N) were in turn estimated from n using SI Equation 3 (correlation plot shown in SI Fig. 4), following the method developed by Beale and Weckhuysen [33]. The relationship between n and nanoparticle size becomes stronger with fewer numbers of atoms, where a greater proportion of the atoms are undercoordinated at the surface and n decreases rapidly.A series of catalysts with including manganese were prepared and compared to the CuO·CuAl2O4 benchmark catalyst. The catalysts were synthesised via controlled co-precipitation in sodium carbonate solutions (with or without addition of manganese). After separation and drying, thermal treatment at 750 °C was necessary for the formation of the spinel structure. Additional catalysts were prepared for comparison in the activity study: Pure (“stoichiometric”) copper aluminate spinel (CuAl2O4; without additional CuO) and CuO·CuAl2O4 catalysts with varying copper metal surface area were obtained by variation of synthesis parameters. The catalytic performance relative to surface area (reported below) demonstrated substantial and complex differences due to the inclusion of Mn. The structures of the (2, 4 and 6 wt%) Mn containing CuO·CuAl2O4 catalysts were therefore studied in detail in comparison to the (0 wt% Mn) benchmark CuO·CuAl2O4 catalyst as oxidic pre-catalysts, and during and after activation by reduction with hydrogen.To explore the catalytic activity of the spinel type CuO·CuAl2O4 catalysts, the batch liquid-phase hydrogenation of butyraldehyde was used as a model reaction. The reaction was performed in an autoclave at 120 °C and 60 bar hydrogen pressure. Before the reaction, the oxidic pre-catalysts were activated in hydrogen at 300 °C for one hour. The yield of butanol is represented as a function of copper nanoparticle surface area in Fig. 1 . Two series of CuO·CuAl2O4 catalysts were measured. In the first series (blue ■), copper nanoparticle surface area was varied by modifying the synthesis conditions, see Section 3.2. In the second series (black ▴), nanoparticle surface area was kept constant with increasing Mn content from 0% to 6. In the first series, a linear correlation of activity with nanoparticle surface area is observed. However, presence of 6 wt% Mn results in significantly lower activity (~48% yield) than the Mn free catalyst (~67% yield) with similar copper surface area (~15 m2/g). In the second series, the yield likewise decreases with increasing manganese content, from 56% at 0 wt% Mn to a minimum of 21% yield at 4 wt% Mn with a slight recovery to 29% yield at 6 wt% Mn. However, while the butanol yield undergoes drastic changes, copper surface area hovers around 9 m2/g, regardless of manganese content. As the activity of catalysts containing Mn is shown not to correlate to Cu0 metal surface area, another factor may be expected also to control catalyst activity. The inclusion of manganese therefore offers a possible additional tool to investigate the active state. Finally, despite exhibiting the lowest metallic copper surface area (6.4 m2/g), the pure spinel phase CuAl2O4 (red ●) prepared as model system yields more butanol than almost all other catalysts. This indicates that the spinel itself, as well as the nanoparticle surface, is important in the reaction mechanism.Manganese has a strong influence on the phase composition and crystallite size of the oxidic pre-catalysts, as shown by XRD analysis (Fig. 2 a and b). Increasing manganese content leads to increased crystallinity of CuAl2O4 spinel (indicated by ●) and CuO phases, indicated by sharper and more intense diffraction peaks. No additional Mn containing phase is formed, but the spinel phase becomes more pronounced after inclusion of Mn, relative to the Mn free catalyst. It is therefore likely that Mn is incorporated into the CuAl2O4 phase which exhibits higher crystallinity than the Mn free catalyst. The increased peak intensity at higher Mn doping is also indicative of the growth of both spinel and copper oxide crystallites. Crystallite diameters for both phases were determined by the Scherrer equation (indicated by ■ and ● in Fig. 2b and later also summarised in Table 3). In the Mn free catalyst, crystallites as small as 9.4 nm (CuO) and 7.3 nm (CuAl2O4) are formed; although the error in spinel particle size determination is large as the diffraction peaks are narrow. At 6 wt% Mn, the crystallite sizes of both phases increase to 16.2 nm (CuO) and 12 nm (CuAl2O4).The oxidic pre-catalyst is reduced with H2 to the active structure, resulting in the formation of Cu0 nanoparticles on the surface of the spinel. The surface area of these nanoparticles was determined by N2O pulse chemisorption for each catalyst (indicated by ▾ in Fig. 2b). Despite the changes in oxidic crystallite sizes, the surface area of copper nanoparticles after catalyst activation remains surprisingly constant (~9 m2/g) with increasing Mn content. To further explore the effect Mn has on the reduction behavior to the active catalyst, TPR, XRD, XANES and EXAFS was carried out.The effect of manganese content on reduction behaviour is very pronounced. TPR studies are shown in Fig. 2c and summarised in Table 3, where the reduction temperature (TM, given by the TPR peak position) is shifted from 319 °C for Mn free CuO·CuAl2O4 to 257 °C for the 6 wt% Mn containing catalyst. The lower temperature of reduction is not only observed in TM, but also in the onset of reduction, which is shifted to lower temperatures too and undergoes a faster rate of reaction with Mn presence, as indicated by the steeper slope of the TCD signal. The degree of reduction of CuII species, determined from the H2 consumption, assuming stoichiometric CuII reduction, is presented in Fig. 2d (▴), alongside the reduction temperature (TM, ●). Up to 97% of total copper in a Mn free CuO·CuAl2O4 catalyst was reduced, whereas 6 wt% Mn decreased the degree of reduction to just 66%. These results suggest that reduction of copper is inhibited in the Mn incorporated spinel. However, the decrease in reduction temperature (TM) suggests that a certain fraction of CuII is simultaneously more easily reduced. It is possible that the presence of Mn results in competing effects which account for this contradictory behaviour.The reduction of an Mn free CuO·CuAl2O4 oxidic catalyst in hydrogen (5% in N2) was studied further using in-situ XRD in order to track the changes in the crystalline phases of the bulk catalyst. The results can be seen in Fig. 3 a, where copper oxide reflections decrease in intensity below 150 °C, but the XRD pattern does not change substantially. Above 150 °C, metallic copper appears while spinel and copper oxide reflections remain visible. At 200 °C all crystalline CuO is reduced and metallic copper dominates the diffractogram. At temperatures > 200 °C the spinel peaks shift to higher angles but remain present up to 600 °C. The shift of the pure spinel (CuAl2O4) reflexes to higher angles is described by Plyasova et al. to be indicative of a contraction of the spinel unit cell caused by reduction of the spinel phase itself [18]. Reduction of pure CuAl2O4 at 300 °C is known to yield metallic copper and a cation deficient spinel phase, into which protons are incorporated that offset the lost positive charge. Intermediate CuI species, Cu2O and CuAlO2, are also known to form [22,34]. The contraction of the spinel lattice is therefore likely due to the replacement of large CuII ions by much smaller protons.For comparison with the Mn free catalyst, 6 wt% Mn-CuO·CuAl2O4 was activated by reduction in hydrogen at 300 °C and also studied using XRD (see Fig. 3b). The CuO reflections are similarly absent in the reduced Mn containing catalyst pattern (grey line), but unlike the Mn free CuO·CuAl2O4, the spinel phase remains distinctive alongside sharp metallic copper peaks. One further essential difference is observed: the spinel peaks are not shifted compared to the oxidic pattern (black line), indicating that the spinel phase with Mn does not undergo a lattice contraction. The absence of a lattice contraction suggests that less CuII is reduced and removed from the Mn containing spinel phase than in the Mn free CuO·CuAl2O4. The reduced lattice seems to be stabilised by the presence of Mn.Since XRD only probes crystalline phases, the reduction was studied using in-situ XAS to investigate the bulk catalyst, including possible amorphous phases. The Cu K-edge XANES spectra of the Mn free CuO·CuAl2O4 can be seen in Fig. 4 a, where the oxidic CuO·CuAl2O4 (black line) exhibits the 1 s → 4p transition at an edge position of 8983.8 eV, (edge position is defined as the maximum point of the first derivative) which lies closer to the edge positions of pure CuO than CuAl2O4 (8983.5 eV and 8986.6 eV respectively, shown in SI Fig. 2). A weak pre-edge at 8976.5 eV is due to the 1 s → 3d transition, forbidden by dipole selection rules, but observed due to 3d/4p orbital mixing [35,36]. This dipole-forbidden transition is characteristic of CuII and is not observed in CuI due to its closed shell d10 configuration [37]. Below 200 °C the peak intensity of the white line and oscillations of the multiple scattering region dampen slightly, but the edge position does not change. The reduction of CuII in the oxidic catalyst occurs between 200 °C and 250 °C, as indicated by the edge shift from 8983.8 eV to 8979 eV, the disappearance of the weak pre-edge feature at 8976.5 eV and the change in oscillations in the multiple scattering region above the edge. At higher temperatures (>300 °C), the catalyst is reduced, with no further spectral changes observed by XAS.Linear combination fitting of the oxidic and reduced catalysts gives more detailed information on the relative contributions to the observed spectra by different copper species. The Mn free CuO·CuAl2O4 is compared to the 6 wt% Mn-catalyst in Fig. 4b with contributions of reference components (CuIIO, CuIIAl2O4, CuI 2O and Cu0, see SI Fig. 3 for fitted spectra) quantified by linear combination fitting of XANES data. Before reduction, the oxidic catalysts are shown to contain CuII in CuAl2O4 spinel and CuO species only, with no CuI contributions to the observed spectra at all. However, the percentage of CuII in the Mn containing pre-catalyst is higher in the spinel phase than in CuO (58:42%), compared to the Mn free catalyst (51:49%). After reduction, CuII is almost completely removed from both CuO and CuAl2O4 phases in the Mn free catalyst, whereas 21% of copper is retained as CuII in the Mn containing CuAl2O4 spinel. This results in 70% of copper species forming Cu0 in the Mn free catalyst, but only 50% Cu0 when 6% Mn is present. The XANES analysis therefore definitively shows that Mn presence results in greater retention of CuII in spinel and lower reduction to Cu0.To assess the role of Mn in the structure, the oxidation state and local coordination of Mn within the catalyst was also investigated using XAS. The XANES of the oxidic and reduced catalysts are shown in Fig. 5 a, alongside three manganese oxide references, with the pre-edge feature A and edge features labelled B1/2, and the white line feature C. The pre-edge feature A is due to the 1 s → 3d transition and becomes sharper and more intense for non-centro-symmetrically coordinated Mn, for example in tetrahedral sites [38]. The pre-edge feature is sharper in the catalyst XANES spectrum than the octahedrally coordinated manganese oxide reference spectra, indicating that the catalyst contains Mn in tetrahedral sites. Comparison with literature examples of Mn K-edge XANES analysis suggests that the reduced catalyst corresponds to a MnAl2O4 spinel, identified by the features A-C which are characteristic of Mn spinel structures [13]. The feature labelled B1/2 are therefore attributed to the symmetry allowed 1 s → 3p transition, which is divided over two peaks due to the distribution of Mn over two sites in the spinel lattice [39].The first derivative is shown in Fig. 5b, from which the first peak (B1) after the pre-edge (A) gives the edge energy E. The K-edge energy shifts are calculated from E-E 0 (where E 0 = 6539 eV) are listed in Table 1 , with the corresponding oxidation state of Mn for each compound. The oxidic Mn containing catalyst has an edge shift of 7.7 eV, indicating an oxidation state between 2 and 3, which does not change after reduction to the activated catalyst. However, although all features A-C are present in the reduced catalyst, the shape of the XANES does undergo a change, with a shift in the position and size of feature C. Although the energy shift of the oxidic catalyst is not higher than the reduced catalyst, based on the position of feature B1, the differences in the spectrum at higher energy indicates contributions from Mn in a higher oxidation state. Linear combination fitting, shown in Fig. 5c, of standard components to the oxidic 6 wt% Mn - catalyst data resolves two 50:50 contributions of MnAl2O4 spinel and MnIVO2 to the measured XANES spectrum.EXAFS fitting was carried out in order to confirm the MnAl2O4 structure identified in the activated catalyst. Fig. 6 shows the k 2-weighted Mn K edge EXAFS data of the 6 wt% Mn - catalyst in k-space (Fig. 6a) and the corresponding k 2-weighted Fourier transformation to R-space (Fig. 6b). The oscillations of the reduced catalyst show a shift to lower wavenumbers which is more pronounced at higher k and the number of contributions also increases. In R-space, the reduced catalyst results in a first shell peak of lower amplitude than the oxidic catalyst, as well as changes is the second shell. The fits for both the oxidic and reduced 6 wt% Mn - catalyst are indicated by the dotted lines in Fig. 6 and the scattering paths and parameters used to generate the fits are summarised in Table 2 . The fit to the oxidic catalyst data was generated using scattering paths from two crystal structures (MnAl2O4 and MnO2) each given a 50% weighting, based on the LCF results, resulting in an R-factor of 0.01. The fit of the reduced catalyst was generated using only MnAl2O4 scattering paths, giving an R-factor of 0.02.In summary, doping the CuO·CuAl2O4 catalyst with 6 wt% Mn results in incorporation of approximately half the Mn into tetrahedral sites of the spinel lattice with oxidation states between 2 and 3. The remaining Mn is located in octahedral sites, either within the spinel lattice or in MnO2 type structure, with an oxidation state of 4. Activation results in reduction of the MnIV to an average oxidation state of 2.5 accompanied by a decrease in the average coordination number (n) from 5 to 4. One explanation is that MnIV in octahedral sites also migrates to tetrahedral sites within the spinel upon reduction.In order to investigate the copper nanoparticles formed on the reduced catalysts, EXAFS analysis of in-situ activated catalysts was carried out. Fig. 7 shows the k 2-weighted Cu K edge EXAFS data of the Mn free CuO·CuAl2O4 and 6 wt% Mn - catalyst in k-space (Fig. 7a) and the corresponding k 2-weighted Fourier transformation to R-space (Fig. 7b). The oscillations of the Mn containing catalyst in k-space are noticeably dampened compared to the Mn free CuO·CuAl2O4. This produces less intense peaks in R-space and is indicative of smaller nanoparticles as a result of Mn presence of the spinel [33]. Assuming hemispherical copper particles, as previously reported by Plyasova et al., with face centered cubic (fcc) structure, the average nanoparticle sizes are quantified Table 3 in terms of diameter and number of atoms [18]. Data fitting to correlate the average coordination number of atoms in the nanoparticles to the size of the nanoparticle is shown in the supporting information (SI Equation 3 and SI Fig. 4). The Mn containing catalyst exhibits smaller nanoparticles, containing just 70 ± 20 atoms with a hemispherical diameter of 1.3 ± 0.2 nm, compared to the Mn free catalyst nanoparticle size of 210 ± 50 atoms and diameter of 2.6 ± 0.3 nm. The presence of 6 wt% Mn therefore results in nanoparticles that are on average half the diameter of those of the Mn free catalyst, with approximately one seventh of the number of atoms.To further investigate the role of spinel, pure CuAl2O4 was prepared (Materials and Methods), reduced and analysed in-situ with IR-spectroscopy coupled with CO-chemisorption, to probe the outermost catalyst surface. The copper aluminate spinel exposed to CO exhibited the IR bands shown in SI Fig. 5 (descriptions of experiments and assignment of IR bands also in SI) [40]. As expected, an activated spinel consists of Cu0 as well as of CuI/CuII ions on its surface. Unfortunately, the oxidized species are not distinguishable from each other by this technique. The experiments indicate reduction of CuI/II ions to metallic copper by CO. The redox reaction with CO at room temperature shows that the surface is highly labile, with multiple oxidation states of copper indicating availability and activity of these species also at the outermost surface.As described in the introduction, the mechanism of hydrogenation over copper spinel catalysts is disputed in the literature. Several different possible active species in the reduced spinel have been proposed: a CuI species, the Cu0 nanoparticles, and a combination of both nanoparticles and the spinel support [17,23–25]. Several accessible oxidation states of copper have been identified during reduction of the co-crystallised CuO·CuAl2O4 mixed phase catalysts. Firstly, the CuO phase appears to be directly transformed into crystalline Cu0, with no intermediate steps or formation of other crystalline species. Previous research on reduction of CuO under industrially relevant conditions also showed that the reduction of CuO to Cu0 occurs in one step [41,42]. However, CO-probing the pure spinel surface with IR measurement indicates the presence of either CuI or CuII after reduction (SI Fig. 5). Plyasova et al. also suggested that not all spinel copper is reduced completely to Cu0 but that some CuII ions are retained within the lattice [43].The more detailed linear combination fitting of XANES data of reduced catalysts definitively showed the formation of a CuI species, likely located within the spinel, as well as significant amounts of CuII remaining in the reduced Mn containing spinel phase (Fig. 4). The reaction mechanism suggested by Bechara et al. involves CuI stabilised in octahedral positions as the active species, as they were able to correlate the amount of CuI ions to activity [23]. However, in this research the amount of CuI species in the reduced catalysts increases slightly with Mn presence (Fig. 4b), whereas activity was shown to decrease (Fig. 1). This indicates that CuI cannot be crucial to the active state, so the research mechanism proposed by Bechara et al. is unlikely to be correct for the catalysts and reaction studied here [23].The cation deficient reduced spinel lattice forms the bulk of the reduced catalyst, with copper metal nanoparticles growing upon its surface. It is also known that copper nanoparticles can be active sites for the dissociative adsorption of H2 [25]. Yurieva et al. went further to suggest that metallic copper provides electrons for the hydrogenation, which leads to the expectation that activity must correlate with copper metal surface area [17]. This is not the case, as catalytic activity is reduced by Mn presence whereas the metal surface area remains constant (Fig. 1, Table 3). A mechanism only controlled by dissociative adsorption of H2 on metallic copper cannot fully explain the observed trend. Furthermore, pure CuAl2O4 with a smaller copper metal surface area yielded significantly more butanol than expected based on a linear trend between surface area and activity (Fig. 1). This indicates that the reaction mechanism does not only involve the copper metal nanoparticle surface, but is also affected by the structure of the spinel and thus more complex.Incorporation of Mn into the spinel lattice has a profound effect on the spinel structure. The lattice is stabilised by raising the relative amount of spinel to copper oxide and increasing the spinel crystallite size. This appears to inhibit the reduction of CuII, causing copper to be retained within the lattice which therefore does not undergo contraction, unlike the Mn free catalyst. At the same time, the reduction of CuII is facilitated in terms of the lower reduction temperature (Tm, Fig. 2c, Table 3). A variety of mechanisms to explain how Mn simultaneously lowers Tm and prevents a significant percentage of CuII from being reduced are possible. Mn is a known redox catalyst with several possible oxidation states and could therefore modify the reduction behaviour of the CuII spinel due to its different redox potential. The spinel structure contains tetrahedral and octahedral cationic sites and Mn is known to occupy both, as either MnII or MnIII in different types of spinels [14]. XANES and EXAFS measurements of the reduced catalyst showed MnII/III in predominantly tetrahedral sites, which are known to be vacated by CuII upon reduction to Cu0 [18]. The MnAl2O4 type local structure identified is known to undergo oxidative transfer of MnII in tetrahedral sites to MnIII in octahedral sites under certain conditions, which are not reduced under 800 °C in 5%H2/Ar [14,44]. It is therefore possible that MnII, in tetrahedral sites, could act as an electronic promoter within the spinel, increasing the availability of electrons to CuII and resulting in a lower temperature of copper reduction.The decrease in reducible copper could be explained by two further effects in the Mn containing catalyst. The simple increase in crystallite size could result in pathways to the spinel surface that are simply too long for copper to migrate. In addition, blocking of Cu migration pathways by MnII/III ions incorporated into tetrahedral sites of the spinel structure is also possible. It is considered likely that a combination of all three mechanisms is at play. A lowered redox potential gives rise to a spinel more active towards reduction and accounts for the earlier onset and lower temperature of CuII reduction in Mn containing catalysts. However, inhibited migration due to larger distances and blocking of pathways could account for the large fractions of CuII that are not reduced in the manganese containing spinel.Yurieva et al. championed an interfacial reaction mechanism where protons are supplied by the lattice and electrons are supplied by Cu0 for the reduction of the reactant molecules, after which the resulting CuII ions migrate back into the lattice [17]. This type of mechanism, involving the redox reaction of CuII ↔ e - Cu0, requires access to multiple copper oxidation states in the activated catalysts, which we have shown to be the case. However, in order for this mechanism to be catalytically relevant, unrealistically high rates of ion migration in and out of the lattice would be required per cycle. We therefore propose a modified mechanism: a predominantly interface-based reaction mechanism, involving both the spinel and copper nanoparticle surface (illustrated in Fig. 8 ). In this mechanism, the electron reservoir of the metallic copper particle is refreshed by the dissociative adsorption and oxidation of H2 on the nanoparticle surface. The resulting protons are then stored inside the spinel phase and are able to migrate quickly through the outermost layers of the lattice to active sites on the surface. Besides the lower reduction degree of copper in the manganese containing catalyst, an additional effect of Mn in such a mechanism could be to reduce proton capacity of the lattice due to occupation of cation sites by MnII/III and non-reduced CuII, which have been shown to be retained in the spinel by XANES analysis (Figs. 4b and 5). The incorporation of Mn ions in the lattice could also block or influence proton migration pathways through the lattice, slowing the resupply of protons to the active sites on the spinel surface. Both, or either, of these effects could account for the reduced activity of the Mn containing catalyst.Multiple techniques have been used to investigate the structure of the active Mn containing and Mn free CuO·CuAl2O4 catalysts for hydrogenation of butyraldehyde. Mn presence results in a stabilisation of the spinel towards reduction of CuII to Cu0 by occupation of tetrahedral sites within the spinel by Mn cations, causing decreased catalytic activity. This has implications regarding the catalytic mechanism and highlights the role of the spinel lattice as a proton reservoir. A modified mechanism is therefore proposed, where the copper nanoparticle acts as a site for H2 dissociation and supplies electrons and the spinel lattice stores and delivers protons, with the catalytic reaction likely occurring at the interface.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 Clariant for funding (to C.D. and K.K.), and NWO and Clariant for funding (NWO LIFT 731.015.407, Launchpad for Innovative Future Technology, PreCiOuS, to M.T. and M.H.). K.K. thanks the University of Amsterdam for an HRSMC (Holland Research School of Molecular Chemistry) fellowship. The authors thank the staff of the I20 and B18 beamlines at Diamond Light Source in Didcot, UK (proposal number SP16558-1) for support and access to their facilities. The authors thank Bas Venderbosch, Jean-Pierre Oudsen and Lukas Wolzak for assistance during synchrotron measurements. The authors would like to thank Ulrike Ammari, Petra Ankenbauer and Bircan Dilki from the microanalytical laboratory at Technical University of Munich for the conduct of elemental analyses.Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcat.2020.12.017.The following are the Supplementary data to this article: Supplementary data 1

Copper aluminate spinel (CuO . CuAl2O4) is the favoured Cr-free substitute for the copper chromite catalyst (CuO . CuCr2O4) in the industrial hydrogenation of aldehydes. New insights in the catalytic mechanism were obtained by systematically studying the structure and activity of these catalysts including effects of manganese as a catalyst component. The hydrogenation of butyraldehyde to butanol was studied as a model reaction and the active structure was characterised using X-ray diffraction, temperature programmed reduction, N2O chemisorption, EXAFS and XANES, including in-situ investigations. The active catalyst is a reduced spinel lattice that is stabilised by protons, with copper metal nanoparticles grown upon its surface. Incorporation of Mn into the spinel lattice has a profound effect on the spinel structure. Mn stabilises the spinel towards reduction of CuII to Cu0 by occupation of tetrahedral sites with Mn cations, but also causes decreased catalytic activity. Structural data, combined with the effect on catalysis, indicate a predominantly interface-based reaction mechanism, involving both the spinel and copper nanoparticle surface in protonation and reduction of the aldehyde. The electron reservoir of the metallic copper particles is regenerated by the dissociative adsorption and oxidation of H2 on the metal surface. The generated protons are stored in the spinel phase, acting as proton reservoir. Cu(I) species located within the spinel and identified by XANES are probably not involved in the catalytic cycle.

The development of the Haber Bosch Process was undoubtedly a major achievement of the 20th Century. Through provision of a route to access synthetic fertiliser, it can be credited with sustaining 40% of the global population. It has been estimated that it is responsible for 2% of the global commercial energy requirement [1]. This figure takes into account the production of reactants for which the hydrogen required is generally prepared via fossil based feedstocks. As a whole the process, which currently runs on a production scale of ca 174 million tonnes per annum, accounts for 670 million tonnes of CO2 emissions per year equating to around 2.5% of worldwide fossil fuel based CO2 emissions [1]. It is therefore a very important major process for which a number of superlatives apply. On the industrial scale, ammonia synthesis is conducted at high pressure (>100 atmospheres) and moderate temperatures, ca. 400 °C. The catalysts employed are either based upon iron or ruthenium. Although thermodynamically ammonia synthesis is favoured by lowering reaction temperature, the conditions employed are dictated by the requirement for acceptable process kinetics. In current commercial application, the process is highly integrated and very efficient. However, with the increasing availability of renewable electricity it is becoming more and more practical to produce the hydrogen required for localised ammonia synthesis via, for example, electrolysis. The possibility for localised sustainable “green” ammonia synthesis which could, for example, be conducted on farmland for provision of on-demand fertiliser is a strong driver for the discovery of more active catalysts which can be applied more easily under such conditions. There is also increasing interest in the application of ammonia as a fuel or hydrogen carrier. There are a number of alternative directions currently being investigated in terms of sustainable ammonia production and these include electrocatalytic approaches [2,3], photocatalytic ammonia synthesis [4,5] and chemical looping routes [6,7].In terms of investigation of catalysts for ammonia synthesis, it is generally the case that individual studies in the literature tend to focus in detail on small subsets of related materials and it is often difficult to benchmark such materials against different material classes reported by other groups due to issues such as differences in experimental procedure. This is a driver for the present manuscript in which we present an empirical overview of the performance of a wide range of supported catalysts which have been selected on the basis of their expected activity. Relevant considerations which have been documented in the literature are highlighted presenting a structured overview of some of the previous literature in relation to the materials screened. Further investigation would involve performance evaluation under conditions of greater relevance to application, such as operation at higher pressure and inclusion of ammonia in the feedstream. The conditions which have been selected correspond to those reported in a number of studies and which we have applied in our previously published studies to discern structure-composition and activity relationships which might otherwise be obscured by operation at higher reaction pressure and/or the inclusion of a low level of ammonia in the reactant feed. Accordingly, the study is intended to be a starting point for further development of active catalytic materials and the systems we have screened have neither been optimised nor characterised in great detail. The data reported suggest potential directions for further investigation where further detailed systematic investigation may, for example, provide the basis for computationally aided design [8].The materials screened within this study were prepared as follows. 5% Ru/Al2O3 was used as commercially obtained (Sigma Aldrich, Ru 5 wt. % on alumina, powder, reduced, dry). The material was pretreated at 500 °C under 60 mL/min of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%).5% Ru/Al2O3 + 1% KOH - approximately 1 g of Ru/Al2O3 (Sigma Aldrich, Ru 5 wt. % on alumina, powder, reduced, dry) was impregnated by dropwise addition of an aqueous solution of 1 wt precent KOH (Sigma Aldrich, reagent grade, 90%, flakes). The material was then dried in air overnight at 110 °C and was pretreated at 500 °C under 60 mL/min of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). 5% Os3(CO)12/SiO2 and 5% Os3(CO)12/γ-Al2O3 - triosmium dodecacarbonyl (Os3(CO)12, Sigma Aldrich, 98%) was supported onto silica (amorphous, precipitated, Sigma Aldrich) or alumina (γ-alumina, Condea Chemie, alumina extrudates) using the method outlined by Collier et al. [9]. The support was impregnated with a solution of Os3(CO)12 in dichloromethane. The volume of dichloromethane required was determined by point of wetness for each support. The material was then dried at 40 °C to remove the dichloromethane to produce a yellow powder. The material was prepared to achieve a 5% loading by weight of osmium. The material was pretreated at 500 °C under 60 mL/min of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). 5% Os3(CO)12 /SiO2 + 1% KOH - approximately 1 g of 5% Os3(CO)12/SiO2 was impregnated by dropwise addition of an aqueous solution of 1 wt percent of KOH (Sigma Aldrich, reagent grade, 90%, flakes). The material was then dried in air overnight at 90 °C. The material was pretreated at 500 °C under 60 mL/min of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). 5% Os3(CO)12/SiO2 dehydroxylated - silica (amorphous, precipitated, Sigma Aldrich) was dried following the method detailed by Collier et al. [9]. The support was dried at 500 °C for 16 h under a flow of nitrogen at 60 mL/min. The ramp rate for heating was 10 °C/min. After 16 h, the material was cooled under nitrogen. Triosmium dodecacarbonyl (Os3(CO)12, Sigma Aldrich, 98%) was then supported onto the dried silica. The support was impregnated with a solution of Os3(CO)12 in dichloromethane. The volume of dichloromethane required was determined from the point of incipient wetness for silica. The material was then dried at 40 °C to remove the dichloromethane to produce a yellow powder. The material was pretreated at 500 °C under 60 mL/min of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). 10% CoRe supported on silica, alumina and zirconia - 5 g of silica (amorphous, precipitated, Sigma Aldrich), alumina (α-alumina, Fisher Chemicals, aluminium oxide-calcined) or zirconia (monoclinic zirconium (IV) oxide, Sigma Aldrich, powder, < 5 μm, 99% metal basis) were impregnated simultaneously with an aqueous solution of ammonium perrhenate (NH4ReO4, Sigma Aldrich, assay, form, powder or crystals, ≥ 99%, 0.55 g) and cobalt nitrate hexahydrate (Co(NO3)2·6H2O, Sigma Aldrich, ACS reagent, ≥ 98%, 0.60 g). The material was stirred for 1 h at room temperature. The material was then dried at 125 °C overnight. Following this, the material was calcined in air at 700 °C (applying a 10 °C/min ramp rate) for 3 h. Prior to reaction, 0.5 g of the sample was pre-treated for 2 h at 600 °C under a 60 mL/min flow rate of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). The material was prepared to give a 10 wt% of CoRe on support with a 1:1 wt ratio of cobalt to rhenium. 5% Re/SiO2 - 5 g of silica (amorphous, precipitated, Sigma Aldrich) was impregnated simultaneously with an aqueous solution of ammonium perrhenate (NH4ReO4, Sigma Aldrich, assay, form, powder or crystals, ≥ 99%, 0.55 g). The material was stirred for 1 h at room temperature. The material was then dried at 125 °C overnight. Following this, the material was calcined in air at 700 °C (applying a 10 °C/min ramp rate) for 3 h. Prior to reaction, 0.5 g of the sample was pre-treated for 2 h at 600 °C under a 60 mL/min flow rate of 3:1 H2/N2 (BOC, H2 99.998%, N2 99.995%). The material was prepared to target a loading of 5 percent by weight of rhenium. 10% CoRe/MgO - 5 g of magnesium oxide (Sigma Aldrich, 325 mesh 99%+ metals basis) was impregnated simultaneously with aqueous solutions of precursors (5 wt.% loading of Co and 5 wt.% loading of Re), then stirred for 60 min at room temperature. The material was then dried at 125 °C for 12 h and then calcined in air at 700 °C using a ramp rate of 10 °C/min for 3 h. 0.5 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 mL/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 600 °C for 2 h using a ramp rate of 10 °C/min.5% Ni2Mo3N/SiO2 - 5 g of silica (amorphous, precipitated, Sigma Aldrich) was impregnated simultaneously with aqueous solutions of precursors (2 wt.% loading of Ni and 3 wt.% loading of Mo) and stirred for 60 min at room temperature. The material was then dried at 150 °C overnight. 0.6 g of this material was then calcined in 5 mL/min flowing N2 at 700 °C using a ramp rate of 10 °C/min for 6 h. 0.4 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 mL/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 700 °C for 2 h using a ramp rate of 10 °C/min. 10% Co/SiO2 - 5 g of silica (amorphous, precipitated, Sigma Aldrich) was impregnated with an aqueous solution of precursor (10 wt.% loading for Co) and stirred for 10 min at room temperature. The material was then dried at 110 °C overnight and then calcined in air at 600 °C using a ramp rate of 10 °C/min for 4 h. 0.5 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 mL/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 600 °C for 3 h using a ramp rate of 10 °C/min. 10% Co/α-Al2O3 - 5 g of α–alumina (α–Al2O3, Fisher Chemicals, aluminium oxide-calcined) was impregnated with an aqueous solution of precursor (10 wt.% loading for Co) and stirred for 10 min at room temperature. The material was then dried at 110 °C overnight and then calcined in air at 600 °C using a ramp rate of 10 °C/min for 4 h. 0.5 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 mL/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 600 °C for 3 h using a ramp rate of 10 °C/min. 10% Mo2N0.78/SiO2 - 5 g of silica (amorphous, precipitated, Sigma Aldrich) was impregnated with an aqueous solution of ammonium heptamolybdate tetrahydrate ((NH4)6Mo7O24.4H2O, Sigma-Aldrich (Germany) Puriss p.a., ACS reagent ≥99.0% (T)) corresponding to 10 wt.% loading for MoO3 and stirred for 10 min at room temperature. The material was then dried at 110 °C overnight and then calcined in air at 450 °C for 2 h using a ramp rate of 10 °C/min. 0.5 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 mL/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 700 °C for 2 h using a ramp rate of 5 °C/min. 10% MoPOMSi/α-Al2O3 and MoPOM/α-Al2O3 - 5 g of α–alumina (α–Al2O3, Fisher Chemicals, aluminium oxide-calcined) was impregnated with an aqueous solution of precursor (either phosohomolydic acid (H + PMo12O40, Sigma-Aldrich USA) or 12-molydosilicic acid (H4Mo12O40Si, Strem Chemicals USA) corresponding to 10 wt.% loading for MoO3) and stirred for 10 min at room temperature. The material was then dried at 110 °C overnight and then calcined in air at 450 °C for 2 h using a ramp rate of 10 °C/min. 0.5 g of this material was placed in a quartz glass reactor tube and charged to the reactor, it then underwent pre-treatment by flowing gas mixture of 60 ml/min N2:H2 (BOC, H2 99.998%, N2 99.995%) (1:3) at 700 °C for 2 h using a ramp rate of 5 °C/min. In relation to terminology, “POMSi” refers to the material prepared using the 12-molybdosilicic acid precursor and “POM” refers to that derived from phosphomolybdic acid. The corresponding silica and zirconia analogues were prepared analogously using silica (amorphous precipitated, Sigma Aldrich) and monoclinic zirconia (Sigma–Aldrich (United Kingdom) zirconium (IV) oxide, powder, <5 μm, 99% metal basis) respectively as supports. The same sample abbreviation indicating the precursor is used as for the case of α–Al2O3.Catalytic activity was evaluated using a fixed bed microreactor operating at ambient pressure. 0.3 – 0.5 g of material was loaded into a quartz reactor tube, held between quartz wool plugs and heated using a Carbolite furnace. Brooks mass flow controllers were used to deliver 60 mL/min of 75 vol % H2/N2 (BOC, 99.98%) reactant gas through the reactor bed at the specified reaction temperatures. Ammonia production was determined by observing the decrease in conductivity of 200 mL of a 0.0018 M solution of H2SO4 which the exit stream of gas flowed through. The rates reported in the present study correspond to steady state reaction conducted over a minimum period of 8 h. Fig. 1 presents the mass normalised rates for ammonia synthesis determined at 400 °C for various materials. All materials were screened under this reaction condition but only a few were found to exhibit activity. It is notable that the K+ promoted Ru/Al2O3 presents the highest activity at 400 °C, with a pronounced enhancement due to K+ doping being evident. This is consistent with the literature in which Ru is considered to be a close to optimum catalyst [10] and which can be further promoted by the addition of alkali metals [11]. Such promoters, which include K+, are believed to function via donating electron density to the Ru surface, an effect which can be seen indirectly in the infra-red spectra of adsorbed N2 molecules [11]. Ru is also known to be a strongly structure sensitive catalyst for the reaction with activity being related to the B5 step site which gives a pronounced particle size dependence [12], although it has been argued that mixed particle size distribution is most effective with larger particles activating mobile hydrogen species which migrate to small Ru particles promoting NHx hydrogenation in the case of Ru/Al2O3 catalysts [13]. Modification of the support is also known to be of importance in the case of Ru catalysts with MgO [14], BN [15] and electrides [16] being reported to enhance performance. In the current study, we have applied a commercial Ru/Al2O3 reference as a benchmark and as such the material is not optimised. Whilst Ru, which forms the basis of the commercial KAAP catalyst [17], has attracted a lot of interest in the literature due to its high activity the Co-Re material which possesses the next highest activity at 400 °C in our study has been seldom studied. In fact, to our knowledge, this is the first report of the performance of MgO supported Co-Re. Previous reports of performance have focussed upon bulk CoRe4 systems [18,19]. Active materials were originally prepared by ammonolysis and the suggestion had been made that the presence of Co stabilised an active rhenium nitride phase [18]. Subsequent studies have shown that catalysts of enhanced activity can be prepared by replacement of the ammonolysis step by 3:1 H2:N2 pre-treatment [19]. When the 3:1 H2:N2 pretreatment mixture is replaced by 3:1 Ar:H2 activity develops following a short induction period [19]. Rhenium nitride has been reported to be an active catalyst which decomposes to yield lower activity rhenium metal during the course of reaction [18,20]. In–situ XAS based studies of bulk CoRe catalysts prepared via the H2:N2 and Ar:H2 pretreatment steps shows that the active state of the catalyst is a complex mixture of bimetallic and monometallic species with no definitive evidence for a nitride phase being present [21]. In that study activity development was related to Co-Re mixing. In the context of the current study, attention is drawn to a previous report centring upon the activity of Al2O3 supported rhenium in which promoted with Cs+ is related to the removal of hydrogen inhibition [20]. Indirectly, this may suggest that application of a basic support may enhance performance and this was the basis for the selection of the MgO support. As a benchmark, we have previously reported an ammonia rate of 943 +/- 44 μmol h−1 g−1 at 400 °C for bulk CoRe pretreated with N2:H2 and run in the same reactor system [19].The fourth most active catalyst under our conditions was formed from SiO2 supported Os3(CO)12. We were interested in inclusion of Os containing systems as historically Os had been identified as an active catalyst. In addition, Os can be found in the same group of the periodic table as Fe and Ru, the two elements on which different commercial ammonia synthesis catalysts are based. Indeed, the activity for ammonia synthesis is found to increase from Fe to Ru and, with Os lying below Ru, it was of interest to further compare Ru and Os despite obvious limitations such as the nature of the precursor etc. Concerns with the application of Os relate to potential toxicity arising from the formation of OsO4 and, historically, element scarcity. To the authors’ knowledge, there have not been very many studies which have reported the performance of Os based systems. One such study, reported a cyclical approach to ammonia synthesis which involved separate pulse sequences of N2 and H2 as a means to obtain high yield at reduced reaction pressure [22]. More recently, a DFT based investigation has been published in which a similar N2 activation barrier over Ru and Os nanoparticles was reported [23] with Ru being the better catalyst due to satisfying the requirements of activation energy, surface vacancy sites and number of step sites for particles of 2–4 nm diameter. In the present study, the continuous feed ammonia synthesis activity of the supported Os based system is interesting and can be directly compared to the 78.5 +/- 0.5 μmol h−1 g−1 which we have measured at 400 °C on bulk osmium powder. Whilst there could be some loss of osmium through volatilsation, decomposition of the supported cluster via an intermediate hydride might occur [24]. The decomposition of supported osmium carbonyl clusters with respect to retention of nuclearity has been controversial [25–27] especially since the Os−Os and Os−CO bond strengths are similar. Collier et al. have argued that the cluster structure may be retained in their study which employed extensively dehydroxylated supports [10]. Hence, we have compared hydroxylated and dehydroxylated SiO2 supports (see Figs. 2 and Fig 3 ) in the current study and observe a relatively small enhancement of rate on the dehydroxylated system at 500 °C with the hydroxylated SiO2 supported system apparently being more active at 400 °C. Further investigation would be required to draw firm conclusions concerning the nature of the active phases. However, this preliminary screening exercise suggests that such further studies to both elucidate structure sensitivity and potentially optimise this catalytic system might be useful avenues of further exploration. If nuclearity could be preserved by judicious choice of preparation route, the application of osmium carbonyl precursors provides a potential route to control ensemble size since a wide range of osmium carbonyl cluster sizes are documented [28]. In addition, the potential application of mixed metal osmium cluster precursors provides a potential route to systematic tuning of the activity of dispersed metal particles.The fifth material to display measureable activity at 400 °C under the reaction conditions employed has been labelled as Ni2Mo3N/SiO2, although we have not established the definite formation of the ternary nitride phase. This phase was targeted in view of the reported high catalytic activity of ternary nitrides [29–33]. Bulk Co3Mo3N, particularly when promoted with low levels of Cs+, has been widely recognised to be a very active catalyst for ammonia synthesis which has been variously ascribed to the result of a scaling relationship relating to N2 adsorption enthalpy whereby the combination of Co and Mo yields an enhanced activity material comparable to the performance of Ru [10], or a N-based Mars-van Krevelen mechanism [34,35] possibly involving an associative mechanism [36]. In this case, Ni2Mo3N was targeted rather than Co3Mo3N due to the fact that it can be more easily prepared without an ammonolysis step (ammonolysis was not employed at all in the present study due to concerns of NH3 retention on the various supports and its subsequent release complicating reaction rate analyses) just employing the 3:1 H2:N2 reaction mixture alone [37]. In addition, by application of a Pechini based route, it has been shown that Ni2Mo3N with comparable performance to Co3Mo3N can be prepared [38]. With this in mind, the activity of the supported material is not surprising although additional studies would be necessary to establish its nature. Fig. 2 presents the mass normalised rates for ammonia synthesis determined at 500 °C for various materials. It is apparent that a wider range of materials exhibit activity at this temperature than at 400 °C, although the ammonia synthesis reaction is less thermodynamically favourable with increasing temperature (the thermodynamically limited yields are 0.44% and 0.129% at 1 atmosphere pressure and 400 °C and 500 °C respectively). Once again, it can be seen that the Ru based systems are the most active. However, the extent of K+ promotion is lost with respect to the lower temperature. This may correspond to loss of K+ as the reaction temperature is increased due to enhanced mobility, but this would have to be established by elemental analysis. Cs+ is acknowledged to be a better promoter than K+ [11] but we did not explore this as in other studies on different systems we have found it to be highly mobile and easily lost from the catalytic phase at elevated reaction temperature. It can also be seen that the Os3(CO)12 derived catalysts are prominent amongst the higher activity materials (for reference the activity of bulk osmium powder was measured to be 282 μmol h−1 g−1 at 500 °C in the same reactor set up). An additional observation to be made in the present study is that there is no evidence of promotion by K+ for the Os system. As discussed earlier, it is also possible that there are some support effects amongst these materials (Fig. 3), although when the error bars in relation to activity data are taken into account, the effect seems relatively small overall. This general observation is in marked contrast to the CoRe systems. In order to facilitate comparison, they are presented in Fig. 4 where a pronounced dependence of activity upon the support identity can be seen. MgO is found to be the best support of those investigated for this system and the origin in this observation may relate to its basic nature as discussed previously. Silica is found to be reasonably good as a support whereas α-Al2O3 and particularly ZrO2 are found to be much less effective with Re/SiO2 exhibiting higher performance. The origin of these differences is not yet apparent and could relate to particle dispersion and/or mixing effects. In view of the relative performance of the CoRe systems it appears that they are worthy of further attention. To date, as for Os based catalysts, they have not been the subject of extensive investigation. As stated previously, an ammonia synthesis rate of 943 +/- 44 μmol h-1 g-1 at 400 °C has been reported for the bulk material [19]. This material is known to possess very low surface area (< 1 m2 g-1) and so on a surface area normalised basis CoRe is a comparatively highly active which warrants further investigation into supported CoRe systems. Unlike the case for the 400 °C tests, at 500 °C the supported Co systems are active. Co is fairly frequently found to be a component of active materials with CoRe being investigated in the present study, and being found to be comparatively active, and the activity of Co3Mo3N being referred to. Within the literature, LaCoSi has been reported to be an effective catalyst [39]. Indeed when a comparison was made between CoMo/CeO2, which is believed to form the supported active Co3Mo3N phase, and Co/CeO2 at 400 °C and 0.9 M P, the Co/CeO2 is initially observed to be significantly more active (4 mmol h-1 g-1 versus < 3 mmol h-1 g-1) although with time on stream the CoMo system maintains performance and the Co system significantly deactivates over the first 100 h on stream to ca, 2 mmol h-1  g-1 [40]. The application of Co/CeO2 has been discussed in terms of low-pressure ammonia synthesis and issues relating to deactivation via sintering of Co nanoparticles has been detailed in the literature very recently [41]. Incorporation of dopamine into the synthesis procedure and the associated removal of the resultant carbon layers has been reported to enhance the activity of a Co/CeO2 catalyst from 3.81 mmol h-1 g-1 to 19.12 mmol h-1 g-1 at 425 °C and 1 MPa with stability being maintained for at least 50 h on stream [41]. This enhancement has been attributed to smaller resultant Co crystallite size, enhanced metal-support interaction and lowered N2 activation energy. In the present study, the pH of the impregnating solution is 3 which is anticipated to be significantly below the point of zero charge of the supports [42]. This will result in a net surface charge and, given that the impregnating solution comprises [Co(H2O)6]2+, it can be anticipated that the Co dispersion would be poor leaving room for potential further optimisation.The final sub-set of materials to compare, relate to those comprising molybdenum. Molybdenum oxide precursors are known to nitride under 3:1 H2:N2 under the pretreatment conditions employed within this study [43]. MoO3 can be nitrided to produce the β-Mo2N0.78 phase which, as reported elsewhere is active for NH3 synthesis (a rate of 35 μmol h−1 g−1 at 400 °C and ambient pressure using a 3:1 H2:N2 reaction mixture has been reported [43]). γ-Mo2N prepared by ammonolysis of the same precursor reportedly exhibits a very similar are which is insensitive to morphology (the pseudomorphic nature of ammonolysis can be used to good effect here with MoO3 precursors of different morphology) [43], although structure sensitivity for ammonia synthesis at ambient pressure and 400 °C has been reported for Mo2N where site time yield ratios of 40:25:1 have been reported for 63, 13 and 3 nm diameter particles respectively [44]. In the present study polyoxometallates have been explored as potential precursors to highly dispersed MoNx phases for which controlled dopant levels (as achieved by the heteroatom) and size and composition could be achieved. Our initial studies have concentrated upon employing phosphomolybdic acid and silicomolybdic acid as Keggin structured precursors containing controlled levels of P and Si “dopant” respectively. As can be seen in Fig. 5 , there is very limited influence of both MoNx precursor and also dopant (in terms of the latter point, the activities are comparable to that of the Mo2N0.78/SiO2 sample, which employs ammonium heptamolybdate as precursor). The composition Mo2N0.78 has not been directly verified and is assumed based upon the anticipated binary molybdenum nitride phase which would result from the nitridation conditions employed within the current study [43]. The significant promotional effect of the inclusion of Ni, and suggested formation of the supported ternary nitride, is readily apparent in this figure as was discussed previously. In this context, it is important to establish that supported Ni is not expected to exhibit ammonia synthesis activity. In terms of benchmarking, the ammonia production rate measured with Co3Mo3N under comparable reaction conditions applying the same reactor is 489 +/- 17 μmol h−1 g−1 [45].In this manuscript, we have undertaken an empirical screening of a wide range of supported ammonia synthesis catalysts. The systems selected have been based on the known previous activity of component phases. Whilst, as might have been expected in terms of the literature, the Ru based systems we have studied were observed to exhibit the highest activity, there are a number of other potentially interesting observations. The activity of osmium based systems seems worthy of further investigation, perhaps employing carbonyl cluster precursors to control ensemble size and composition associated with preparation methods developed to retain cluster nuclearity. Supported CoRe systems, which demonstrate pronounced dependence upon the identity of the support, are also interesting candidates for further investigation, as are supported ternary nitrides.We wish to acknowledge funding in the area of ammonia synthesis from the Engineering and Physical Sciences Research Council through grant EP/L02537X/1.

The present study presents an empirical screening study of the catalytic performance of a variety of supported materials for ammonia synthesis at 400 and 500 °C. Amongst the materials tested, those derived from Ru/Al2O3 exhibited the best performance. Supported Os and CoRe catalysts also demonstrated comparatively high activities indicating them to be potentially worthy of further investigation.

Nowadays, hydrogen is used in several industrial applications. Primally it is an essential material required in a variety of chemical processes such as the oil refining, the production of ammonia and methanol, as well as the synthesis of many polymers. Additionally, hydrogen is also used in other industrial sectors, namely: glass and electronic production, metallurgy and also food industry [1,2]. Because of its properties, hydrogen is also recognized as one of the most important energy sources, that can be an alternative to the carbon-based fuels. Presently, most of the hydrogen is produced from fossil fuels, which is responsible for the emission of significant amounts of CO2 to the atmosphere [2–4]. Therefore, it is so important to develop alternative methods of hydrogen production. Among a variety of different processes, water electrolysis is considered as an environment-friendly method. If the electrical energy needed to carry out the electrolysis, originate from the renewable sources (wind or solar energy), the process meets the “zero-emission” criteria. In this case the obtained hydrogen is the cleanest energy carrier, which can be used to store the excess of the electricity. Moreover, water electrolysis allows to produce the high purity H2.Three types of electrolysis technologies are currently commercially available namely: alkaline water electrolysis (AWE), proton exchange membrane (PEM) and solid oxide electrolysis cells (SOEC) [2,4]. Since the alkaline technique was developed first, it is considered as one of the most mature methods of hydrogen production. However, the high energy consumption and the low energy conversion efficiency results in the high production costs of the alkaline water electrolysis. Because of that, the contribution of AWE and other electrolysis techniques to the total H2 production technologies, remains only at the level of about 5% [4]. Developing of new electrode materials, characterized by low overpotential for the hydrogen evolution reaction (HER) can allow to decrease the production costs and improve the cost-effectiveness of the alkaline process. That is why, currently, a lot of research concerns to investigate the catalytic activity for the HER of various materials [5–12].It is well known that Pt and the Pt-based materials are characterized by the superior catalytic activity for the HER. However, their utilization in the large-scale production is impossible because of their high costs and scarcity. In the literature one can find many reports on the electrocatalytic activity of the less noble metals e.g. Ni, Co, Fe. It was found that materials composed of the mixture of two or more metallic components possess better catalytic properties compared to the electrodes made of pure metals. Therefore, many scientific reports concern the investigation of the HER performance on the various binary and ternary alloys cathodes, as for example: Ni–Co, Ni–Fe, Ni–Mo, Co–Mo, Ni–W, Ni–Mo–Cu, Ni–Co–Cu, Co–Ni–Mo and Ni–Mo–W [13–19]. The improvement in the electrocatalytic performance of the alloy electrodes may be caused by both: the increase of the real surface area and the increase of the intrinsic activity of the alloy material. Another way to enhance the electrocatalytic activity of the non-noble metal electrodes is to introduce solid particles into the metal matrix. The literature data confirm enhanced electrocatalytic performance of different composite material e.g.Co–W/CeO2, Ni–CeO2, Ni–W/TiOx, Co/Ni–MoO2, and Co–Ni-graphene [20–29].Ni and Ni-based materials are widely used as cathodes in the alkaline water electrolysis. This is a consequence of their high corrosion resistance in the concentrated alkaline media [13,30]. Due to low hydrogen evolution overpotential [31], molybdenum is the willingly used alloy component of electrode materials for the HER.The aim of the present paper was to investigate the electrocatalytic properties for the HER of Ni–Mo/WC composite coatings. It was demonstrated that tungsten carbide is characterized by the “platinum-like” catalytic properties. This phenomenon is explained as the result of the modification of the tungsten lattice by carbon, in the way that the surface electronic properties of WC resemble those of Pt [32,33]. Therefore, materials containing WC are consider as potential alternatives to Pt for the HER processes [29,32–34]. In our research the influence of the addition of WC nanoparticles into the Ni–Mo alloy matrix on the electrocatalytic performance for the HER in alkaline media was thoroughly investigated. The Ni–Mo/WC composites were obtained by the electrochemical deposition. This method is a convenient technique that allows to prepare the nanocrystalline alloys and gives the possibility to regulate their composition by fixing a proper electrolyte composition and process parameters. The structure, composition and surface morphology of the studied coatings were investigated using XRD, SEM and EDS methods. XPS technique was used to examine the surface composition of the Ni–Mo/WC composites. The electrocatalytic activity for the HER was evaluated based on the cathodic polarization measurements and the electrochemical impedance spectroscopy (EIS). The cyclic voltammetry (CV) and chronopotentiometry (CP) technique was used to determine the long-term behaviour and the stability of the obtained composite coatings during the HER process.Ni–Mo/WC composite coatings with different Mo and WC content were electrodeposited from the solutions with the composition presented in Table 1 . The value of the electrolyte pH was adjusted to 4.5 with the solution of H2SO4. WC with the average grain size of 150–200 nm was obtained from Sigma-Aldrich. The amount of WC nanopowder that was added to the plating solution was fixed at 2.7 g dm−3. Such concentration of the nanopowder allowed to obtain composite coatings with significant WC content. Higher concentrations of WC nanoparticles resulted in their agglomeration and accumulation of the nanopowder at the vessel bottom. In order to ensure the homogenous dispersion of the solid particles in the electrolyte, the ultrasonic treatment (30 kHz, 50 W) and mechanical agitation (500 rpm) was used for 1 h before starting and during the electrodeposition process. The studied coatings were obtained in the galvanostatic regime at a current density of 4.5 A dm2. The electrodeposition was carried out at a temperature of 25 °C. The coatings were deposited on carbon steel disc electrodes with the surface area of 1.7 cm2. The Pt plate was used as an anode. During the deposition process the electrodes were placed horizontally in the vessel, parallel to each other. Before the deposition process, the cathode surface was mechanically polished with successive grades of abrasive paper and finally with a diamond paste. Then the samples were degreased with acetone, rinsed with doubly distillate water and dried. The deposition time was set at 30 min.In the investigation of the electrocatalytic performance for the HER of the Ni–Mo/WC composites, Ni–Mo alloy coating was a reference sample. The Ni–Mo deposit was obtained under similar conditions as the composite coatings, from the solution without the addition of WC nanopowder.Scanning electron microscopy (SEM) method (Quanta 3D 200i Microscope and FEI Helios G4 PFIB CXe DualBeam Microscope) was used to investigate the surface morphology of the studied composite and alloy coatings. The composition of the obtained deposits was analyzed by the Oxford Energy Dispersive X-ray Spectrometer (EDS) coupled to the scanning microscope. The element distribution map was obtained using Bruker XFlash 630 EDS Detector. Coatings cross-sections were prepared using Helios G4 PFIB CXe DualBeam Microscope. The place of investigation was protected with thin Pt layer obtained by focused ion beam induced deposition. By ion milling the material in front of the protective layer was removed and then the obtained cross-sections were polished with ion beam.X-ray diffraction technique was adopted to examine the structure of the investigated coatings. Siemens D5000 diffractometer with Cu Kα1 radiation (λ = 0.15406 nm) was applied. The measurements were performed in the 2θ range of 20–100° with a step of 0.02.X-ray photoelectron spectroscopy (XPS) studies were carried out using a SPECS PHOIBOS-100 hemispherical spectrometer equipped with a Mg source (1253.6 eV) operating at 250 W for high resolution spectra. The analyzed area of the sample was approx. 1 cm in diameter. The surfaces of catalyst samples were analyzed in the “as received” form and after gently Ar+ ion beam sputtering (1 keV, 1.3 μA/cm2). Spectra were processed and fitted by SPECLAB 2 and CasaXPS ver. 2.19 software using Gaussian-Lorentzian curve profile and Shirley baseline. The C 1s peak at 284.8 eV was used as the reference.The electrocatalytic properties for the HER of the Ni–Mo/WC and Ni–Mo coatings were investigated in 1 M KOH solution at a temperature of 25 °C. The measurements were performed with the Gamry Reference 600 potentiostat. A standard three-electrode cell configuration with Pt counter electrode and Ag/AgCl reference electrode was applied. During the electrochemical tests, the Pt counter electrode was separated from the rest of the system by the Nafion™ 117 membrane. The measured values of the potential were converted to a reversible hydrogen electrode (RHE) using the following formula: E (V vs RHE) = E (V vs Ag/AgCl) + 1.023 V.The potentiodynamic polarization test were performed at a scan rate of 1 mV s−1. The curves were collected in the potential range of −0.58 V vs RHE up to the open circuit potential. The obtained data were corrected on the IR ohmic drop. The solution resistance values were determined based on the EIS results. Before the experiment, each sample was held for 10 min in 1 M KOH solution at a potential of −0.58 V vs RHE. This was to reduce the metal oxides that might formed on the metal surface after the deposition process.EIS measurements were conducted in the frequency range of 10 mHz to 10 kHz. A sinusoidal signal of 5 mV amplitude was used. The experiments were performed at a selected cathodic overpotential. Before the measurement, each sample was held for 10 min in 1 M KOH solution at a potential of −0.58 V vs RHE. The obtained data were analyzed with Gamry Echem Analyst software.CV curves were collected in a potential range of 0.12 to −0.58 V vs RHE. The measurements were performed for 100 cycles, at a scan rate of 50 mV s−1.Chronopotentiometry measurements were conducted at a constant current density value of 100 mA cm−2. The CP curves were recorded for 60 h. Fig. 1 presents the SEM images of the studied Ni–Mo/WC and Ni–Mo coatings. The surface of Ni–Mo alloy coating obtained from the solution without an addition of WC nanoparticles, is compact and homogenous. The coating is characterized by globular and cauliflower-like structure (Fig. 1a). According to EDS analysis, the alloy deposit contained about 22,8 wt% of Mo (Table 2 ). The Ni-Mo-WC composite coatings obtained from solution containing 2.7 g dm−3 of WC nanopowder, also show globular structure, however they are less compact and regular in comparison to the alloy deposit. The surface of the composite coatings with higher WC content (Fig. 1b and c) is rough and highly developed. Ni–Mo/WC 3 deposit obtained from the solution with the highest molybdate concentration is characterized by smother surface, however globular grains are also visible (Fig. 1d). Fig. 2 presents the distribution of the individual elements (Ni, Mo and W) on the surface of Ni–Mo/WC 2 and Ni-Mo-WC 3 composites. For both coatings the Ni and Mo atoms are rather uniformly distributed on the composite surface. However, it can be observed, especially on the surface of Ni–Mo/WC 2 coating with higher WC content, that the intense signal originating from W atoms is followed by more intense signal of Mo and less intense signal of Ni. This phenomenon is probably the result of the adsorption of molybdate ions on the surface of the WC nanopoarticles in the plating solution.The obtained EDS results reveal that the concentration of MoO4 2− ions in the plating solution affects not only the amount of codeposited Mo, but also the amounts of WC nanoparticles incorporated into the coating. As one might expect, an increase in the amount of the molybdate ions in the solution led to a rise in Mo content in the composite coating from 7.1 wt% for MoO4 2− concentration of 0.009 mol dm−3 to 27.7 wt% for MoO4 2− concentration of 0.025 mol dm−3 (Table 2). On the other hand, higher amounts of molybdate ions in the plating solution resulted in lower WC content in the composite coating. According to EDS analysis, the W content amounted to 21.2 and 8.7 wt% for the coating deposited from the solution with the lowest (0.009 mol dm−3) and highest (0.025 mol dm−3) MoO4 2− concentration, respectively (Table 2). A negative influence of high molybdate concentration on the incorporation of nanoparticles into the metallic matrix, was observed also in our earlier research [35]. This phenomenon may be caused by a decrease in the surface positive charge of WC nanoparticles due to the adsorption of MoO4 2− ions on their surface.The Mo/Ni ratio calculated for the Ni–Mo/WC 3 composite coating and Ni–Mo alloy deposit obtained from solution with the same molybdate concentration (0.025 mol dm−3) was 0.38 and 0.29 respectively. This indicates, that the Mo codeposition was facilitated by the presence of WC nanoparticles in the electrolyte.The local cross-sections of the studied coatings were presented in Fig. 3 . Despite the same deposition time (30 min) a significant differences in the coatings thickness was observed. These changes are related to the Mo content in the Ni–Mo matrix. An increase in the amount of codeposited Mo resulted in the smaller coating thickness due to lower current efficiency of the deposition process. A similar effect was observed in our earlier research of Ni–Mo–ZrO2 composite coatings [35]. The cross-section obtained for Ni–Mo alloy containing ~22.8 wt% Mo (Fig. 3a) revealed a compact and homogenous structure. The coating thickness was about 7 μm. The thickness of the composites with lower Mo content was grater - about 17 μm for Ni–Mo/WC 1 (~7.1 wt% Mo) and 13 μm for Ni–Mo/WC 2 (15.4 wt% Mo). Moreover, the cross-sections of Ni–Mo/WC 1 and Ni–Mo/WC 2 composites containing high amounts of incorporated WC nanoparticles, demonstrate less compact structure with empty gaps occurring in the coatings volume. Fig. 4 shows the distribution map of the individual elements in the cross-section of Ni–Mo/WC 2 composite.It can be seen, that WC nanoparticles are distributed in the whole coating volume, however they also tend to accumulate into bigger aggregates. The cross-section obtained for Ni–Mo/WC 3 composite revealed a compact, however, significantly thinner coating - thickness of ~1 μm. The low thickness is related to high Mo content (~27.7 wt %) in the Ni–Mo matrix. According to the literature, the first stage of Mo deposition is an incomplete reduction of molybdate ions to Mo oxides of lower valence, which are characterized by low overpotential for the hydrogen evolution reaction [36]. Therefore, their presence on the cathode surface results in the intensification of the hydrogen evolution during the deposition process and causes the decrease of the current efficiency. Moreover, intensification of the hydrogen evolution process leads to formation of higher amounts of gas bubbles on the cathode surface, which hinder the adsorption and reduction of metal ions.XRD patterns of the studied coatings are presented in Fig. 5 . In the pattern obtained for the Ni–Mo alloy coating one broad peak at the 2θ angle of about 42° is visible. The peak corresponds to the (111) plane of nickel FCC structure (ICDD no. 01-071-4654). However, its maximum is shifted to the lower 2θ value in comparison to the pure nickel pattern. This phenomenon is characteristic for the formation of Ni–Mo solid solution [37]. The small peak at the 2θ of about 82.4° originating from the Fe phase of the steel substrate is also visible in the Ni–Mo pattern. In the diffractograms of Ni–Mo/WC composite coatings, the peaks corresponding to Ni–Mo solid solution are broadened and their intensity decreased. This indicates smaller size of crystallites of composite coatings. A decrease in the crystallite size due to the incorporation of solid particle into the metal matrix was observed also by other researchers. This phenomenon was explained as the result of inhibition of the crystals growth and enhancing of crystals nucleation by solid particles adsorbed on the metal matrix during the deposition process [38,39]. For the studied Ni–Mo/WC composites the crystalline size is determined by two factors, namely the amount of incorporated WC nanoparticles and Mo content in the Ni–Mo matrix. It has been demonstrated in the literature, that an increase in the Mo content results in the crystallite size refinement of Ni–Mo alloy [37,40]. In the XRD pattern of Ni–Mo/WC 3 coating, low intensity of the peak corresponding to Ni–Mo solid solution and high intensity of the peak originating from the Fe phase of the steel substrate (at the 2θ of about 44.7 and 82.4°), is also the consequence of low deposit thickness (~1 μm). In the diffractograms recorded for Ni–Mo/WC composites additional peaks corresponding to WC phase (ICDD no. 00-025-1047) also appeared.XPS method was used to analyse the surface composition of the studied composite and alloy coatings. Table 3 presents contents of the individual elements on the as-received surface (without Ar+ ions sputtering) of the Ni–Mo/WC and Ni–Mo coatings. The as-received surface of all coatings was covered with the layer of carbon adsorbed from the air. This contamination carbon appeared in the most outer surface layer with the thickness less than 1 nm. As it can be seen in Table 3, the oxygen content is significantly higher for the composite coatings than for the Ni–Mo alloy. This indicates, that the Ni–Mo/WC coatings were characterized by more oxidized surface in comparison to Ni–Mo deposit. The oxygen content in the surface layer increased with the rise of molybdate concentration in the plating solution. The surface of Ni–Mo/WC 3 composite contained the highest oxygen content (38.1 at. %). For Ni–Mo alloy coating the Mo/Ni ratios calculated based on XPS and EDS results are the same (around 0.3). In case of composite coatings Mo/Ni ratios calculated based on XPS results are higher than the ratios resulting from the EDS analysis. This shows that the surface of Ni–Mo/WC composites was enriched in Mo in comparison to the coating bulk.The percentage shares of individual forms of the elements in the studied coatings was evaluated based on the deconvolution of Ni 2p, Mo 3d and W 4f spectra. Fig. 6 presents the Ni 2p3/2/spectra obtained for the Ni–Mo/WC and Ni–Mo coatings. The spectrum of metallic nickel can be fitted using three peaks: the main asymmetric peak at 852.6 eV and two broader satellites peaks at about 3.7 and 6.1 eV above the main contribution [41–43]. For a better fit, in our work, the surface of Ni–Mo coating after Ar+ sputtering was used as a reference spectrum of metallic Ni. In this case the maximum of the main peak was located at 852.3 eV and the maxima of the satellite peaks were shifted by 3.7 and 6.5 eV above (Fig. 6a). Based on our earlier research [35,44], we expected that besides metallic Ni also Ni(OH)2 would appear on the surface of the studied coatings. Therefore the part of the spectrum above the binding energy of 854 eV was analyzed assuming the presence of Ni(OH)2. A two-component deconvolution procedure was used to fit the Ni 2p3/2 core level spectra with Ni(OH)2 envelope. The major peak attributed to Ni(OH)2 component was found at the binding energy of 859.9 eV and a satellite peak appeared 5.8 eV above. This results agreed well with the data presented in the literature [45]. The obtained data revealed that the surface layer of the as-deposited Ni–Mo alloy coating contained mainly metallic Ni (90.8 at.% of the total Ni) (Table 4 ). Different situation was observed for Ni–Mo/WC composites. In case of Ni–Mo/WC 2 and Ni–Mo/WC 3 deposit, the oxidized Ni was the main component of the coatings surface layer. An increase in the molybdate concentration in the plating solution resulted in higher Ni(OH)2 contents on the composite surface. For the as-deposited Ni-Mo-WC 3 coating 90.4% of the total Ni were oxidized in the surface layer.Fitted high-resolution spectra recorded for Mo 3d regions were presented together with Shirley background line in Fig. 7 . The spectra were deconvoluted assuming the presence of the following components: Mo(0), MoO2, Mo2O5 and MoO3. Deconvolution was performed using the procedure presented by Baltrusaitis et al. [46]. And applied already in our previous research [35]. The binding energies of 227.45, 229.30, 230.60 and 232.2 eV were ascribed to the Mo(0), Mo(IV), Mo(V) and Mo(VI), respectively. Similarly to Ni(0), for a better fit, the surface of Ni–Mo coating etched with Ar+ ions was used as a reference spectrum of metallic Mo (Fig. 7a). The XPS data obtained based on Mo 3 d spectra revealed similar relationship as for those observed for nickel. On the as-deposited surface of Ni–Mo alloy coating the share of metallic Mo amounted to 73.5%. The content of metallic Mo on the as-deposited surface of the composite coatings was lower and it decreased with an increase in the molybdate concentration in the plating solution – from 53.1% for the Ni–Mo/WC 1–5.4% for the Ni–Mo/WC 3. Mo2O5 was the main oxidized form of Mo in the surface layer of the studied coatings. The exception was the Ni–Mo/WC 1 composite, for which MoO3 was the main form of the oxidized Mo. Fig. 8 a shows the W 4f spectra obtained for the WC nanopowder. Since, according to the literature data, WC can oxidize in the air [47] the spectrum was deconvoluted into three components: WC (32.0 and 34.0 eV), WO2 (33.0 and 35.0 eV) and WO3 (35.91 and 38.11 eV). The spectra recorded for the Ni–Mo/WC coatings were presented in Fig. 8b – d. It can be seen from Fig. 8 that for the composite coatings, the maximum of the highest peak appears at lower values of binding energy than in case of WC nanopowder. In Fig. 8e a reference W 4f spectrum fitted with consideration of metallic W, WC and W oxide was presented [48]. Comparing the reference spectrum with the spectra recorded for the studied composites, it might be concluded that metallic W was present on the surface of Ni–Mo/WC coatings. This indicates that during the deposition process, some of W atoms was reduced to metallic tungsten. This assumption seems to be reasonable, since during the deposition process reducing conditions occur at the cathode surface. Tran et al. [49], in their research also observed the reduction of W atoms to metallic tungsten in Mo–W bimetallic carbide. The highest percentage share of metallic W was calculated for the Ni–Mo/WC 1 composite (93.5 at.%) (Table 4). Ni–Mo/WC coatings obtained from the solutions with higher molybdate concentration were characterized by lower contents of W0 in the surface layer (76.9 and 74.1 at. % for the Ni–Mo/WC 2 and Ni–Mo/WC 3 deposit, respectively). Opposite relationship was observed for the oxidized W. Similarly as for Ni and Mo, the share of the Wox rises with an increase in the molybdate concentration in the plating solution.The obtained XPS results indicates that the incorporation of WC nanoparticles significantly influences the surface of the Ni–Mo alloy coating. The surface of Ni–Mo/WC composites is more oxidized in comparison to Ni–Mo deposit. This effect becomes particularly apparent for the composites with high Mo content in the Ni–Mo matrix.It was proposed in the literature [50–53] that the hydrogen evolution process in alkaline media proceed through the following reactions: (1) M + H2O + e− → MHads + OH− (2) MHads + H2O + e− → H2 + M + OH− (3) 2MHads → 2M + H2 In the presented mechanism the proton discharge electrosorption - Volmer reaction (1) is followed by electrodesorption - Heyrovsky reaction (2) and/or H recombination - Tafel reaction (3). The value of Tafel slope (bc) is determined by the rate-limiting reaction, so it can provide some insight into the mechanism of the hydrogen evolution process. According to the general model of the HER mechanism, the Volmer reaction control the process speed if bc is around 120 mV dec−1. The Heyrovsky or Tafel step determines the HER rate if Tafel slope is 30 or 40 mV dec−1, respectively.The potentiodynamic curves recorded for the studied Ni–Mo/WC and Ni–Mo coatings are shown in Fig. 9 . The kinetics parameters determined from the obtained data are presented in Table 5 . In case of the composite with the lower Mo content (Ni–Mo/WC 1) and the Ni–Mo alloy, two regions with different slopes can be distinguished on the obtained curves. In both cases, in the lower potential range (region 1 in Fig. 9) the higher values of bc were calculated. This phenomenon is most probably caused by the oxide layer on the coatings surface. Other authors also observed the influence of the surface oxide layer on the hydrogen evolution process on metal electrodes [54–56]. XPS results confirmed the presence of oxides on the surface of all studied coatings. However, Ni–Mo alloy and Ni–Mo/WC 1 composite were characterized by lower thickness of the oxide film (lower oxide content – Table 3) in comparison to other Ni–Mo/WC coatings. It is possible that thinner oxide layer, especially in case of Ni–Mo alloy was also more compact and could inhibit the electron transfer in the low potential range. This resulted in very high value (590 mV dec−1) of Tafel slope determined for Ni–Mo alloy for the region 1 of the polarization curve. For the region 2 the bc value was 130 mV dec−1, which suggest that Volmer reaction determined the hydrogen evolution rate on the Ni–Mo alloy in the higher potential range. In case of Ni–Mo/WC 1 coating, the determined value of Tafel slope was 208 and 104 mV dec−1 for the region 1 and 2 (Fig. 9), respectively. The lower value of bc in the low potential range in comparison to Ni–Mo alloy may indicate the less compact oxide layer on the surface of the composite coating. The value of Tafel slope obtained for the region 2 indicates that, similarly to the alloy coating, the Volmer reaction determined the rate of the hydrogen evolution on the Ni–Mo/WC 1 composite in the higher potential range. Different behaviour was observed for the composites with higher Mo content. The polarization curves recorded for Ni–Mo/WC 2 and Ni–Mo/WC 3 coatings were characterized by one value of Tafel slope in the whole potential range. According to XPS data the surface oxide layer of these composites was characterized by greater thickness and different composition in comparison to Ni–Mo and Ni–Mo/WC 1 coatings. In this case, the region with significant limitations of the electron transfer for low potentials was not observed. The calculated Tafel slopes are 153 mV dec−1 for Ni–Mo/WC 2 and 163 mV dec−1 for Ni–Mo/WC 3. These values are greater than the expected value of ~120 mV dec−1. Similar results were also observed for other composite materials and explained by the presence of the outer oxide film, which could affect the charge transfer on the electrode surface [22,57,58]. Considering the XPS results, which confirmed the high oxygen content on the surface of Ni–Mo/WC 2 and Ni–Mo/WC 3 coatings, this explanation seems reasonable. The results presented in Fig. 9 reveal different catalytic activity for the hydrogen evolution reaction of the studied composite and alloy coatings. Additionally the obtained results indicate that the presence of the surface oxide layer (with different thickness and composition for different coatings) significantly affects the electron transfer process on the electrode surface/solution interface influencing at the same time the course of the recorded potentiodynamic curves.The exchange current density (i0) is the parameter that provides the information about the catalytic activity of the electrode material. It can be seen from Table 5 that the Ni–Mo/WC composite coatings are characterized by significantly higher values of i0 in comparison to Ni–Mo alloy deposit. This confirms the beneficial effect of the WC particles on the catalytic activity of Ni–Mo alloy deposit. Moreover, it can be observed that the determined i0 value increased with the rise in the Mo content in the composite coating. The highest value of the exchange current density was calculated for Ni–Mo/WC 3 composite indicating its superior catalytic activity to the other studied coatings. In order to hydrogen evolution proceed with a measurable rate a certain overpotential is required. Therefore, a good method of comparing the electrocatalytic activity of different materials is to determine the overpotential needed to achieve a fixed value of current density (i.e. a hydrogen production rate). This gives an information about the amount of the energy required to produce a fixed amount of hydrogen for each catalysts. In Table 5, for each studied coating, the overpotential needed to achieve the current density of 10 mA cm2 (η10) was presented. The η10 values reveal the similar trend in the electrocatalytic activity as the trend observed based on the j0 values. The required overpotentials were lower for Ni–Mo/WC composites than for Ni–Mo alloy coating. Ni–Mo/WC 3 electrode was characterized by the lowest value of η10, which confirms its highest catalytic activity for the HER among the studied coatings.In order to further investigate the hydrogen evolution process on the studied coatings and evaluate their electrocatalytic activity, the electrochemical impedance technique was applied. For each coating the EIS spectra were recorded at three different values of cathodic overpotential, which corresponded to different hydrogen evolution rates (current densities range from ~5 to ~ 20 mA cm−2). In order to receive a physical picture of the processes occurring at the electrode/solution interference the obtained EDS must be fitted with a proper model. The literature data reveal that three different equivalent circuits have been mostly proposed to model the processes that occur during the HER [54,55,59–64]. The one time constant model (1 TC) presented in Fig. 10 a describes the hydrogen evolution process when the response of hydrogen adsorption is not manifested on the spectra. In this model, R s is the solution resistance, R 1 is the charge transfer resistance and CPE 1 is the parameter associated with the double layer capacitance. If the system is characterized by two time constants, the two time constant parallel model (2TCP) or the two time constant serial model (2TCS) model have been used to describe the HER. The 2TCP model (Fig. 10b) reflects the response of a system in which both time constants are related to the HER kinetics. It is assumed that the high frequency (HF) time constant τ1 (CPE1 – R1) corresponds to the HER charge transfer kinetics, while the low frequency (LF) time constant τ2 (CPE2 – R2) is related to the hydrogen adsorption. Both time constants depend on the applied overpotential. In the 2TCS model (Fig. 10c) only the LF time constant is related to the HER kinetics and it change with overpotential. The HF time constant corresponds to the surface porosity and it is potential independent.In our research, due to ensure the more accurate fit of the collected data, in all adopted models the constant phase element (CPE) was used instead of the capacitor. The use of the CPE is often required because of the distribution of the relaxation times as a result of the surface inhomogeneities such as the surface roughness and porosity or when the adsorption or diffusion takes place [45]. The impedance of the CPE is defined as: (4) Z CPE = Y 0 −1 (jω) -n where Y 0 is a time constant parameter (Ω−1 s-n cm−2), ω is the angular frequency of the AC signal and n is the CPE exponent. Fig. 11 presents the EIS spectra collected for the studied coatings. The one time constant respond was recorded only for Ni–Mo/WC 1 composite for the data collected at the overpotential of −336 and −386 mV (Fig. 11a). In this case the data were fitted with the model presented in Fig. 10a. The other recorded spectra were characterized by the presence of two time constants. Since applying the 2TCP and the 2TCS models resulted in almost identical goodness of fit expressed by the Chi-squared value (χ2), deciding which model is correct required a careful analysis of the obtained data.The experimental data recorded for Ni–W/WC 1 composite at the lowest overpotential (−286 mV) were fitted with the 2TCP model. The same model was also apply for the all EIS data collected for Ni–Mo/WC 2 coating (Fig. 11b). The resulting EIS parameters are presented in Table 6 . It can be seen, that in case of data fitted with the 2TCP model, the value of Y 01 parameter is relatively constant and the values of R 1 decrease with an increase in the overpotential. This is a characteristic behaviour of the time constant related to the HER charge transfer kinetics. Moreover, the Y 02 values are higher than Y 01 and they increase with overpotential. At the same time, the values of R 2 decrease with an increase in the overpotential. This observations confirms that the low frequency time constant (CPE 2 – R 2 ) is related to the response of the hydrogen adsorbed on the electrode surface [54,55].Different behaviour was observed for the Ni–Mo/WC 3 composite. It can be seen from Fig. 11c that the diameter of the high frequency semicircle does not change with overpotential. This suggests that the HF time constant is related to the surface porosity and the LF time constant corresponds to the charge transfer kinetics of the HER. Similar tendency was also observed for Ni–Mo alloy coating (Fig. 11d). Therefore the EIS data collected for Ni–Mo/WC 3 composite and Ni–Mo alloy were fitted with the 2TCS model. Similar behaviour of Ni–Mo/WC 3 and Ni–Mo electrodes might be the result of the relatively high Mo content in both coatings (~28 and ~23 wt% for Ni–Mo/WC 3 and Ni–Mo, respectively), which had influenced the structure of the Ni–Mo alloy matrix. Additionally, in case of Ni–Mo alloy, in the spectrum recorded at the highest overpotential value (−625 mV), a characteristic inductive loop is visible in the low frequency region (Fig. 11d). This phenomenon was also observed by other researchers and explained as the result of adsorption of large amounts of hydrogen on the cathode surface [21,59,62,65]. In this case the experimental data were fitted with the equivalent circuit presented in Fig. 10d (2TCSI), containing two additional elements related to the adsorption process – inductance L and resistance R L . The resulting EIS parameters calculated for Ni–Mo/WC 3 composite and Ni–Mo alloy are also presented in Table 6. The EIS spectra of each studied coating were collected at different cathodic overpotentials, so the obtained values of resistance cannot be directly compared. However the obtained results allow to conclude, that an increase in the Mo content in composite coating resulted in a decrease in the value of the charge transfer resistance. Even though the EIS spectra of Ni–Mo/WC 3 coating were recorded at lower cathodic overpotentials in comparison to other studied coatings, this composite was characterized by the lowest values of the charge transfer resistance. This proves its highest catalytic activity among the studied materials. These results are consistent with the trend observed based on the cathodic polarization measurements. The relatively low values of charge transfer resistance of Ni–Mo alloy coating were the results of high cathodic overpotentials for which the EIS spectra were collected.The obtained EIS parameters that corresponds to the HER charge transfer kinetics, gives the possibility to evaluate the catalyst real surface area, which is the true electrochemically accessible area for the hydrogen to adsorb. The real surface area can be estimated from the value of double layer capacitance (C dl ). Assuming that the average double layer capacitance of a smooth metal surface is 20 μF [54,55,57,59,62], the real surface area can be calculated as A real = C dl /20 cm2. The double layer capacitance can be determined by the following relation [66]: (5) C i = [ Y 0 i / ( R s − 1 + R i − 1 ) ( 1 − n i ) ] 1 / n i The real electrochemically active area of the material allow to calculate the roughness factor (R f ) of its surface. This parameter characterize the ratio of the real surface area to the geometric surface are, so it can be calculated as R f = A real /A geometric . Knowing the surface roughness factor it is possible to evaluate the intrinsic catalytic activity (defined as i/R f ) of the catalyst. The calculated values of double layer capacitance and R f parameter for the studied coatings are presented in Table 7 . The obtained results reveal that Ni–Mo/WC composite coatings are characterized by higher values of the surface roughness factor in comparison to Ni–Mo alloy. The highest value of R f was obtained for Ni–Mo/WC 3 composite. This was the result of the presence of WC nanoparticles as well as the high Mo content in the Ni–Mo matrix. Table 7 also illustrates the comparison of the intrinsic catalytic activity of the studied coatings. The ratios i/R f were calculated for the current densities determined from the polarization curves at a set overpotential of 150 mV. This is a value of the overpotential that can be expected during the hydrogen production in the electrolyser. The highest i/R f ratio was also calculated for Ni–Mo/WC 3 coating. This indicates that the superior catalytic activity of the Ni–Mo/WC 3 electrode is the result of both high real surface area and high intrinsic catalytic activity of the composite coating. The other studied coatings were characterized by slightly lower i/R f ratios than Ni–Mo alloy. The presented considerations indicate that both factors, namely, the presence of WC nanoparticles as well as Mo content in Ni–Mo alloy matrix affected the catalytic activity of the studied composite materials.In order to evaluate the long term behaviour and stability of the studied materials during the HER, two electrochemical techniques have been applied, namely, cyclic voltammetry (CV) and chronopotentiometry.The CV measurements allowed to investigate the electrocatalytic behaviour of the studied coatings under the influence of the cyclically changing potential in the range of 0.12 to −0.58 V vs RHE. The results of CV investigations were presented in Fig. 12 and Table 8 . For the Ni–Mo/WC 1 and Ni–Mo/WC 2 composites the electrochemical behaviour was similar for the initial and final cycle (Fig. 12a and b). In case of the Ni–Mo/WC 1 coating the value of current density recorded at the potential of −0.58 V for the 100th cycle slightly decreased in comparison to the 1st cycle (Table 8). For Ni–Mo/WC 2 the measured current densities were similar for the 1st and 100th cycle. Higher values of current densities recorded for Ni–Mo/WC 2 coating in comparison to Ni–Mo/WC 1 deposit confirmed better electrocatalytic properties for the HER of composite with higher Mo content. Different electrochemical behaviour was observed for Ni–Mo/WC 3 composite and Ni–Mo alloy. In this case the electrochemical performance changed clearly with the increase in the number of cycles (Fig. 12c and d).For Ni-Mo-WC 3 composite the current density recorded at −0.58 V vs. RHE increased from 74.6 mA cm−2 for the 1st cycle to 89.5 mA cm−2 for the 100th cycle. This indicates that the changes occurring on the coating surface as a result of the subsequent CV cycles positively affected the electrocatalytic properties for the HER of the Ni-Mo-WC 3 composite. Opposite relationship was observed for the Ni–Mo alloy coating. In this case the recorded current densities decreased significantly when increasing the number of cycles (from 52.2 mA cm−2 for the 1st cycle to 22.8 mA cm−2 for the 100th cycle), which proves the deterioration of the catalytic activity.In order to understand the observed behaviour, the surface of Ni–Mo/WC 3 and Ni–Mo coating after 100 cycles of CV measurement was analyzed using the XPS method. The recorded Ni 2p3/2, Mo 3d and W 4f spectra (Fig. 13 a–c) were deconvoluted according to the procedures described in the previous chapter. Based on the obtained results the percentage shares of the individual forms of the elements were calculated. In Table 9 the surface composition of the Ni–Mo/WC 3 and Ni–Mo coatings in the as-deposited form and after 100 cycles of CV measurements was compared. In both cases, a decrease in the content of metallic Ni and a rise in the amount of Ni(OH)2 was observed on the coating surface after the 100 cycles of CV. According to the literature data the formation of α-Ni(OH)2 may occur at the potential of about −0.9, −0.8 V vs SCE (0.1, 0.2 V vs RHE) [67,68]. Since our experiment was performed at the potential range of 0.12 to −0.58 V vs RHE, Ni oxidation to Ni(OH)2 could appear during the subsequent CV cycles. The obtained XPS results indicate the irreversible nature of this process. At the same time, an increase in the content of metallic Mo on the surface of both kinds of coatings after 100th cycles was observed. In case of Ni–Mo alloy coating, the Mo0 content amounted to 98.8% of the total Mo. The hole oxidized Mo was in the form of MoO2.The observed changes on the surface of Ni–Mo alloy coating led to the deterioration of the catalytic activity for the HER. For Ni-Mo-WC 3 composite the content of Mo0 in the surface layer increased from 5.4% for the as-deposited coating to 12.7% for the deposit after 100 of CV cycles. Additionally, it was observer that Mo2O5 was the predominant oxide on the surface of as-deposited coating (51% of the total Mo), while on the surface of the coating after CV measurements 66.6% of the total Mo was in the form of MoO3. Moreover the subsequent CV cycles resulted in a complete reduction of the oxidized W. The literature data reveal that MoO3 possess interesting catalytic properties and has been widely used as a catalytic material for the hydrogen evolution reaction [69–75]. Therefore the observed improvement of the catalytic activity of Ni-Mo-WC 3 composite coating after subsequent CV cycles may be the result of an increase in the MoO3 content in the electrode surface layer.The long term catalytic activity and stability during the HER was evaluated based on the chronopotentiometry method. The measurement was performed for Ni–Mo/WC 3 composite and for comparison for Ni–Mo alloy electrode. CP curves were recorded for 60 h at a set current density of 100 mA cm−2. The obtained results are presented in Fig. 14 . For both tested coatings the measured potential, after an initial slight increase, remained at a constant level throughout the entire experiment. The potential required to provide the fixed current density was significantly lower for the Ni–Mo/WC 3 electrode (~-0.66 V vs RHE) than for Ni–Mo alloy(~-0.91 V vs RHE). These results confirm the improvement of catalytic activity due to WC nanoparticles incorporation into Ni–Mo alloy matrix.The obtained results indicate that both factors, namely, the presence of WC nanoparticles as well as Mo content in Ni–Mo alloy matrix affected the catalytic activity of the studied composite materials. Both parameters strongly influenced the surface composition and surface roughness factor of the Ni–Mo/WC catalysts. The composites containing higher amounts of WC nanoparticles and at the same time with lower Mo content (Ni–Mo/WC 1 and Ni–Mo/WC 2) were characterized by slightly lower values of the intrinsic catalytic activity in comparison to Ni–Mo alloy. This indicates that their improved catalytic activity is the consequence of an increase in the surface roughness factor due to incorporation of WC nanoparticles. The XRD results revealed more nanocrystalline structure of the composite coatings in comparison to Ni–Mo deposit. An increase in Mo content in the Ni–Mo alloy matrix also results in the crystallite size refinement, therefore both parameters – WC incorporation and codeposition of Mo affected the structure and the value of roughness factor of the studied composites. The highest value of R p parameter was calculated for Ni–Mo/WC 3 catalyst. This was the consequence of the WC incorporation (WC content was lower than in other composites, but still relatively high - 8.7 wt% of W) and the highest content of Mo in the Ni–Mo matrix among the studied coatings. On the other hand the presence of WC nanoparticles resulted in more oxidized surface of composite coatings in comparison to Ni–Mo alloy. The thickness of the oxide layer on the composite surface was also enhanced by higher Mo content in Ni–Mo matrix. Ni–Mo/WC 3 catalyst was characterized by the most oxidized surface, which was the result of both the presence of WC nanoparticles and the highest Mo content in the Ni–Mo matrix. High content of oxides on the surface of Ni–Mo/WC 3 composite resulted in its higher intrinsic catalytic activity in comparison to other studied coatings. The CV and XPS results suggest that this effect was the consequence of high concentration of MoO3 on the composite surface.The catalytic activity for the HER of Ni–Mo/WC composites with different Mo and WC content and, for comparison, of Ni–Mo alloy electrode was investigated in 1 M KOH solution. The SEM analysis revealed that the incorporation of WC nanoparticle into Ni–Mo alloy matrix resulted in less compact and regular surface of the composite coatings. The amount of incorporated nanoparticles was affected by the molybdate concentration in the plating solution. All studied coatings were composed of nickel FCC phase. However, lower intensity and broadening of the peak corresponding to Ni–Mo matrix in the patterns of Ni–Mo/WC coatings, indicates smaller crystallite size of the composites in comparison to Ni–Mo alloy. The XPS results revealed that the surface of Ni–Mo/WC composite electrodes was more oxidized than the surface of Ni–Mo alloy. This indicates, that the thickness of the outer oxide layer was greater for the composite coatings. Moreover it was discovered, that the oxide layer thickness increased with the rise in the Mo content in Ni–Mo metal matrix.The cathodic polarization measurements revealed that, in case of the all studied electrodes, the hydrogen evolution rate was determined by the Volmer reaction. Moreover, the process was affected by the outer oxide layer presented on the coatings surface. The performed electrochemical tests proved that the incorporation of the WC nanoparticles into the Ni–Mo matrix resulted in the improvement of the electrocatalytic properties for the HER. The catalytical activity increased with an increase in the Mo content in the alloy matrix of the Ni–Mo/WC composites. The highest catalytic activity was identified for the Ni–Mo/WC 3 composite, characterized by the highest Mo content and the most oxidized surface among the studied coatings. According to EIS results this improvement was the consequence of both high real surface area and high intrinsic catalytic activity of Ni–Mo/WC 3 electrode.It was discovered that the catalytic activity of Ni–Mo/WC 3 composite increased after subsequent CV cycles. XPS analysis revealed that this was a consequence of an rise in the MoO3 content in the oxide layer on the electrode surface.The obtained results allow to conclude that the catalytic activity of Ni–Mo/WC composites was determined by two factors, namely, the presence of WC nanoparticles and the Mo content in the Ni–Mo metallic matrix. The presence of WC nanoparticles and high content of Mo in Ni–Mo/WC 3 composite resulted in highly oxidized surface and led to higher catalytic activity for the HER.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 co-financed by statutory activity subsidy from the Polish Ministry of Education and Science for the Faculty of Production Engineering of Wroclaw University of Economics and Business (Department of Inorganic Chemistry; No 501-110-0310005000) and for the Faculty of Chemistry of Wrocław University of Science and Technology (Department of Advanced Material Technologies (K26W03D05); grant number 8211104160).

The catalytical activity for the hydrogen evolution reaction (HER) of the electrodeposited Ni–Mo/WC composites is examined in 1 M KOH solution. The structure, surface morphology and surface composition is investigated using the scanning electron microscopy, X-ray diffraction and X-ray photoelectron spectroscopy. The electrocatalytic properties for the HER is evaluated based on the cathodic polarization, electrochemical impedance, cyclic voltammetry and chronopotentiometry methods. The obtained results prove the superior catalytic activity for the HER of Ni–Mo/WC composites to Ni–Mo alloy. The catalytic activity of Ni–Mo/WC electrodes is determined by the presence of WC nanoparticles and Mo content in the metallic matrix. The best electrocatalytic properties are identified for Ni–Mo/WC composite with the highest Mo content and the most oxidized surface among the studied coatings. The impedance results reveal that the observed improvement in the catalytic activity is the consequence of high real surface area and high intrinsic catalytic activity of the composite.

The development of industrial fields has been progressing since the industrial revolution in the eighteenth century. The discharged wastewaters from industrial activities contain many organic and inorganic pollutants that are highly toxic and persistent [1,2]. Phenolic compounds are one of the major organic water pollutants produced from different industries, including petrochemicals, polymers, dyes, and pharmaceuticals. Particularly, phenol is a persistent organic pollutant with toxic and carcinogenic properties [3]. The levels of phenol are restricted by the environmental protection agency (EPA) to less than 0.1 ppm in wastewater discharges. Consequently, significant attention has been paid to removing phenol from industrial wastewater, and many methods have been developed in this regard [4].Advanced Oxidation Processes (AOPs) are widely used for the effective removal of organic pollutants in wastewater. AOPs have the potential to degrade various types of organic pollutants in wastewater. AOPs are based on the generation of active free radicals, such as hydroxyl radicals (HO), that degrade organic pollutants to nontoxic products [5]. The use of heterogeneous catalysts in AOPs, with the involvement of a strong oxidizing agent such as hydrogen peroxide (H2O2), is commonly studied by researchers to improve the degradation of organic pollutants [6].Spinel ferrites MFe2O4 (M = Co, Ni, and Cu, etc.) are interesting materials with unique physical and chemical properties. They are widely used in various applications, including biomedical, sensors, supercapacitors, microwave absorption, and catalysis [7]. Among various ferrites, cobalt ferrite (CoFe2O4) has received considerable attention for its excellent magnetic properties, narrow optical bandgap, good chemical stability, ease of preparation, nontoxicity, and good catalytic performance [8]. Cobalt ferrites have been used as catalysts for various reactions [9].The present study investigated the application of cobalt-based ferrite nanoparticles as catalysts for phenol degradation at neutral pH. Various cobalt-based ferrites were prepared, namely: pure ferrites (CoFe2O4), mixed ferrites (CoxNi1-xFe2O4), and P-modified ferrites (P-CoFe2O4). In the preparation of these ferrites, various metal precursors and phosphorus sources were used and the differences among the ferrites produced were revealed. To the best of authors' knowledge, there are no reports available in the literature on the preparation and catalytic performance of P-modified ferrites.In this study, the catalysts were prepared by the sol-gel auto-combustion method and characterized using FTIR, XRD, TEM, SEM, DR/UV–Vis and H2-TPR techniques. The ferrites activities for phenol degradation under catalytic and photocatalytic reaction conditions were examined and compared.The nitrate precursors of Co(NO3)2˖6H2O and Fe(NO3)3˖9H2O were purchased from Merck, whereas the chloride precursors of CoCl2˖6H2O and NiCl2˖6H2O were obtained from ACROS. The phosphorus precursors were ortho-phosphoric acid (H3PO4, Scharlau) and diammonium hydrogen phosphate ((NH4)2HPO4, Sigma-Aldrich).A series of cobalt-based ferrite catalysts were synthesized using the previously reported sol-gel auto-combustion method of ferrite nanoparticles [5]. As presented in Table S1 (Electronic Supplementary Information, ESI), three different sets of the catalysts were synthesized, namely: pure-, mixed-, and P-modified cobalt ferrite nanoparticles. The pure CoFe2O4 ferrites were synthesized using either chloride (CoCl2.6H2O) or nitrate (Co(NO3)2 .6H2O) metal precursors. The chemical formula of the mixed ferrites is CoxNi1-xFe2O4, where x= 0.3, 0.5, and 0.7. These three mixed ferrite samples were synthesized using chloride precursors of cobalt and nickel. The P-modified (0.7 and 1.5 wt%) ferrites (P-CoFe2O4) were synthesized using either H3PO4 or (NH4)2HPO4 as phosphorus sources.A stoichiometric amount of the divalent metal ion (M = Co and Ni) precursor and the iron (Fe) precursor were dissolved separately in distilled water. The two solutions were then mixed, and a certain amount of citric acid (CA) was added to the solution. The molar ratio of M:Fe:CA was 1:2:3. The solution was heated up to 80 °C with continuous stirring, followed by the addition of ammonium hydroxide to adjust the solution's pH to 8. The mixture was heated to form a gel that was then combusted to form an ash-like powder. Finally, the obtained catalyst was easily ground to a fine homogeneous powder using a mortar and pestle. The P-modified ferrites were also prepared using the same sol-gel synthesis method. Four CoFe2O4 catalysts were synthesized with 0.7 wt% or 1.5 wt% loading of phosphorus using either H3PO4 or (NH4)2HPO4 precursors. In the same solution of the metal precursors, the corresponding volume of phosphorus precursor was used and the same synthesis process was followed.The prepared catalysts' purity and structure were analyzed by X-ray diffraction (XRD) using PANalytical Powder Diffractometer (X'Pert PRO) with Cu-Kα radiation at 40 kV, λ= 1.5406 Å and 40 mA. FTIR spectra were collected in the range of 400–4000 cm−1 using Bruker ALPHA-Platinum ATR. The scanning electron microscopy (SEM) technique was used to characterize the prepared catalysts' morphological features by FEG QUANTA 250 operated at 30 kV. Transmission electron microscopy (TEM) measurements were conducted with FEI Tecnai 20 operated at 200 kV. Diffuse reflectance (DR) spectra were recorded using Cary 5000 UV-VIS-NIR Spectrophotometer. H2 temperature-programmed reduction (H2-TPR) measurements were performed using a Quantachrome ChemBET-TPR/TPD instrument.The catalytic activities of the various ferrites solids were assessed for the degradation of phenol in water at room temperature. A typical catalytic reaction was performed in a 150 mL beaker containing 95 mL of 200 ppm of phenol solution, 5 mL of 30% H2O2, and 60 mg of the catalyst under stirring at room temperature. During the reaction, samples of 1 mL of the reaction mixture were withdrawn at specific time intervals and filtered through 0.2 μm nylon membrane filters. The collected filtrates were analyzed by a HPLC system (Shimadzu) to monitor the changes in phenol concentration at λ = 280 nm. The separation was performed using a C18 column (Restek, 150 × 4.6 mm) with a flow rate of 1 mL/min, and a mobile phase composed of 64% water, 35% methanol and 1% acetic acid. The same experimental set-up was used to test phenol's catalytic degradation under photocatalytic conditions using a photoreactor with an Osram-metal halide lamp (400 W, 350–750 nm). The following Eq. (1) was used to calculate the % of phenol degradation: (1) Degradation % = C o − C C o × 100 where C o is the initial concentration of phenol and C is the concentration of phenol after a certain time of the reaction. The kinetics of degradation reactions were studied using the following first-order expression described by Eq. (2): (2) ln C C o = − k app t where C o is the initial concentration of phenol, C is the concentration of phenol after a certain time of the reaction, t is the time (min), and k (min−1) is the rate constant of the reaction.To get an initial insight into the degradation reaction, control catalytic experiments were carried out as described above using the prepared CoFe2O4 catalysts only (without H2O2). The results are displayed in Fig. S1, which clearly shows that no degradation of phenol occurred in the absence of H2O2. This indicates that phenol does not directly react onto the CoFe2O4 surface. However, as discussed in Section 3.2, phenol degradation is caused by CoFe2O4 catalyst in the presence of H2O2.The phase and crystallinity of the CoFe2O4 nanoparticles synthesized from nitrate and chloride precursors were determined using powder XRD. The results are presented in Fig. 1(A), which shows diffraction peaks at 2θ = 18.5, 30.4, 35.8, 37.4, 43.4, 53.8, 57.3, 62.9 and 74.4°. The diffraction peaks can be assigned to (111), (220), (311), (222), (400), (422), (511), (440) and (533) reflection facets that confirmed the spinel cubic structure [10–12]. The observed XRD patterns agreed with the PDF card 22–1086 of CoFe2O4 and with literature data [9]. The XRD patterns of both samples demonstrated their high purity and crystallinity. However, the nitrate-derived CoFe2O4 sample peaks are more sharp and intense, which indicates a higher crystallinity for this sample than the chloride derived CoFe2O4. An average primary crystallite size was calculated using the Scherrer formula given by the following Eq. (3): (3) L = Kλ B cosθ where K is the Scherrer constant, which is set to 0.89 (spherical shape), λ is the wavelength (nm) of the X-ray source, B is the peak width at half maximum (radian), θ is the diffraction angle, and L is the crystallite size (nm). The calculated crystallite sizes using Eq. (3) were 24.8 and 34.0 nm for the chloride and nitrate derived CoFe2O4 samples, respectively.Two vibrational metal-oxygen modes at about 400 and 600 cm−1 are typically observed for ferrites samples [5,13]. The absorption band located at ~ 600 cm−1 (v 1 ) corresponds to the stretching vibrational mode at the tetrahedral sites in the lattice, whereas that at ~ 400 cm−1 (v 2 ) corresponds to the vibrational mode at the octahedral site [12]. For the prepared CoFe2O4 samples, v 1 band was detected at 540 cm−1, but v 2 band was not observed due to instrumental limitations, as shown in Fig. 1(B), which compares the FTIR spectra of the synthesized samples from nitrate and chloride precursors. Remarkable differences were observed in the FTIR spectra of the two CoFe2O4 samples. The CoFe2O4 sample obtained from the nitrate precursor mainly showed the v 1 IR band, whereas additional bands at ~ 1400, 1560 and 3200 cm−1 were observed in the spectrum of the CoFe2O4 sample prepared from the chloride precursor. The first two bands (1400 and 1560 cm−1) can be assigned to O-H bending vibrational modes, and the third broad IR absorption band (3200 cm−1) can be associated with the O-H stretching vibrational modes of water molecules. Such structural differences are expected to induce different physical and chemical properties to the CoFe2O4. Fig. 2 shows high-resolution SEM images of CoFe2O4 samples prepared from the nitrate and chloride precursors. These images clearly show significant differences in the morphological features of the CoFe2O4 ferrites prepared using different precursors. The morphology of the CoFe2O4 sample, synthesized from a chloride precursor, consisting of large agglomerates with cubic and polyhedral structures, is represented in Fig. 2(A) [14,15]. On the other hand, the CoFe2O4 particles prepared from the nitrate precursor consists of regular spherical shapes, as presented in Fig. 2(B). Comparing the SEM images, it appears that the chloride derived particles are more dispersed and larger than the nitrate derived particles. The different morphologies obtained from these CoFe2O4 samples agree with the reported findings of Sinkό et al. [15], which showed that different morphologies were present in synthesized CoFe2O4 samples from different precursors. Nitrate salts have been proposed to be more powerful precursors for the sol-gel phase than chloride salts. In addition, condensation reactions in the nitrate solution are more intensive.The powder XRD patterns of the mixed CoxNi1-xFe2O4 ferrites are presented in Fig. S2(A). The patterns exhibited distinctive diffraction peaks that confirm the cubic structure of the prepared ferrites [16]. The observed peaks correspond to (111), (220), (311), (222), (400), (422), (511), (440) and (533) reflections. Using the Scherrer formula, the crystallite sizes were found to be 27.2, 31.0, and 23.7 nm for Co0.3Ni0.7Fe2O4, Co0.5Ni0.5Fe2O4, Co0.7Ni0.3Fe2O4, respectively. Fig. S2(B) displays the FTIR spectra of the mixed ferrites (CoxNi1-xFe2O4), which are similar to the FTIR spectra of the pure CoFe2O4 sample as shown in Fig. 1(B). In addition to the v 1 characteristic IR band at ~ 560 cm−1, other bands were observed at 1400 and 1560 cm−1 due to to O-H bending vibrational modes, and at ~ 3200 cm−1 attributable to the O-H stretching vibrational modes of water molecules [17].The SEM images of the three CoxNi1-xFe2O4 samples are depicted in Fig. S3, which clearly show differences in the morphological features of the synthesized mixed ferrites due to the variation in their chemical composition. The Co0.7Ni0.3Fe2O4 particles showed irregular flakes. At the same time, agglomerated cubic and polyhedral-like particles were observed in Co0.5Ni0.5Fe2O4 and Co0.3Ni0.7Fe2O4 microimages. A similar observation was reported for CoxNi1-xFe2O4 synthesized from chloride precursors via a hydrothermal process [16].High-resolution TEM analysis was also carried out for the CoxNi1-xFe2O4 nanoparticles. The TEM images showed that Co0.5Ni0.5Fe2O4 and Co0.3Ni0.7Fe2O4 particles possessed cube-like structures, as observed in Fig. S4(c) and (e), while Co0.7Ni0.3Fe2O4 particles exhibited spherical structures as shown in Fig. S4(a). Co0.3Ni0.7Fe2O4 nanoparticles were well-dispersed, while Co0.5Ni0.5Fe2O4 and Co0.7Ni0.3Fe2O4 nanoparticles were agglomerated. The TEM measurements revealed that the average nanoparticle size was ~ 24, 28, and 21 nm for Co0.3Ni0.7Fe2O4, Co0.5Ni0.5Fe2O4, and Co0.7Ni0.3Fe2O4 samples, respectively. The TEM images of these ferrites possessed well-defined lattice fringes. The d-spacing values were estimated and found to be in the range of 0.262–0.266 nm, and correspond to the 311 plane, which agree with the discussed XRD data, and confirm the spinel phase present in the mixed ferrites.A large number of literature reports is available on the catalytic properties of P-modified oxides, but there are no reports on the preparation and catalytic properties of P-modified CoFe2O4 ferrites to the best of our knowledge. Powder XRD was used to analyze the phase structure of synthesized P-CoFe2O4 catalysts, and the findings are presented in Fig. 3 . Typical peaks of the spinel phase confirming the cubic structure of these ferrites are indicated and the intensities of the XRD peaks of P-CoFe2O4 samples are reduced compared to the pure CoFe2O4, which indicates that P-addition affected the crystallinity of the CoFe2O4. Using the Scherrer formula, the crystallite sizes were found to be 20.9 and 24.6 nm for the 0.7 wt% and 1.5 wt% P-CoFe2O4 using (NH4)2HPO4 as a phosphorus source. On the other hand, the crystallite sizes were found to be 22.9 and 22.8 for the 0.7 wt% and 1.5 wt% P-CoFe2O4 using H3PO4 as a source of phosphorus. It is worth noting that in the latter material, 1.5 wt% P-CoFe2O4 using H3PO4, two peaks were observed ~ 28° and 33° which can be attributed to a cobalt phosphate crystal phase [18,19].Fig. S5 shows the FTIR spectra of four P-CoFe2O4 ferrite samples. The ferrite characteristic v 1 IR band was observed in all spectra at ~ 560 cm−1. The vibrational bands corresponding to the O-H bending modes were detected at 1400 and 1560 cm−1, while the broad O-H stretching vibrational band was observed at ~ 3200 cm−1. In addition, there is an absorption band ~ 1025 cm−1, which can be attributed to the phosphate group (PO4 3−) vibrational stretching mode [20–22]. The intensity of this band increased with increasing the phosphorus content from 0.7 to 1.5 wt%.The SEM images of the P-CoFe2O4 synthesized using different phosphorus precursors are shown in Fig. S6. The agglomeration of the cubic-like crystals is clearly shown in these images. A remarkable change can be seen in Fig. S6 as the morphology of P-CoFe2O4 became less homogenous, more porous, and possessed a rougher surface compared to the pure CoFe2O4 sample (Fig. 2(A)).The photocatalytic activity of semiconducting materials is related to their optical absorption ability, which plays an important role in the determination of photocatalytic efficiency. Therefore, optical studies of all the synthesized ferrite catalysts were performed by using UV–Vis diffuse reflectance spectroscopy. The Tauc transformation of DR/UV–Vis spectra allows determining the optical band gap energies (Eg) using the following Eq. (4): (4) α h ν 1 n = A h ν − E g where α is the absorption coefficient, h is Planck's constant, v is the light's frequency, A is a constant “band tailing parameter”, E g is the band gap in electron volt, and n is a number that represents the type of transition. The DR/UV–Vis spectra of the prepared ferrites are presented in Fig. S7. The band gap energy values (Table S2) were estimated from the specta by extrapolating each curve's linear region to intersect the energy axis. The prepared ferrite samples were found to exhibit E g values in the range of 1.78–1.84 eV, which agreed with band gaps energies previously reported [23]. Both CoFe2O4 samples prepared from nitrate and chloride precursors showed the same band gap value of 1.78 eV, implying that the used precursor does not affect the band gap energy CoFe2O4. On the other hand, a slight increase in the E g values was observed when Ni and P atoms were introduced into the CoFe2O4 ferrite. As presented in Table S2, increasing the Ni content in the mixed CoxNi1-xFe2O4 ferrites increased the E g values, demonstrating blue shift as a function of Ni substitution.Catalytic and photocatalytic reactions were conducted at room temperature and neutral pH using pure, mixed and P-modified cobalt ferrite catalysts. Fig. 4 shows the variations in the catalytic activities toward phenol degradation observed for the pure CoFe2O4 ferrites prepared from nitrate and chloride precursors. Interestingly, the various precursors have caused entirely different catalytic activity against the degradation of phenol. The catalytic activity of CoFe2O4 prepared using the chloride precursor is very high, revealing a complete degradation of phenol within 80 and 15 min under catalytic and photocatalytic conditions, respectively. On the other hand, the CoFe2O4 synthesized from the nitrate precursor is catalytically inactive but shows a gradual increase in phenol photodegradation during the reaction time, reaching nearly 39% after 5 h of reaction. The rate of phenol degradation reaction followed the first-order kinetic model (Eq. 3), and calculated k values of these reactions are presented in Table S3.The higher catalytic activity of the CoFe2O4 prepared from the chloride precursor is possibly attributed to an increased concentration of surface active sites compared to the CoFe2O4 prepared from the nitrate precursor. Such differences in the catalytic activity can also be related to the different structural and morphological properties of the CoFe2O4 samples, as shown in Fig. 2, which in turn are linked to site activity. Cubic-like structures were observed for the CoFe2O4 prepared from the chloride precursor (Fig. 2(A)), while a completely different structure characterized the CoFe2O4 prepared from the nitrate precursor (Fig. 2(B)). Previous reports demonstrated that ferrites' morphologies play an important role in enhancing their properties and expanding their applications [24,25]. For instance, different morphologies were obtained using other precursors to prepare ZnFe2O4 ferrite, which was used as a catalyst to degrade safranine-O and remazol brilliant yellow dyes [25]. Nanorods samples achieved the highest catalytic activity compared to nanoflowers and hollow microspheres samples [25].In addition, the difference in the catalytic activity of CoFe2O4 ferrites prepared from nitrate and chloride precursors can be due to variations in their redox properties. Measurements of H2-TPR were carried out to examine the reducibility of CoFe2O4 catalysts produced from chloride and nitrate precursors. The findings are shown in Fig. S8, which typically indicate that these ferrites have been reduced at a high-temperature range. The nitrate precursor H2-TPR profile of CoFe2O4 shows two peaks of reduction: the main peak at 534 °C due to the production of metallic cobalt (Co0) and Fe3O4, while the second peak at ~ 658 °C could be correlated with the production of FeO and Fe0 [26]. On the other hand, the TPR profile of CoFe2O4 from the chloride precursor is considerably different. It starts at 442 °C and is followed by a broad peak at 658 °C and another peak at 860 °C. The distinct TPR profiles demonstrates the difference in the reduction ability (related to the metal cation – oxygen anion bond strength) of these ferrites, and hence the difference in their catalytic activity. The first H2 reduction peak maximum temperature of the CoFe2O4 prepared from the chloride precursor catalyst appears lower than that of the nitrate precursor. This suggests that the redox ability of the CoFe2O4 prepared from the catalyst of the chloride precursor is higher, which led to its higher catalytic activity compared to the CoFe2O4 from the nitrate precursor.The influence of incorporating nickel into the CoFe2O4 nanoparticles on phenol degradation is demonstrated in Fig. S9(C-D). Clearly, the three mixed CoxNi1-xFe2O4 (x= 0.3, 0.5, or 0.7) ferrites were very active compared to the performance of the pure cobalt CoFe2O4 ferrite for both catalytic and photocatalytic reactions. It seems that Co0.3Ni0.7Fe2O4 is slightly more active than the other two mixed ferrites. Complete degradation of phenol was achieved in less than 30 min using Co0.3Ni0.7Fe2O4, while pure CoFe2O4 removed phenol in about 80 min. A possible explanation for this result is that incorporating Ni into the CoFe2O4 structure improved the ferrite's redox properties, which consequently enhanced its catalytic activity. In the case of photoassisted conditions, the three mixed CoxNi1-xFe2O4 (x= 0.3, 0.5, or 0.7) solids exhibited high catalytic activities, showing a complete degradation of phenol within less than 10 min. It is worth mentioning that Co0.3Ni0.7Fe2O4 completely degraded phenol within less than 5 min. Table S4 presents the rate constants of the catalytic reactions of the mixed CoxNi1-xFe2O4 (x= 0.3, 0.5, or 0.7) ferrites.The P-CoFe2O4 ferrites were prepared using (NH4)2HPO4 and H3PO4 precursors with two phosphorus loadings, 0.7 and 1.5 wt%. The catalytic activities of these samples toward the degradation of phenol were compared, and results are displayed in Fig. S9(E-F) and Table S5. The catalytic activities of the P-CoFe2O4 appear higher than the pure CoFe2O4 under both catalytic and photocatalytic conditions. The phosphorus modification significantly enhanced the catalytic activity towards phenol degradation. No reports are available on the preparation and catalytic properties of P-modified CoFe2O4 ferrites. However, the effect of P-modified cobalt oxide (CoO) was investigated to enhance the catalytic oxidative dehydrogenation of propane [27]. Recently, the post-treatment of copper ferrite (CuFe2O4) using H3PO4 showed an improvement in its catalytic activity towards phenol degradation [22]. Obviously, the catalytic activities of P-CoFe2O4 were significantly improved under photoassisted conditions. The highest photocatalytic activity towards phenol degradation was achieved by 1.5 wt% P-CoFe2O4 prepared using (NH4)2HPO4.The catalytic performance of copper ferrite (CuFe2O4) catalysts was demonstrated as highly efficient Fenton-like reagents to treat water from organic pollutants [5]. The degradation of pollutants proceeded by the decomposition of H2O2 to HO• radicals in the presence of CuFe2O4 catalyst. In the current work, CoFe2O4 catalyst may promote similar Fenton-like degradation reactions in which HO• radicals attack phenol molecules, producing new radicals (PhO∙) that may further decompose into smaller molecules such as CO2 and H2O according to the following reactions (5)–(7): (5) H 2 O 2 + CoFe 2 O 4 → 2HO • (6) HO • + PhOH → PhO • + H 2 O (7) PhO • + H 2 O 2 → intermediates → CO 2 + H 2 O Under photocatalytic conditions, the reaction is initiated when CoFe2O4 catalyst absorbs a photon, which leads to the promotion of an electron in the conductive band (eCB −) and a positive hole in the valence band (hVB +). In addition to HO• radicals, the produced eCB − and hVB + enable the reduction and oxidation reaction, respectively, that degrade phenol molecules: (8) CoFe 2 O 4 + hv → e CB − + h VB + (9) H 2 O 2 + e CB − → OH − + HO • (10) PhOH + h VB + → intermediates → CO 2 + H 2 O (11) PhOH + e CB − → intermediates → CO 2 + H 2 O (12) PhOH + HO • → intermediates → CO 2 + H 2 O In this study, AOP was utilized for phenol degradation using different cobalt-based ferrite catalysts. Pure CoFe2O4 and mixed CoxNi1-xFe2O4 (x= 0.3, 0.5, and 0.7) cobalt ferrites were successfully synthesized using the sol-gel auto-combustion method. The ferrite catalysts were prepared using chloride and nitrate precursors. Phosphorus modified cobalt ferrites (P-CoFe2O4) were also synthesized using different phosphorus sources: H3PO4 and (NH4)2HPO4. The prepared ferrites were characterized using powder XRD, FTIR, SEM, TEM, DR/UV-Vis and H2-TPR techniques. The rates of reactions of the pure ferrites prepared from chloride precursors were higher than those prepared using nitrate precursors. The catalytic degradation of phenol was successfully achieved at ambient conditions. Mixed ferrites (CoxNi1-xFe2O4) and P-modified ferrites (P-CoFe2O4) showed higher catalytic performance than pure CoFe2O4 ferrites for the degradation of phenol. The rate of photocatalytic reactions of phenol degradation was comparatively higher than the catalytic reactions for all of the prepared catalysts. Some catalysts achieved complete degradation of phenol under photo-induced conditions in less than 5 min.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 the Research Office at Khalifa University for funding under the project No. LTR14013. The authors would also like to thank Prof. Abbas Khalil from the United Arab Emirates University (UAEU) for the H2-TPR measurements. Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2020.106267.

Nowadays, cobalt-based catalysts are becoming increasingly attractive for various reactions. In this work, the catalytic and photocatalytic activities of different cobalt ferrites were investigated for the degradation of phenol in water at neutral pH. Pure CoFe2O4 and mixed CoxNi1-xFe2O4 ferrites were synthesized using different precursors (chloride and nitrate). CoFe2O4 catalysts derived from chloride precursors showed a higher catalytic efficiency compared to those derived from nitrate precursors. Significant enhancement in the catalytic activities was observed when Ni was incorporated into CoFe2O4. Furthermore, the catalytic activities of phosphorus-modified cobalt ferrites (P-CoFe2O4) significantly improved the photocatalytic degradation of phenol.

Ammonia nitrogen, which is a collective term for NH3 and NH4 +, is a common nitrogen compound. In nature, high concentrations of ammonia nitrogen found in water and soil come mainly from the improper disposal of human waste and livestock excretion, industrial effluents and the excessive use of fertilizers. Although ammonia nitrogen is necessary for plant growth, in excess, it causes eutrophication of rivers, lakes and inland seas, causing serious damage to the environment [1]. Thus, purification of water with high concentrations of ammonia nitrogen is required to ensure environmental sustainability [2].Catalytic wet air oxidation (CWAO) over supported precious metal catalysts and mixed metal oxides is a promising technique for purification of water containing ammonia nitrogen because ammonia nitrogen is selectively oxidized to harmless gaseous compounds [3,4]. However, temperatures far greater than the boiling point of water and high pressures are needed for the reaction to occur [5,6], and this is a major disadvantage.Recently, oxidative decomposition of organic pollutants with O3 in water has gained much attention [7,8] because the reaction proceeds even at low temperatures and ambient pressure. To enhance the decomposition efficiency, the reaction is often performed in the presence of catalysts, which is called catalytic ozonation [9–11]. Catalytic ozonation is applicable to the decomposition of ammonia nitrogen in water, but there are only a few reports. Our group has previously reported the catalytic ozonation of ammonia nitrogen in water [12]. We have tested eight metal oxides (MOx, M = Co, Ni, Fe, Sn, Mn, Cu, Mg, and Al) and found that Co3O4 is the best catalyst in terms of selectivity to gaseous compounds and stability in the reaction solution. However, the decomposition rate of ammonia nitrogen over Co3O4 is only the third best among the tested metal oxides [12]. Subsequently, Co3O4-MgO [13] and MgO [14] have been reported to be more active than Co3O4. However, they have relatively high selectivity to NO3 −, which is a problem in wastewater treatment. In addition, we recently reported that MgO acts as a reactant but not as a catalyst for the decomposition of ammonia nitrogen with O3 [15]. MgO dissolves in the reaction solution, maintaining weakly basic conditions (pH ≈ 9.5) by neutralizing the H+ formed during the decomposition of ammonia nitrogen (NH3 + 4O3 → NO3 − + H2O + H+ + 4O2) with OH−, which is formed upon dissolution of MgO [15]. Thus, the development of catalysts that are more active, selective and stable is necessary for realizing commercial applications.Supported noble metal catalysts have been applied for various total oxidation reactions with O2 [16]. In catalytic ozonation, some supported noble metal catalysts have been used for decomposition of organic pollutants in water [9,10]. Li et al. [17] have found that Pd/CeO2 promotes the decomposition of pyruvic acid in water via catalytic ozonation. Gomes et al. [18] have reported enhancement of the decomposition rate for 4-hydroxylbenzoic esters as well as improvement of chemical oxygen demand (COD) removal by modifying TiO2 with noble metals, especially Pt and Pd. However, to the best of our knowledge, the application of heterogeneous noble metal catalysts for the catalytic ozonation of ammonia nitrogen in water has not been reported yet.In the present study, we tested six Al2O3-supported noble metal catalysts (Ru, Rh, Pd, Ag, Ir, and Pt) for the catalytic ozonation of ammonia nitrogen in the presence of Cl−. We found that Pd was the best metal in terms of activity and selectivity and then further investigated the influence of support on the catalytic performance. In addition, factors controlling the catalytic performance of the supported noble metal catalysts are discussed in relation to the electronic structure of the metals, namely, the d-band center relative to the Fermi level, which is known as the Hammer-Nørskov d-band model.All materials were of analytical grade and were used without further purification. RuCl3·nH2O, RhCl3, PdCl2, IrCl3, AgNO3, H2PtCl6 and activated carbon (AC) were purchased from FUJIFILM Wako Pure Chemical Co. Al2O3 (AEROXIDE® Alu C), SiO2 (AEROSIL® 300) and TiO2 (AEROXIDE® TiO2 P25) were provided by Nippon Aerosil Co., Ltd., and CeO2 (type-A) was provided by Daiichi Kigenso Kagaku Kogyo Co., Ltd.All catalysts were prepared by using a conventional impregnation method. To 10 mL of an aqueous solution of noble metal salt with a concentration of 10 g L−1 (as metal), 2 g of Al2O3 were added. The resulting suspension was vigorously stirred at 353 K for 2 h, and then the water was completely removed from the suspension in an oven at 373 K overnight. The solid was then calcined at 523 K for 3 h and then reduced in an H2 gas flow (20 mL min−1) at 723 K for 2 h. The loading amount of the noble metal was fixed at 5 wt%. For the supported Pd catalysts, SiO2, TiO2, CeO2, and AC were used as supports in addition to Al2O3. Details of the characterization of the catalysts can be found in the Electronic Supporting Information (ESI).The reaction solution containing 10 mmol L−1 of NH4 + was prepared using NH4Cl. Although the initial pH of the reaction solution was adjusted to 7 by adding an aqueous KOH solution, the pH was always ~3 after the reaction. Since the pK a of NH4 + is 9.25, NH4 + was the predominant species in the reaction solution during the reaction. Thus, NH4 + will be used to express ammonia nitrogen throughout the paper.Catalytic ozonation of NH4 + in water was performed in a flask connected to a gas flow line and traps (Fig. S1). Typically, 0.1 g of the catalyst was added to the reactor containing 100 mL of the reaction solution, and then the suspension was heated in an O2 gas flow with vigorous stirring. After the temperature of the suspension reached 333 K, the gas was changed to a mixture of O3/O2 (O3 concentration, 1.88 mmol L−1 and total flow rate, 100 mL min−1) to start the reaction. O3 was generated from O2 using an ozone generator (Tokyu Car Co. SO-03UN-OX). A small portion of the reaction solution was withdrawn at regular intervals and analyzed using two ion chromatographs (Tosoh Co. Ltd., IC-2001) to determine the concentrations of NO3 − and NH4 +. The details of the analytical conditions are described in the ESI.Selectivity to NO3 − (SNitrate) was calculated using Eq. (1): (1) S Nitrate % = Formed NO 3 − Consumed NH 4 + × 100 Since NO2 − was not detected in the reaction solution, and the gaseous products (GasN) were not analyzed in this study, based on N-material balance the selectivity for Gas-N (SGas-N) was calculated by subtracting SNitrate from 100% (Eq. 2): (2) S Gas − N % = 100 − S Nitrate The catalytic performance of the Al2O3-supported noble metal catalysts is shown in Fig. 1 , where the catalytic activity was evaluated in terms of the initial reaction rate per unit amount of metal (r 0, mol h−1 molmetal −1) and the selectivities were obtained after 6 h of reaction. Ir/Al2O3, Rh/Al2O3 and Pt/Al2O3 catalysts show moderate activity for the reaction, whereas the activities of Ag/Al2O3 and Rh/Al2O3 are negligible. On the other hand, Pd/Al2O3 exhibits high activity (r 0 = 20.4 mol h−1 molmetal −1) and selectivity to Gas-N (90%) for the reaction.To determine the intrinsic activity of each noble metal, the catalytic activities were compared on the basis of the turnover frequency (TOF), which is defined as the initial decomposition rate of NH4 + per metal atom exposed on the surface (Eq. (3)): (3) TOF h – 1 = Initial decomposition rate of NH 4 + mol h – 1 g catalyst – 1 Number of metal atoms exposed on the surface mol g catalyst – 1 The amount of chemisorbed CO is commonly used to estimate the number of atoms exposed on the surface of supported noble metal catalysts. However, this method was inapplicable to some catalysts tested in this study because little or no CO was chemisorbed [19]. Thus, the particle size distribution obtained from TEM images (Fig. S2) was used to estimate the number of metal atoms exposed on the surface. Details of the estimation are given in the ESI. As seen in Fig. S3, the TOF for Pd/Al2O3 is much higher than that of the other catalysts, and is more than 3-times higher than those of the second-best catalysts (Rh/Al2O3 and Pt/Al2O3), indicating that Pd is highly active for the reaction.In some catalytic reactions over supported noble and transition metal catalysts, a trend in the catalytic activity is associated with the position of the d-band center (ε d) relative to the Fermi level (E F), called the Hammer-Nørskov d-band model, because d-electrons of metals often play an important role in chemisorption. Also, the state of the d-electrons determines the stability of an intermediate in a transition state [20,21]. Thus, the TOF for the catalytic ozonation of NH4 + over the supported noble metal catalysts was plotted as a function of the position of the d-band center relative to the Fermi level (ε d – E F) calculated by Nørskov and co-workers [21]. As seen in Fig. 2 , the TOF has a volcano-type dependence on the ε d – E F values. It has been reported that the bond energies of metal–hydrogen (M-H) and metal–oxygen (M-O) bonds correlate with the value of ε d – E F and that the larger ε d – E F is, the lower the bond energies of the M-H and M-O bonds are [22]. Although a detailed reaction mechanism for the catalytic ozonation of NH4 + over the supported noble metal catalysts is still unknown, dissociation of the NH bond of NH4 + and activation of O3 to give M-H and M-O, respectively, on the metal particle must be involved in the reaction. Since Pd has a moderate value of ε d – E F in comparison with those of other noble metals tested in this study, activation of the reactants to give PdH and PdO occurs relatively smoothly, and the formed intermediate does not cause poisoning of the active Pd sites. Thus, the moderate nature of Pd may be the reason for the high catalytic activity for the catalytic ozonation of NH4 +.To improve the performance of the supported Pd catalyst, the effects of the support were investigated using SiO2, TiO2, CeO2 and activated carbon (AC), in addition to Al2O3 (Fig. 3 ). The catalytic activity significantly changes depending on the support, whereas the selectivity is barely affected. Pd/AC shows only a negligible activity, and Pd/TiO2 is less active than Pd/Al2O3. It should be noted that Pd/SiO2 and Pd/CeO2 are much more active than Pd/Al2O3 is. Pd/CeO2 shows a slightly higher selectivity for Gas-N than Pd/SiO2. High selectivity for Gas-N is advantageous for the purification of wastewater containing ammonia nitrogen. Thus, Pd/CeO2 is preferable to Pd/SiO2. In addition, the Pd in Pd/CeO2 was less soluble than Pd in Pd/SiO2 was (Table S1), which indicates that Pd/CeO2 is highly stable under the reaction conditions. In fact, Pd/CeO2 could be reused for the reaction without significant reduction in the catalytic performance (Fig. S4), and the reaction stopped after the catalyst was removed from the reaction solution (Fig. S5). Thus, we conclude that Pd/CeO2 is the best catalyst for the ozonation of NH4 + in water.As described so far, Pd/CeO2 exhibits high activity and selectivity to Gas-N as well as high stability for the catalytic ozonation of NH4 + when the reaction is performed in the reaction solution prepared using NH4Cl. However, the catalyst was completely inactive in the reaction solution prepared using (NH4)2SO4. This fact implies that Cl− is involved in the reaction over Pd/CeO2. To understand the role of Cl− in the reaction, changes in the concentration of Cl− in the solution during the catalytic ozonation of NH4 + over Pd/CeO2 were investigated. The reaction was performed in the reaction solution prepared using NH4Cl. As shown in Fig. S6, the concentration of Cl− decreased as that of NH4 + decreased. A similar behavior has been reported for the catalytic ozonation of NH4 + over Co3O4 [11,15]. In the reaction over Co3O4, Cl− is consumed to form chloramines (NH3–x Cl x , x = 1–3) due to the reaction of NH4 + with ClO−, which forms from the reaction of Cl− with O3 (Cl− + O3 → ClO− + O2). Thus, it is plausible that the formation of chloramines occurred over Pd/CeO2. A similar reaction has been reported for the ozonation of ammonia nitrogen in the presence of Br− but in the absence of any catalysts [23,24]. However, the reaction rate is slower in the presence of Cl− than it is in the presence of Br− [24].To confirm the formation of chloramines, we performed the following experiments. Catalytic ozonation of NH4 + over Pd/CeO2 was performed in a reactor connected to three traps containing a strongly basic KOH solution (pH 11), in which a fine powder of MgO was dispersed, placed at the outlet of the reactor (Fig. S1). Since chloramines are less soluble in water and easily undergo decomposition in strongly basic solutions to give Cl− [25], the chloramines formed during the reaction come out of the reactor along with the flow of the O3/O2 mixture and undergo decomposition in the trap solution. Thus, we checked whether or not Cl− was present in the trap solution. After 6 h, Cl− which would have a concentration of 7.3 mmol L−1 if it was present in the reaction solution, was found in the trap solution. Thus, chloramines were formed during the reaction.As mentioned above, NO3 − was the only ionic product found in the reaction solution, and therefore, the rest of the products must be GasN, which includes chloramines and likely other nitrogen compounds (N2, N2O, NO, and NO2). Although the decomposition of NH4 + to Gas-N is desirable for the purification of wastewater, we performed quantitative evaluation of the selectivities to chloramines and other Gas-N for the reaction over Pd/CeO2.To determine the quantity of chloramines remaining in the reaction solution after 6 h, at which the conversion of NH4 + was 99% or more (Fig. 4 ), the pH of the reaction solution was increased to 11 by adding an aqueous KOH solution after the reaction. After increasing the pH, 0.7 mmol L−1 of Cl− was present in the reaction solution. As mentioned above, 7.3 mmol L−1 of Cl− was present in the trap solution. Since the initial concentration of NH4 + was 10 mmol L−1 and 1.0 mmol L−1 of NO3 − formed in the reaction solution after 6 h, the selectivities to chloramines, other Gas-N and NO3 − were calculated to be 80%, 10%, and 10%, respectively, as only monochloramine (NH2Cl) of the three possible chloramines formed (Fig. 4). If dichloramine (NHCl2) and trichloramine (NCl3) had formed, the selectivity to chloramines would have been lower, and that to other Gas-N would have then been higher.Co3O4 is one of the most active and selective metal oxide catalysts ever reported for the catalytic ozonation of NH4 + in the presence of Cl− [11]. Thus, we compared the catalytic performance of Pd/CeO2 with Co3O4 under the same reaction conditions (Fig. 4). As seen in Fig. 4, the activity of Pd/CeO2 is about four times higher than that of Co3O4 at 333 K. Even when the reaction was performed at 293 K, the decomposition rate over Pd/CeO2 was still faster than that over Co3O4 at 333 K (Fig. 4). In addition to the catalytic activity, it should be noted that the selectivity of Pd/CeO2 to Gas-N is higher than that of Co3O4 (90% and 83% for Pd/CeO2 and Co3O4, respectively). Therefore, we may conclude that Pd/CeO2 must be the best heterogenous catalyst for the catalytic ozonation of NH4 + so far reported to the best of our knowledge.Among the noble metals tested in this study, which included Ru, Rh, Pd, Ag, Ir and Pt, Pd showed the highest activity and selectivity to gaseous compounds in the decomposition of NH4 + by O3 in water and in the presence of Cl−. The moderate value of the d-band center relative to the Fermi level (ε d – E F) for Pd is favorable for the activation of the reactants, leading to the high catalytic performance of supported Pd. The support for Pd had a large impact on the catalytic performance, and CeO2 was the best support to give a catalyst with high activity, selectivity and stability. Pd/CeO2 was more active and selective to gaseous compounds, which included predominantly chloramines, than Co3O4, which is one of the best catalysts so far reported to the best of our knowledge.Philip Anggo Krisbiantoro: Conceptualization; Data curation; Formal analysis; Writing-original draft.Tomokazu Togawa: Conceptualization; Data curation; Formal analysis.Koki Kato: Conceptualization; Data curation.Jiequiong Zhang: Data curation.Ryoichi Otomo: Writing-review & editing.Yuichi Kamiya: Formal analysis; Investigation; Project administration; 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.P.A. Krisbiantoro received a scholarship from Indonesia Endowment Fund for Education (LPDP scholarship) from the Ministry of Finance, Republic of Indonesia. The analysis of TEM was carried out with JEM-2010 microscope (JEOL) at Faculty of Engineering, Hokkaido University, supported by Material Analysis and Structure Analysis Open Unit (MASAOU). Supplementary material Image 1 Supplementary data to this article can be found online at https://doi.org/10.1016/j.catcom.2020.106204.

Al2O3-supported precious metal catalysts were tested for the oxidative decomposition of NH4 + in water with O3 and the effects of support on the catalytic activity of Pd were investigated. Pd/CeO2 was found to be the best catalyst in terms of activity, selectivity and stability and was more active and selective than Co3O4, which is one of the best catalysts reported so far. The reaction proceeded only in the presence of Cl−. The moderate value of the d-band center relative to the Fermi level of Pd was found to be favorable for the activation of reactants, leading to a high catalytic performance.

Commonly used catalysts include various metal complexes, salts and metal oxides, as well as transition metals. All of these compounds belong to groups of noble or critical metals [1–3]. Currently, they are widely applied in the developing technologies of wind, nuclear, geothermal and biomass power plants, solar photovoltaics, electric vehicles, batteries, fuel cells, and carbon capture and energy storage systems, and may be used in both homogeneous and heterogeneous catalysis [3]. Therefore, having regard to sustainable development and the rational use of resources, the replacement of metals such as platinum, palladium, ruthenium, silver and gold with other types of catalytically active species is an important issue [4]. In the development of new catalytic materials, there is growing interest in the application of abundant elements such as iron, nickel, cobalt, aluminum, or copper, which are considered reasonable choices in both academic and industrial settings.The use of cobalt in homogeneous and heterogeneous catalysis is attracting great interest, because pure cobalt, as well as its oxides CoO, Co2O3 and Co3O4, exhibit good reactivity and stability [5–8]. On the other hand, from the point of view of sustainable development, cobalt-based catalysts are an interesting alternative because of the low price of cobalt and abundant sources of its minerals [9]. The easy accessibility of cobalt precursors, the predictable redox mechanisms of prepared catalysts, and their resistance to autooxidation – similar to the resistance of second- or third-row transition metals – are reasons for the growing interest in the preparation and use of cobalt-based catalysts in oxidation-reduction and coupling reactions [5,6,10–12].To increase catalytic efficiency, the improvement of activity, reusability, and selectivity is essential for the design of catalysts. In the case of heterogeneous materials, it is possible to design a material with the desired structure and suitable dispersion of active sites [4,13,14]. An attractive new field of study, enabling the provision of such materials in a more economically feasible manner, is biomimetics. The term biomimetics describes the emulation of nature's systems, elements and models to solve the complex human problems [15–17]. A good example of organisms which serve as inspiration for materials science is the marine sponges. These simplest animals possess three-dimensional skeletons made of spongin, chitin, calcium carbonate, or silica, depending on the species [18]. Currently, the most interesting for materials science purposes – especially for catalysis – are those sponges with spongin-based and chitin-based three-dimensional fibrous skeletons [19–24]. These open up new possibilities of creating naturally prefabricated supports for catalytic materials, which are comparatively cheap and renewable. Besides, the carbonization of these scaffolds broadens the range of their potential applications.As a naturally produced biomaterial, spongin represents a renewable source, which gives it an advantage over other commercially available materials. Their unique chemical and structural properties eliminate the necessity of molding and functionalizing the structure before application. Interestingly, the carbonization of spongin-based skeletons is still poorly described in the literature [25,26]. There are only two papers considering applications of carbonized spongin-based scaffolds functionalized with various metal oxides. In previously published work by Szatkowski et al. [25], they carbonized spongin-based skeletons at 650 °C, and then functionalized the resulted biocarbon with manganese(IV) oxide. In another study, by Petrenko et al. [26], carbonization of the spongin-based skeleton took place at a higher temperature of 1200 °C, leading to turbostratically disordered carbon in the form of graphite, which was functionalized with copper(I) oxide to obtain a catalyst for the reduction of 4-nitrophenol.These investigations encourage us to expand the existing knowledge by applying spongin-based skeletons in the development of novel materials for oxidation-reduction reactions. In that event, the main aim of this study was the preparation of biocarbons in low carbonization temperatures using spongin-based skeletons as a biocarbon source and evaluation of their functionalization ability with cobalt oxides. The physicochemical properties of the obtained composites were investigated in detail and then compared to the properties of biocarbons before functionalization. Moreover, prepared composites have been tested as potential catalysts in oxidation-reduction reactions, namely the oxidation of styrene, the decolourization of rhodamine B, and 4-nitrophenol reduction. Catalytic tests included reusability and kinetic studies.Commercial sponges were obtained from INTIB GmbH (Germany). 4-Nitrophenol, styrene, acetonitrile, sodium borohydride, rhodamine B and tert-butyl hydroperoxide were purchased from Sigma Aldrich (Reagent Plus standard). Cobalt nitrate hexahydrate and hydrogen peroxide were purchased from VWR (Germany). All reagents were used without any additional purification.First, the spongin-based skeletons were washed and cleaned as described in previous publications [21,22]. Then the cleaned skeletons were subjected to a carbonization process. The carbonization was carried out at various temperatures (400, 500, and 600 °C) using an R 50/250/13 tube furnace (Nabertherm, Germany) with a heating rate of 10 °C/min; heating was followed by a one-hour plateau and then cooling by thermal inertia to 50 °C. The whole procedure was carried out under a nitrogen atmosphere with a flow of 10 mL/min. In the next step, carbonized skeletons were functionalized with cobalt using the sorption-reduction method. Sorption was carried out by immersing 100 mg of the selected carbon material in 50 mL of an aqueous solution of cobalt nitrate hexahydrate with a cobalt ion concentration of 0.085 M. Sorption was conducted for one hour under continuous stirring of the reagents. Afterwards, 50 mL of 0.1 M water solution of sodium borohydride was added to this mixture, initiating the reduction, which was carried out for one hour. Then the solid material was recovered from the mixture by filtration, washed several times with water, and dried in a dryer at 100 °C overnight. The sorption-reduction procedure was repeated three times. After the procedure, the resulted materials were submitted to ultrasonic treatment (15 min) to check the stability of the metal-containing phase.The morphology and microstructure of the functionalized solids were investigated using a Zeiss EVO40 scanning electron microscope (Germany). The surface composition of the prepared materials was analysed by energy-dispersive X-ray spectroscopy (EDS) using a Princeton Gamma-Tech unit equipped with a digital prism spectrometer. Energy-dispersive X-ray fluorescence spectrometry has been carried out to evaluate the chemical composition of prepared samples using Epsilon 4 spectrometer equipped with a high-resolution silicon drift detector (SDD), typically 135 eV@ Mn-Kα (Malvern Pananalytical). The surface area, pore volume and average pore size were determined using an ASAP 2020 instrument (Micromeritics Instrument Co.). All samples were degassed at 120 °C for 24 h in a vacuum chamber before measurement. The surface area was determined by the multipoint BET (Brunauer–Emmett–Teller) method using adsorption data under relative pressure (p/po). The crystalline structure of the obtained materials was determined by a wide-angle X-ray diffraction method. A Rigaku Miniflex 600 analyser (Rigaku, Tokyo, Japan) operating with Cu Kα radiation (α = 1.5418 Å) was used. The patterns were obtained over an angular range of 10–80°, with measurement steps of 0.02° and 1 s dwell time. The analysis was based on the International Centre for Diffraction Data (ICDD) database. The XPS spectroscopy has been performed using a Prevac spectrometer (Prevac Ltd). Spectra were collected using a hemispherical Scienta R4000 electron analyser. Scienta SAX-100 x-ray source (Al Kα, 1486.6 eV, 0.8 eV band) equipped with the XM 650 X-Ray Monochromator (0.2 eV band) were used as a complementary equipment. The pass energy of the analyser was set to 200 eV for survey spectra (with 500 meV step) and 50 eV for regions (high-resolution spectra): Co 2p, O 1s and C 1s (with 50 meV step). The base pressure in the analysis chamber was 5·10−9 mbar. During the spectra collection, it was not higher than 3·10−8 mbar. Fourier-transform infrared spectroscopy (FTIR) was carried out in ATR mode (Vertex 70, Bruker, Germany). The analysis was performed over a wavenumber range of 4000–400 cm−1 (at a resolution of 0.5 cm−1). The influence of pH on the zeta potential was measured using a Zetasizer Nano ZS instrument equipped with an autotitrator (Malvern Instruments Ltd.) by analysing 0.01 g of catalyst in 25 mL of 0.001 mol/L NaCl solution at 25 °C. The suspensions' pH was automatically adjusted by an automatic titrator using sodium hydroxide (0.2 mol/L) or hydrochloric acid (0.2 mol/L). The zeta potential was obtained from the electrophoretic mobility by the Smoluchowski equation [27].The prepared materials were used as catalysts to treat the organic contaminants styrene, rhodamine B and 4-nitrophenol, in oxidation, decolourization and reduction, respectively. The reaction conditions are shown in Table 1. Detailed information about the catalytic tests is given in Supplementary note 1.The chemical composition and physicochemical properties of the prepared materials were investigated using SEM-EDS and XRF. In Fig. 1, SEM images with EDS maps and the corresponding numerical data are shown.The cobalt percentage by weight varies with the support used; it takes the values 15.64%, 36.93% and 21.84% for Co3O4@C400, Co3O4@C500 and Co3O4@C600, respectively. The variation in cobalt content is related to differences in the elemental composition of carbonized supports as shown in Supplementary note 2, the increase in carbonization temperature results in an increase of carbon content and the changes of the content of other heteroatoms, which can impact the affinity of carbonized biocarbons towards cobalt [28,29]. Taking into consideration the conditions of the functionalization process and properties of used biocarbons, the general conclusion has been found that the chemical composition of C_500 °C support and its highest surface area results in the best affinity to bind the cobalt species during the functionalization process.The presence of the heteroatoms may be explained by the spongin-based skeleton's chemical composition, which is built from amino acids (for more detail, please see Supplementary note 2). The existence of silicon in the form of silicon dioxide has been noticed either in biocarbons and materials after functionalization. This small amount of SiO2 is related to the presence of this element in the sponge structure [26]. On the other hand, aluminum and iron may be incorporated into the skeleton during the sponge's growth [30]. Moreover, the elemental composition was also analysed using the XRF spectroscopy, and corresponding data are gathered in Fig. 1E. The difference in the measured elemental content can be ascribed to the methodological approach of each technique, where during XRF analysis, the X-ray beam penetrates deeper into the sample. Likewise, XRF analysis results show the existence of iodine and bromine (the presence of these elements was not shown using the EDS method) and iron in all tested catalysts. The existence of iodine and bromine is a consequence of bromo- and iodothyronines presented in the amino acid chains of spongin-based material [18]. These results provide evidence about the diversity of elements, which will affect the variety of functional groups and enhance catalytic ability. The variation in the percentages by weight of sulfur, nitrogen and silicon dioxide in prepared composites is related to the different coverage of fibers with the metal-containing phase and the different carbonization temperature of support used to functionalization. Thus, differences in pyrolysis conditions affect the number, type and variety of chemical moieties existing on the surface of the carbonized scaffolds and then final catalysts [26].From the presented SEM results (Fig. 1), it is visible that the fibers are mostly tightly covered by the metal-containing phase. However, this phase is absent on some parts of the fibers, probably due to the irregular diffusion of the substrates and the process of mechanical cutting of the materials. The EDS mapping shows that the deposition of cobalt is uniform. A similar observation applies to the oxygen distribution, which is related to the chemical nature of the metal-containing phase consisting of cobalt oxides, as will be discussed in the XRD analysis. Moreover, to investigate the structure and morphology of the prepared catalyst, SEM images of the supports and functionalized catalysts were recorded at higher magnification, as shown in Fig. 2.From the SEM images in Figs. 1 and 2, it is apparent that the carbonized spongin-based skeleton can be expected to be good support due to its unique fibrous organization. The fibers create a system of open porous channels with different shapes (such as rectangular, pentagonal, hexagonal), and the mesh diameter varies from several tens of micrometers to approximately 300 µm. From the observation of the fibers' morphology before and after carbonization (Fig. 2), it is visible that the superficial microfibrillar arrangement – which is typical of non-carbonized spongin-based scaffolds obtained from commercial sponges (Fig. 2A) – has vanished. However, based on an HRTEM study, it was previously shown that even after carbonization at 1200 °C the triple helices of collagenous origin in spongin are preserved. The development of a mesoporous surface is not yet observed, as described elsewhere [26].Changes in the morphology of fibers after functionalization can be observed. Examining more closely the morphology of the surface of the fibers (Fig. 2C, E, G), it is seen that each catalyst exhibits a different form of the metal-containing phase. Independently of the support used, the tendency to form rod-like and round-shaped aggregates are well visible. For the material Co3O4@C400, the fibers are covered with rod-like structures (marked with arrows in Fig. 2C) with a length of approximately 1 µm. The surface of the Co3O4@C500 catalyst is characterized by the presence of sizable agglomerates with uneven shapes. For this material, similar rod-like structures can also be distinguished (marked with arrows in Fig. 2E), but they are significantly more prolonged and larger. Fig. 2G depicts the surface of the fibers of the Co3O4@C600 catalyst, which is covered with a phase containing two distinct structures: tiny rods (marked with an arrow) and aggregations of round-shaped particles (also marked with an arrow). These intriguing differences can be explained by the differences in physicochemical properties of applied supports (for comparison, see Supplementary note 2) as well as differences in the content of the cobalt-containing phase. With the increasing content of cobalt, the aggregation of cobalt oxide particles is apparent. Thus, the rod-like structures form larger aggregates, and round-shaped particles start to be formed and subsequently become a majority within crystallites.Besides the catalysts' morphology, the well-developed surface is crucial to ensure the proper diffusion and mass transfer. Therefore, the surface area analysis was carried out based on the nitrogen adsorption isotherms, and the obtained results are shown in Fig. 3.BET analysis results reveal that the prepared catalysts possessed a relatively small surface with an average mesopore size varied from 19 to 26 nm. The value of surface area differs with catalyst, and it is the smallest for Co3O4@C600 and the highest for Co3O4@C500 catalyst. Nevertheless, owing to the fact that the most trustworthy value of BET surface area of the precursor of carbonized material – spongin-based scaffolds – is equalled 3.5 m2/g (r2 = 0.9998) [18], it is visible that carbonization followed by functionalization does not lead to the collapse of the structure and reduction of porosity. Moreover, comparing the results with data obtained for biocarbons, it is visible that functionalization results in decreasing the surface area of resulted materials, which is a consequence of the type of the functionalization process. The prepared catalysts possess a bigger pore size than corresponding biocarbons, but the volume of pores decreased. Observed for these materials isotherm type of IVa with hysteresis loop H3 is typical for plate-like aggregates where the pore structure mainly consists of macropores which are not entirely filled with pore condensate.Evaluation of the chemical nature of the metal-containing phase of the prepared catalysts is crucial to considerations of their catalytic ability. For this purpose, XRD analysis was carried out, and the corresponding diffraction patterns are shown in Fig. 4, and the diffractograms of the biocarbons are present in Supplementary note 2.As was confirmed by the SEM + EDS studies, the results of XRD analysis show, as expected, a characteristic pattern corresponding to cobalt oxides (Fig. 4). The metal-containing phases of the prepared catalysts consist mostly of Co3O4 and also of CoO, which is confirmed by the presence of diffraction patterns characteristic for various phases of Co3O4 – such as Co3O4 (112), Co3O4 (211), and Co3O4 (004) (ICDD no. 1285798). Also, only for the Co3O4@C600 catalyst, the diffraction peak characteristic for CoO2 (312) (ICDD no. 14149) is observed, while for Co3O4@C400 and Co3O4@C500 an additional diffraction pattern corresponding to the CoO (200) phase (ICDD no. 22408) can be distinguished. Moreover, the intensity of diffractograms pattern varied with catalyst and is the highest for Co3O4@C500 catalyst and the lowest for Co3O4@C400. Such results are in agreement with the content of the metal-containing phase. It should be noted that the differences between the diffractograms of the biocarbons and functionalized materials are significant (for comparison, see Supplementary note 2). The decrease in the intensity of the broad peak corresponding to C(002) (ICDD no. 9011577) after cobalt oxide capping can be attributed to the heavy atom effect of cobalt [31,32].Considering that the surface composition of prepared catalysts has a crucial impact on the overall catalytic properties, the surface of the catalysts was investigated by XPS. The surface composition of prepared materials supports the results of EDS and XRF methods (for comparison, see Supplementary note 3, Fig. S5). The spectra for Co 2p and C 1s are shown in Fig. 5.The cobalt signal has a typical profile with Co 2p3/2 and Co 2p1/2, showing the satellite peaks associated with Co2+ ions. This magnetic effect of Co2+ is typical for the cobalt oxide-based materials, and it is well described in the literature [33–35]. Either Co 2p3/2 and Co 2p1/2 profiles indicate that the surface species are a mixture of cobalt oxide species, mostly Co3O4 based on XRD results and the addition of cobalt hydroxide. Interestingly, the presence of cobalt hydroxide was not confirmed using the XRD analysis. It can be assumed that cobalt hydroxide is a surface feature related to moisture adsorption from the atmosphere.Based on the XPS deconvoluted spectra, it is apparent that the Co3O4@C600 catalyst is characterized by the lowest content of surface cobalt hydroxide and higher cobalt oxide. The catalytic activity of cobalt oxide is beyond doubt; therefore, a higher content of surface cobalt oxide could be related directly to higher catalytic activity. For C 1s profile, an asymmetric spectrum was obtained, whose deconvolution led to three secondary peaks: the principal peak is ascribed to C‒C bond with the energy around 285 eV for all materials. The other two peaks at higher binding energies: 286–287 eV and 289 eV; are referred as C‒O and C‒O‒O bonds, respectively, and are associated with carbonates species adsorbed from the atmosphere on the sample surface. Despite the similarities, for the Co3O4@C600 catalyst, the peak assigned for C‒O‒O bond is well separated from the principal peak due to a very low signal corresponding to C‒O (blue subpeak in Fig. 5D). That coincides with the highest content of surface cobalt oxide observed in Co 2p spectra (Fig. 5C) and could be evidence of cobalt-support interactions.As was mentioned before, higher temperatures of support preparation result in the preparation of different biocarbons that interact differently with cobalt. Nevertheless, it should be pointed out that the prepared solids differ from each not in the composition and oxidation state of cobalt-containing phases, but only in the content of these particular species, which is in agreement with the results of other performed analysis. The values of binding energies slightly vary between the catalysts, which can be explained by variation in elemental composition.The surface functional moieties of the catalysts were investigated using ATR-FTIR analysis. The spectra obtained are shown in Fig. 6.An intense broadband characteristic for stretching vibrations of O‒H and N‒H groups (wavenumbers in the range 3500–3450 cm−1) was observed. From a comparison with the FTIR spectrum of the pure spongin-based skeleton (for more details see [23,36]), it is evident that in these regions three peaks corresponding to O‒H and N‒H vibrations (3200–3000 cm−1) together with peaks derived from C‒H (2950–2850 cm−1) should be present. The observed strong overlapping of broad peaks characteristic for O‒H groups indicates the existence of cobalt-bound hydroxyl groups and probably hydrogen-bound water molecules. The peaks at 1440–1435 cm−1 visible for all examined solids can be assigned to the O‒H bending vibrations characteristic for carboxylic groups. The peaks in the wavenumber range 1380–1370 cm−1 presumably result from bending vibrations of phenolic O‒H bonds. Bands characteristic for stretching vibrations of S˭O bonds are well visible in the wavenumber range 1340–1350 cm−1. These peaks provide additional confirmation of the presence of sulfur within the structure of the prepared solids. Bands characteristic for stretching vibrations of carbonyl groups (wavenumber 1770–1760 cm−1) and bending vibrations of aromatic C‒H groups (wavenumbers in the range 800–700 cm−1) are present in each of the presented spectra. Besides, the FTIR analysis confirms the presence of cobalt oxide in the structure of the catalysts, based on a band characteristic for Co‒O vibrations at 540 cm−1, and another at 640 cm−1 associated with O‒Co‒O vibrations. The existence of these bands is a consequence of the influence of phenolic groups, which stabilize the process of the formation of Co3O4 particles during the reduction [12,13,37,38].Interestingly, although the superficial techniques (EDS, FTIR) reveal the absence of silica or its presence in only small amounts (less than 0.55 wt%), in the XRD pattern, peaks corresponding to silica compounds are well distinguished. This suggests that silica moieties are located in the bulk of the prepared catalysts.Electrokinetic behavior was analysed to better and indirectly understand the colloidal properties of surface activity by measuring the zeta potential in NaCl solution as an electrolyte at various pH values, as illustrated in Fig. 7.The zeta potential of Co3O4@C400, Co3O4@C500, and Co3O4@C600 catalysts is negative in the pH range 8–4 and becomes positive when the pH decreases below 4. The measured isoelectric points are 3.00 for Co3O4@C400, 3.40 for Co3O4@C500 and 3.50 for Co3O4@C600. This increase in the value of the isoelectric point for biocarbons with higher carbonization temperatures might be explained by changes in the content of electron-donating groups such as hydroxyl, amino or alkyl groups. The tendency can be observed that for Co3O4@C400 material, which has a significantly lower oxygen content, the isoelectric point has been measured at the lowest pH. Interestingly, comparing the curves of zeta potential vs pH of prepared catalysts with results obtained for biocarbons – the shift of isoelectric point towards higher values is observed (see Supplementary note 2 Fig. S4). Moreover, with increasing carbonization temperature, the IEP of prepared composites is shifted towards lower values, showing the opposite tendency when comparing the results presented for prepared catalysts to those obtained for biocarbons. Such results might be a consequence of the change of surface functional groups after the functionalization process, which is confirmed when FTIR spectra are compared.The dependence of the zeta potential on pH results from the protonation/deprotonation of the surface groups. In an aqueous solution, the surface of cobalt oxide is covered with hydroxyl groups, as was also shown by the FTIR measurements. Therefore, with the cobalt oxide surface denoted as ≡M‒OH, the protonation (1) and deprotonation (2) reactions can be written as follows [39,40]: (1) ≡M‒OH + H+ ↔ ≡M‒OH2 + (2) ≡M‒O– ↔ ≡M‒OH + H+ In general, the metal-containing phase of the prepared catalysts is hydrated in an aqueous environment. Thus at acidic pH (up to the isoelectric point), the surface of the catalyst is positively charged (together with the hydroxyl, carboxyl and amide groups of the support). However, with increasing pH, the deprotonation reaction occurs, and the surface functional groups then become negatively charged.Styrene oxidation was chosen as a model reaction to test the catalytic activity of the prepared solids. According to the literature, the solvent has an essential role during the catalytic oxidation of styrene and can influence the mechanism of oxidation [41,42]. Acetonitrile was chosen as a solvent because of its polarity. At the beginning of the investigations, hydrogen peroxide was applied as an oxidant. The results obtained are shown in Supplementary note 4. After these unsuccessful attempts, the oxidant was changed to tert-butyl hydrogen peroxide. The results are presented in Table 2.From the data in Table 2, it is apparent that the conversion of styrene ranges from 86.75% for the Co3O4@C400 catalyst to 93.2% for Co3O4@C600. Interestingly, the change of oxidizing agent led to a change of mechanism of the reaction because the epoxide is the main product, formed with a selectivity depending on the catalyst used (69.7% for Co3O4@C500, 74.6% for Co3O4@C400 and 72.4% for Co3O4@C600). Benzaldehyde was not observed as a product. However, from the GC results, it is evident that benzoic acid and phenyl acetyl aldehyde are the main minor products. The fact that Co3O4@C400 has the lowest oxidation activity of all investigated catalysts may be related to the fact that it has the lowest content of cobalt oxide within its structure. However, the most surprising result is that the highest oxidation activity was produced by the Co3O4@C600 catalyst, which has a lower content of cobalt oxide than Co3O4@C500. It appears that the content of cobalt oxide is not the main factor affecting the catalytic activity. One of the possible explanations for this is the structure of the two catalysts: the Co3O4@C500, despite having the highest content of cobalt, the metal-containing phase forms huge agglomerates, which can hamper the diffusion of the substrate and decrease the contact area between the reagents and the surface of the catalyst. In contrast, the Co3O4@C600 catalyst, with a lower cobalt oxide content, is characterized by diversified morphology, the highest content of surface cobalt oxide, and the largest average pore size. Thus, it seems to offer better access to the catalyst's surface.Because the Co3O4@C600 catalyst possesses the highest catalytic activity, this material was used to evaluate reusability. The same catalyst was used repeatedly five times. After each catalytic cycle, the catalyst was washed with acetonitrile and acetone and dried for 12 h. The results indicate decreasing selectivity of styrene oxide formation, from 72.4% in cycle 1–33.6% in the fifth cycle (Table 2). The conversion of styrene also decreases, from 93.2% in the first cycle to 54.1% in the last catalytic run. Detailed data related to styrene and TBHP conversion and selectivity for the formation of styrene oxide and the respective TON values, are presented in Table 2.The most striking observation to emerge from the data concerns the changes in the conversion of TBHP with the catalysts used. The highest conversion is observed for the Co3O4@C600 catalyst, which also gives the highest styrene conversion. Co3O4@C400 and Co3O4@C500 produced similar TBHP conversion values (approximately 50%). For the Co3O4@C600 catalyst, both styrene and TBHP conversion decreased during the reusability tests. The variation in TBHP conversion may be the main reason for the decrease in selectivity for styrene oxide as well as the general decrease in the conversion of styrene. The decline of activity during the catalytic cycle is apparent. However, in the case of the TON value, a decrease occurred after the third cycle. Surprisingly, the TON value for the Co3O4@C500 catalyst is the lowest of all, which agrees with the proposed finding that the morphology of the catalyst surface significantly impacts the catalytic ability of the tested material.While the calculated rate constants are compared, it is apparent that in the reaction with Co3O4@C600 catalyst, the calculated rate constant exhibits the highest value – which is in accordance with all previously mentioned results. Moreover, during catalytic cycles, the value of the rate constant decreases by approximately 30% compared to the rate constant calculated for the first cycle. Important to note is the fact that when the values of rate constants calculated for other catalysts are compared, it is visible that the k value obtained after the fifth reaction cycle is still higher than the value calculated for the reaction with Co3O4@C400 catalyst. On the other hand, the value of the rate constant calculated for the reaction with Co3O4@C500 is significantly higher than for Co3O4@C400 catalysts, resulting from high styrene and TBHP conversions. Therefore, despite the high cobalt-containing phase and the highest BET surface area of Co3O4@C500, the best catalytic activity in the styrene oxidation was achieved by Co3O4@C600. That could be explained by the synergistic effect of diverse morphology of the surface's fibers with the highest content of surface Co3O 4 and the largest average pore size that impacts the catalytic activity.Furthermore, to evaluate the possible effect of the supports on the catalytic activity, the carbonized materials were also applied as catalysts. The results showed that spongin-based scaffolds carbonized at different temperatures exhibit activity in styrene oxidation. Interestingly the C_600 material possesses 90% conversion and 60% selectivity in styrene oxidation. While comparing results of Co3O4@C600 in the same reaction, they are slightly higher in both cases. Therefore, it could be concluded that functionalized materials are responsible for the high catalytic activity performance where improvement seems to be associated with selectivity. For biocarbons obtained at lower temperatures, the functionalization with cobalt species provides enhancement of catalytic properties in both conversion and selectivity; therefore, the role of cobalt cannot be denied. For more information, see Supplementary note 5.Based on the currently available literature [43–47], a mechanism for the oxidation of styrene by carbonized spongin-based skeletons functionalized with cobalt oxides is proposed, as shown in Fig. 8.Firstly, TBHP molecules coordinate the Co sites, which results in the formation of CoIII-oxo species. At the next stage, those species are transformed into active CoIII-peroxo species due to the presence of TBHP molecules. Then the interaction between the CoIII-peroxo species and the styrene C˭C bonds results in the formation of peroxo metallocycles. At the last stage, the styrene oxide is formed due to the breaking of these peroxo metallocycle species, with simultaneous regeneration of CoIII sites. The lack of benzaldehyde as a reaction product implies that direct cleavage of the C˭C bond of styrene does not occur.After the successful oxidation of styrene, the prepared catalysts were applied to the decolourization reaction of rhodamine B in water to check the catalytic behavior of the prepared materials in an aqueous environment. Rhodamine B is a model dye frequently described in the literature [10,21,48,49]. The UV–Vis spectra recorded for the reactions carried out with the catalysts Co3O4@C400, Co3O4@C500, and Co3O4@C600 are shown in Fig. 9.First, it is important to note that the sorption process was not considered because, at the pH at which the reaction occurs, the surface groups of each catalyst are positively charged, as is the rhodamine B molecule. Consequently, Fig. 9 shows that, at acidic pH, after two hours of reaction, the λmax value typical for rhodamine B decreases gradually; only for the Co3O4@C400 catalyst is a peak still weakly visible after two hours (Fig. 9A). The Co3O4@C500 catalyst exhibits the best decolourization ability; this seems to be related to the fact that its structure contains the highest percentage by cobalt weight. Surprisingly, issues related to the impairment of catalytic activity by forming large crystallites and their agglomeration do not seem to play an important role for this catalyst in this reaction.Additionally, the acidic pH of the solution prevents the aggregation of rhodamine B molecules by the formation of electrostatic interactions between the xanthenes and the carboxyl groups of the dye molecules. On the other hand, it has to be noted that the repelling forces between positively charged functional groups of the catalysts (both the support and metal-containing phase) and the cationic form of the rhodamine B molecules do not affect the decolourization efficiency. Therefore, it can be assumed that the prepared catalysts serve as a place of formation of active radicals, which attack the dye molecules. The formation of hydroxyl radicals follows a pattern similar to the Fenton process, in which the Co2+ cations react with hydrogen peroxide to form hydroxyl radicals, as part of the reaction scheme outlined by the reactions (3)–(5) [50,51]: (3) Co2+ + H2O2 + H+ ↔ Co3+ + OH ∙ + H2O (4) OH ∙ + H2O2 + ↔ HOO ∙ + H2O (5) Co3+ + HOO ∙ ↔ Co2+ + O2 + H+ This finding is supported by the fact that when the decolourization reaction was carried out in the absence of a catalyst, the application of hydrogen peroxide did not lead to significant decolourization even after six hours of reaction (Fig. 9D), due to the formation of hydroxyl ions instead of hydroxyl radicals. Thus, it may be assumed that the oxidation mechanism involves a reaction between the rhodamine B molecule and hydroxyl radicals, which leads to the formation of intermediate products, followed by decomposition to final degradation products. The products of rhodamine B oxidation are discussed in detail in references [52–55].Similarly to the case of styrene oxidation, the Co3O4@C600 catalyst was subjected to reusability tests, as described in Supplementary note 6. These promising results show that a cobalt oxide-based catalyst can be successfully applied in the decolourization reaction several times with an eco-friendly oxidant without losing activity. This is especially important considering that in the existing literature, the removal of rhodamine B has usually been achieved using complicated oxidation systems or photocatalytic processes [21,56,57].The successful application of the prepared catalysts in oxidation reactions motivated us to investigate the possibility of using them for reduction reactions. Therefore, the prepared materials were applied in the standard reduction of 4-nitrophenol to 4-aminophenol in water. The changes in the absorbance in time are shown in Fig. 10. Interestingly, this reaction does not occur without the addition of a catalyst due to kinetic restrictions. Thermodynamically, the reaction between 4-NP and sodium borohydride is possible (E 0 is −0.76 V for 4-NP/4-AP and −1.33 V for H3BO3/BH4) [26,58,59].All of the prepared catalysts exhibited similar catalytic activity in the reduction of 4-nitrophenol. Independently of the catalyst used, 100% reduction was achieved after five minutes of reaction. Therefore, to investigate which material has the best catalytic activity, the reduction's kinetics was calculated based on a pseudo-first-order model (see Supplementary note 1). The calculated rate constants are similar for each catalyst; the values are 0.642 min−1 for Co3O4@C400, 0.742 min−1 for Co3O4@C500, and 0.755 min−1 for Co3O4@C600. However, only for the Co3O4@C600 material is there a visible induction period. This may be a consequence of the change of the local surface charges of the catalyst, as well as slower diffusion of the substrate towards the catalyst surface. However, considering the preparation of the support, where an increase in carbonization temperature led to the enhancement of the skeletal structure of the support [26], the size of fibers and diameter of channels do not change significantly with the support used. Thus, the diffusion of the substrate should not be a limiting step. Therefore, the rearrangements of the surface charges are believed to impact the existence of an induction period. This view agrees with the zeta potential measurements (compare Fig. 6), which demonstrate that at basic pH, the surface of the catalyst is deprotonated and negatively charged. Nevertheless, the 60 s induction period does not affect the reaction time and has essentially no impact on the rate constant. This catalytic behavior of the Co3O4@C600 material is in line with the previously presented results for the oxidation of styrene and rhodamine B dye (see Fig. 9 and Supplementary note 6). For comparison, the results of 4-nitrophenol reduction using biocarbons are presented in Supplementary note 7.Because the Co3O4@C600 material exhibits the best catalytic ability to reduce 4-nitrophenol, this catalyst was chosen to investigate reusability. The results are presented in Fig. 11 in the form of a plot of C t /C 0 vs time.The findings indicate the prolongation of the reaction time with each cycle, from 5 min for the first run to 13 min for the fifth run. It is important to note that the induction period obtained with the Co3O4@C600 catalyst in the first run was not observed in the succeeding catalytic cycles. Interestingly, these results can be correlated with the irreversible change in the catalyst's surface charges after the first reaction cycle. Also noteworthy is the significant decrease in the rate constant after each run, which is a consequence of the prolonged reaction time (for comparison see Fig. 11B).Considering the shape of the curves of C t /C 0 vs. time (Fig. 11A), the rate constant for cycles 2–5 was calculated for the time 4–13 min. It was also found that the calculated reduction efficiency in the 5th minute for cycles 2–5 was high, at approximately 90%. This results from the rapid conversion within the 5 min period (well visible in Fig. 11A, from the steep C t /C 0 vs. time curve for each cycle). The increase in the time taken for the reaction to terminate may be explained by blocking the active sites of the catalyst and partial deactivation of the catalyst surface and loss of catalyst during recovery and filtration.Finally, a reaction mechanism is proposed. According to the literature, there are two main routes of 4-NP reduction using metal particles. The first is related to surface-mediated hydrogen transfer [60], and the second occurs by electron transfer via metal particles [61]. In both mechanisms, the sorption of the reactants occurs before the hydrogen or electron transfer. The reduction occurs due to electron relocation from a BH4 − anion to a 4-NP molecule via the catalyst surface, in so-called surface-mediated electron transfer. In this case, the metal-containing phase is considered as an electron reservoir, and thus the size of the cobalt oxide grains has a significant effect on the catalytic activity [62,63]. This is in line with the results obtained for the Co3O4@C500 catalyst. However, the other reduction path should also be taken into consideration. Therefore, based on the kinetic study and current literature [60,62–64], the authors suggest that the overall reduction mechanism over cobalt-based catalysts involves firstly the production of hydrogen radicals by electron transfer from borohydride, and then the addition of hydrogen species to 4-NP molecules (see Fig. 11C).Furthermore, the catalytic properties of the prepared materials in styrene oxidation and 4-nitrophenol reduction were compared with other cobalt-based catalysts. Data are presented in Table 3.In summary, three different cobalt oxide-based catalysts were prepared by the carbonization of spongin-based skeletons at temperatures of 400, 500, and 600 °C, followed by metallization via the sorption-reduction method. Such an approach of low-temperature spongin carbonization to obtain functional catalysts has been presented for the first time. The spongin-based skeletons are a good and stable source of a renewable, fibrous 3D material with predictable morphology and properties. Moreover, the utilization of spongin-based skeletons fits well into the biomimetic approach, thus generating promising possibilities in designing novel organic and inorganic materials suitable for a wide range of applications. The exhaustive physicochemical analysis of prepared materials has been shown with the comparison to properties of biocarbons. The presence of cobalt oxide as the main cobalt species on the surface of these materials has been proved. Consequently, these composites have been successfully tested as potential catalysts in oxidation-reduction reactions in different conditions. It has been shown that all of these materials possess superior catalytic properties in the oxidation of styrene, decolourization of rhodamine B and reduction of 4-nitrophenol – comparable to other materials described in the literature. The promising catalytic activity of the presented materials can be ascribed to synergistic effects of the support and the metal-containing phase. However, the results show that the temperature of the carbonization process affects the chemical composition and structure of the final product. The catalytic study demonstrated that H2O2 is an inactive oxidant during the oxidation of styrene, and that changing the oxidant to TBHP results in successful oxidation of styrene to styrene oxide, with high selectivity. The reusability tests carried out for the Co3O4@C600 catalyst showed the possibility of its repeated application without significant loss of activity in each tested reaction. The study demonstrates the possible application of carbonized spongin-based scaffolds functionalized with cobalt oxide as catalysts in different types of oxidation and reduction reactions. Sonia Żółtowska: Conceptualization, Investigation, Writing - original draft, Writing - review & editing, Visualization. Juan F. Minambres: Investigation, Writing - review & editing. Adam Piasecki: Investigation. Florian Mertens: Investigation, Resources. Teofil Jesionowski: Supervision, Writing - review & editing, 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 work was supported by the National Science Centre, Poland, project Etiuda no. 2019/32/T/ST8/00414, and the Ministry of Education and Science, Poland. Sonia Żółtowska and Teofil Jesionowski would like to thank Professor Monika Mazik of TU Bergakademie Freiberg for assistance with the catalytic tests. The XPS research was carried out with the equipment purchased thanks to the financial support of the European Regional Development Fund in the framework of the Polish Innovation Economy Operational Program (contract no. POIG.02.01.00-06024/09 – Centre of Functional Nanomaterials).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jece.2021.105631. Supplementary material .

This study concerns an application of spongin-based scaffolds of commercial sponge origin as a naturally structured precursor of carbon material. Further functionalization with cobalt via a simple sorption-reduction method resulted in the preparation of novel catalysts tested in oxidation-reduction reactions. The structure and chemical composition of the prepared materials were investigated in detail, demonstrating the presence of carbonized fibers tightly covered with a metal-containing phase mainly composed of Co3O4. The fibrous structure with open porous canals provides good accessibility for substrates to the surface of the catalysts. Biocarbon material obtained at 600 °C exhibited good catalytic ability in the oxidation of styrene (with high selectivity for the formation of styrene oxide) and rhodamine B compared with other prepared catalysts and biocarbons. Interestingly, all of the prepared materials exhibit favorable activity in the reduction of 4-nitrophenol. A reusability study showed good activity even after the fifth catalytic cycle in both oxidation and reduction reactions. The study proved the adaptability of spongin-based scaffolds to prepare biocarbons with high potential to be used as a support for various catalytic applications.

Most of the chemical reactions for the production of the molecules and materials, such as in the fields of chemical manufacturing, energy conversion, environmental remediation, and human health care, make use of catalysts, either in a homogeneous or heterogeneous process. 1 Generally speaking, the homogeneous operation mode suffers from catalyst reusability and severe catalyst losses. Heterogeneous catalysis offers a much more convenient separation of products from catalysts and is thus vital to many chemical industries. 2–5 Modern society has an ever-increasing demand for environmentally friendly catalytic processes to improve the efficiency of the chemical industry, to reduce emission of pollutants, and to insist sustainable development strategy. In the face of present and future challenges of green chemistry, one of the most important tasks is to develop new catalysts. 6 Metal materials are the most widely used heterogeneous catalysts, with catalytic performances strongly dependent on their composition, size, morphology, and structure. 7 Significant development of modern surface science and computational methods in recent years has made it possible to understand the fundamental factors that govern catalytic performances of metals. 8 Up until now, extensive effectors have been made to improve the catalytic performances of metallic catalysts via rational and precise design of powerful metallic catalysts. In addition, several review articles dealing mainly with the principle of preparation of metallic materials and the correlation between the catalytic performances and their fundamental aspects (structural and electronic properties) have been published. 9–14 Inspired by natural materials with special functions resulting from their unique composition and/or size, morphology, and structure, the design of catalysts with a controllable composition and/or morphology has been the subject of great attention because of their fascinating functions and enhanced properties. In terms of heterogeneous metallic catalysts, extensive in-depth studies have revealed that metals have specific catalytic activity for different kinds of reactions as a result of the diverse radial expansion of d bands of metals and thus the different adsorption strengths of substances on their surfaces. Besides the composition, control over their size, morphology, and structure is essential and necessary for developing superior catalysts, which will result in enhanced intrinsic activity, increased surface-active area, and an improved adsorption model of substances on the surface of catalysts. 15,16 On the basis of such concepts, herein we summarize the recent developments in designing highly efficient metal catalysts with a particular focus on their structure (Figure 1 ). Meanwhile, some novel metal-catalytic reaction systems are also discussed.As we know, the catalytic performances of metal catalysts depend strongly on their surface properties. Therefore, their activity and even selectivity can be controlled by tuning the morphology because it can determine the number and nature of the exposed surface atoms serving as active sites. 14 Recent breakthroughs in the synthesis of nanostructured materials have achieved control of the morphology of materials that are relevant for catalyst design. Many groups have paid intensive attention and made great progress in exploiting metal catalysts with controllable morphology. In the following section, we will discuss some salient features of morphology-controllable synthesis of metallic catalysts using examples mainly from our research.Heterogeneous catalysis is a phenomenon that is exclusively dependent on the reactivity of surface atoms. As a result, a high surface-area-to-volume ratio is desired, that is, more surface-active atoms exposed to reactants. In term of metal catalysis, this demand can be equated with a high degree of dispersion of the metal or a very small metal particle size. The most frequently applied method of metal catalyst preparation is solution-based reduction of metallic precursors. Because of the isotropic structures of metals, they are prone to grow into bigger nanoparticles (NPs). Because the reaction between metallic precursors and reductants is usually highly exothermic, particle aggregation inevitably occurs owing to the high local temperature. Consequently, the metal catalysts prepared by the traditional solution-phase synthetic method usually have low surface areas and broad size distributions, which are harmful to activity, selectivity, and even thermal stability. The ability to synthesize monodispersed metal catalysts permits one to improve their catalytic properties. There are a number of examples demonstrating the influence of catalyst particle size on the reaction performances and discovering the aspects of particle-size-dependent phenomena, which can hitherto be helpful in interpreting structure-sensitive reactions. Therefore, size-controllable synthesis of uniform metal NPs enables the study of size effects on the properties of metal catalysts. Meanwhile, size-controllable synthesis of monodispersed metal catalysts by modified chemical reduction methods is also highly desirable.Abundant publications have already highlighted the benefits of using stabilizers such as surfactants and polymers, which can bind strongly to metallic precursors for precisely tuning the metal particle size. 17 However, the resultant metal NPs always present very limited catalytic activities because of the lack of a “clean” NP surface because of the presence of strongly binding capping agents. Usually, high-temperature treatments are required to remove those organic capping agents, which often inevitably cause deformation and aggregation of metal NPs, leading to the deterioration of catalytic performances. Oleylamine-capped Pd NPs can be synthesized by reduction of palladium (II) acetylacetonate [Pd(acac)2] with tert-butylamine borane in the presence of oleylamine. 18 The as-prepared oleylamine-capped Pd was present in the form of uniformly ultrafine NPs with an average particle size of about 2.8 nm (Figure 2 A). Recovery of these oleylamine-capped Pd NPs through subsequent reactions was rather challenging because of their colloidal properties. To utilize the activity of the Pd NPs while allowing easy reuse, the oleylamine-capped Pd NPs were immobilized into channels of mesoporous silica such as SBA-15. After that, the supported Pd@SBA-15 was subjected to acetic acid treatment for removing oleylamine molecules, leading to a stabilizer-free “clean” Pd surface. One-pot dynamic kinetic resolution (DKR) of racemic alcohols to chiral acetates is a powerful tool to synthesize enantiomerically pure alcohols. However, improving the matchability of metal-catalyzed racemization and enzymatic resolution is still a difficult task. 19 The size-controlled 2.8 nm Pd NPs can deliver a sufficiently high racemization rate for racemic secondary alcohols to ensure the continuous feed of the faster-reacting enantiomer to the immobilized lipase B, leading to an optically pure product with an excellent yield under microwave irradiation (Figure 2B).Control of metal precursors reduction kinetics is a key process for achieving metal shape control. To tackle the metal surface pollution problem, we developed an effective capping-agent-free approach to synthesize size-controllable metal NPs by ultrasound-assisted reduction of M(NH3)6 2+ (M = Co, Ni) with borohydride. 20 The particle size can be controlled by adjusting either the ultrasound power or the ultrasonication time. Because of the strong coordination of NH3 to metallic ions, the reduction process is very slow, resulting in relatively large metal particles (>100 nm), which is unfavorable for the catalytic activity. Taking into account the weaker coordination between halide ligands and metallic ions than that between NH3 and metallic ions, [CoX4]2− formed in the presence of KCl and Bu4PBr was used as a metallic precursor for the synthesis of Co-B amorphous alloy. 21 The simple reduction of [CoX4]2− with borohydride resulted in monodispersed and uniformly spherical NPs (Co-B-X) with an average particle size about 55 nm (Figures 3A and 3B). A series of controlled experiments demonstrated that both KCl and Bu4PBr play a key role in fabricating such monodisperse Co-B NPs with crack channels (inset in Figure 3A) and higher surface B content. On the one hand, KCl and Bu4PBr provide halide anions (Cl− and Br−) to form [CoX4]2− by coordinating with Co2+. On the other hand, the stabilizing effect of Bu4P+ ions can prevent the agglomeration of Co-B clusters. The as-synthesized Co-B-X amorphous alloy was subjected to Heck-olefination of iodobenzenes (Table 1 ) under ligand-free conditions in a mixed solution containing dimethylformamide (DMF)/water = 1/1 as the solvent and K2CO3 as the base, which exhibited the enhanced activity up to two times with the conventionally prepared Co-B-C (Table 1, entries 1 and 2). Besides the larger number of Co active sites, the more electron-enriched Co in the Co-B-X resulting from electron-donation of B also plays a promoting effect, which allows a more favorable oxidative addition of the metallic Co to the carbon-halogen bond. Such a catalyst can be used repetitively 11 times with only a slight loss of activity (8%) for butyraldehyde hydrogenation to n-butanol (Figure 3C). Both the enhanced surface B content and the uniform particle size of Co-B-X led to a prominent increment in its thermal stability, leading to a higher durability than the conventionally prepared Co-B-C (Figure 3C).We have synthesized ultralong single-crystalline Cu nanowires (CuNWs) with excellent dispersibility via thermal reduction of copper acetylacetonate (Cu(acac)2) in a liquid-crystalline medium. 22 In a typical run of synthesis, hexadecylamine (HAD) and cetyltriamoninum bromide (CTAB) were mixed at an elevated temperature to form an ordered liquid crystal (Figure 4 A). Then, Cu(acac)2 rapidly coordinated with Br− from CTAB and HAD, leading to the metal moieties enriched within the tubular channels. Subsequently, metal clusters and particles were built in these channels after reduction. Cu could grow up along one direction to form the ultralong nanowires morphology as shown in Figures 4B and 4C. The average diameter of the nanowires was ∼78 nm, and their lengths varied from tens to hundreds of micrometers. On the basis of the superconductivity of CuNWs, we developed some composite materials with enhanced performances in both catalysis and energy storage, such as CuNWs-ZIF-8 with a core-shell structure, CuNWs-TiO2 with direct growth of rutile nanorods, and a CuO nanotube-graphene sandwich structure. All of these unique architectures could be synthesized by the microwave-assisted thermal reaction because CuNWs were induced as super-hot surfaces by microwave irradiation, which facilitated the nucleation and growth of crystals. Moreover, the strong interfaces created by microwave synthesis between two components could favor the charge transfer during a catalytic reaction or electrochemical reaction. For example, uniform ZIF-8 NPs were prepared by direct growth and coverage on CuNWs surfaces, which exhibited high catalytic activity and stability in H2 production via NH3BH3 hydrolysis. Such one-dimensional CuNWs could offer a rapid electron transfer channel. Meanwhile, those ZIF-8 NPs rapidly transferred H− and H+ ions toward Cu active sites. By a similar process, TiO2 nanorods could also assemble and grow directly on the surface of CuNWs to form a nanorod-nanowire structure; see the field emission scanning electron microscopy (FESEM) and transmission electron microscopy (TEM) images and scheme in Figures 4D–4F. 23 In this composite, TiO2 nanorods promoted light harvesting via multiple reflections to generate more and more photoelectrons. The one-dimensional CuNWs facilitated the transfer and gathering of the photoelectrons as a result of the Schottky barrier at the interface illustrated in Figure 4G, leading to a high H2 production rate during photocatalytic water splitting. The H2 production rate reached up to 5,104 μmol h−0 g−g with an apparent quantum yield (AQY) of 17.2%, which was remarkably high among the noble-metal-free TiO2-based photocatalysts and even exceeded the activity of Pt loaded TiO2 nanorods.Recently, triangular metals have attracted increasing attention owing to their excellent properties from the extremely high anisotropy. 24–27 However, fabrication of a highly anisotropic structure is not thermodynamically favorable since most of the bulk metals display a face-centered cubic shape. Controlled colloid chemistry reaction is one of the main techniques for synthesizing triangular metals in which the capping ligand has proven to be critical. However, the capping ligand alone cannot guarantee the symmetry breaking.Chen et al. developed a rapid seedless growth process to synthesize monodisperse triangular gold nanoplates with sharp tips and tailored edge lengths (Figures 5A and 5B). 28 Systematic studies reveal that iodide ions can promote the formation of triangular gold nanoplates by both selective binding onto the Au {111} facets and oxidative etching to remove other less stably shaped impurities by forming tri-iodide ions (I3 −), leaving behind dominant planar-structured nuclei (Figure 5C).Soon afterward, Gangishetty et al. synthesized AuPd bimetallic nanotriangles through reduction of K2PdCl4 onto the as-prepared Au nanotriangles using ascorbic acid as the reducing agent (Figure 6 A). 29 During Suzuki coupling reactions, the plasmonic Au nanotriangles were used to harvest light to improve the catalytic activity of Pd (Figure 6B). Upon exciting the surface plasmons in AuPd nanotriangles using green LEDs with wavelengths near the maximum of the plasmon band, a significant improvement in the reaction rate of Suzuki was observed (Figure 6C). The nanotriangular structure can initiate strong plasmonic heating effects because of the sharp features and can also provide a much greater number of active sites for catalytic reactions because of the plate-like morphology. Despite the efficiency, the sharp-featured triangles are unstable since severe deactivation was observed after two cycles of Suzuki reactions.Wang et al. successfully prepared well-defined CuNi nanocubes by co-reduction of Cu and Ni metallic precursors (Figure 7 A). 30 The key process was supposed to apply borane morpholine as a reducing agent, resulting in the explosive generation of H2 molecules. More interestingly, the obtained CuNi nanocrystals transformed from octahedral to cubic morphology with increasing borane morpholine. Some chemicals derived from borane morpholine in the synthetic systems seemed to act as capping agents and selectively adsorb onto the (100) facets, making these facets thermodynamically more favorable by reducing their interfacial free energies through chemisorption. A3 coupling reaction with alkyne, aldehyde, and amine was chosen as a model reaction to investigate the dependence of catalytic activity on the facets. Under the same reaction conditions (4 h, 120°C), the catalytic activity of Cu50Ni50 nanocubes was 2.2 times that of Cu51Ni49 octahedra (Figure 7B). Density functional theory (DFT) calculations were performed to elucidate the different catalytic activities of CuNi nanocrystal among different exposed facets. It was found that the surface energy of the (100) facet is always higher than that of the (111) facet at the same molar ratio, which could account for the higher catalytic activity of Cu50Ni50 nanocubes than Cu51Ni49 octahedra.In comparison with a convex structure, the nanocatalysts with an excavated structure expose a larger accessible surface area and higher density of low-coordinated atoms (e.g., edges, corners, and kinks), which provide more active sites and thus promote the activity. Very recently, an approach to the synthesis of Pt3Co nanocubes was developed, which allows the control of the concavity of their surface by just altering the feeding amount of deionized water (Figures 8A–8C). 31 The formation of deeply excavated Pt3Co nanocubes was attributed to the combination of facet-selective capping and oxidative etching (Figure 8D). For both methanol and formic acid oxidation reactions, such a nanocube catalyst with a high degree of concavity exhibited high electrochemical surface area (ECSA) and catalytic activity (Figure 8E), which was attributed to the large surface area, the high energy facets, and low-coordinated atoms.Most of the metals prepared traditionally by solution-phase synthesis are present in the form of solid particles. In general, the particle size should be reduced to be as small as possible to obtain more efficient metal catalysts. However, very tiny metal particles usually add problems in catalyst separation and also induce agglomeration because of high surface energy, leading to a decrease in catalytic efficiency. Porous materials represent a new class of powerful catalysts because they offer advantages over their dense counterparts in terms of increased surface area, easy recovery, and controlled porosity. In the upcoming section, we will discuss the recent progress in the synthesis of porous metals and the studies of their catalytic performances.Nanoporous Au can be synthesized by means of the dealloying of Au-Ag alloys with nitric acid. 32 The monolithic Au consists of a three-dimensional network of ligaments with diameters from 10 to 50 nm (Figure 9 A), depending on the preparation conditions. Meanwhile, it contains a large fraction of low-coordinated Au on the surface of the porous material. As a result of a porous structure permeable for reactants and stable without any support, such Au catalysts delivered selectivities above 97% and high turnover frequencies at temperatures below 80°C in gas-phase selective oxidative coupling of methanol to methyl formate (Figures 9B and 9C). The surprising reactivity of nanoporous Au was ascribed to the unaltered selective surface chemistry of Au and the efficient dissociation of O2.Skeletal rapidly quenched (RQ) metals can be synthesized by alkali leaching of RQ alloys, a similar route to that for the preparation of Raney nickel. Qiao’s group reported a skeletal RQ Fe dealloyed from the RQ Fe50Al50 alloy, in which the Fe-Fe coordination number (CN) is 4.0, only half of the standard Fe-Fe CN in body-centered-cubic (bcc) Fe. Moreover, the Fe-Fe distance (R) is expanded to 2.50 Å, significantly longer than 2.48 Å in the bcc Fe (Figure 10 A). 33 The skeletal RQ Fe is highly reactive and can be used to synthesize a ɛ-Fe2C-dominant catalyst in low-temperature Fischer-Tropsch synthesis (LTFTS) at 423–473 K (Figure 10A). The structural peculiarities of this skeletal RQ Fe (nanocrystalline dimensions, low CN, and expanded lattice) are essential to overcome the seemingly insurmountable hindrance so that the carbidation of metallic Fe to ɛ-Fe2C is kinetically limited at a low temperature, taking into account that the ɛ-Fe2C phase is stable only at low temperature. The as-prepared ɛ-Fe2C-dominant catalyst exhibits superior activity for LTFTS in comparison with the reported Fe and Co catalysts. Moreover, this catalyst displayed activity comparable to that of the noble Ru catalyst, together with the high selectivity to liquid fuels and robustness without the aid of electronic or structural promoters (Figures 10B and 10C). By means of the same approach, the skeletal RQ Ni with peculiar undercoordinated site (UCS) abundant and tensile-strained structural characteristics can also be synthesized. 34 This catalyst has superior activity in the low-temperature COx methanation, in which the turnover frequency (TOF) of CO2 is about eight times that of the highest TOF of CO2 ever reported at 473 K. The DFT calculations reveal that the CO activation barrier decreases when the Ni–Ni distance expanded from 2.49 to 2.51 Å with tensile strain on the Ni (111) surface. The superior activity confirms that the UCSs are the active centers for COx methanation and the tensile-strain effect can further accelerate the rate-limiting CO dissociation step. Therefore, more efforts should be aimed at fabricating undercoordinated catalytic materials.Since the first discovery of mesoporous Pt in 1997 by Attard and coworkers, 35 mesoporous metal catalysts have attracted much interest because of their high porosities, large surface areas, tunable pore sizes, narrow pore-size distributions, high electroconductivities, and excellent activity-structure relationships. To date, various mesoporous metals in bulk, thin film, and powder forms have been synthesized based on different strategies including electrochemical depositions, galvanic replacement reactions (GRRs), as well as soft- and hard-templating techniques. 36–43 Besides enhanced surface areas, mesoporous metals possess a concave inner surface, which can improve catalytic selectivity through changing the adsorption model of reductants. 20 Recently, Yamauchi’s group developed a robust, scalable synthetic strategy to generate mesoporous noble metals (e.g., Pt and Rh) via chemical reduction on general polymeric micelle templates, which is different in concept from traditional soft-templating and hard-templating approaches. 39–41 Preparing high-surface-area Rh at mild conditions is extremely challenging because its surface energy is larger than that of other noble metals, such as Pt, Au, and Pd. The first synthesis of mesoporous Rh was achieved successfully a Rh precursor salt (Na3RhCl6) on self-assembled polymeric PEO-b-PMMA micelle templates (Figures 11A and 11B). 41 Ascorbic acid is used as a reducing agent, and DMF/H2O is selected as a mixed solvent. This synthesis strategy generally involves five steps (Figure 11C): (1) the addition of water causes the PEO-b-PMMA to self-assemble into spherical micelles with a PMMA core and a PEO shell; (2) Na3RhCl6 is dissolved in the solution containing Na+ and [Rh(H2O)3-xCl6-x](3-x)− and then interacts with the aqua complexes with the PEO moieties to form PEO-b-PMMA/[Rh(H2O)3-xCl6-x](3-x)− composite micelles; (3) the Rh ions are reduced to form solid Rh nuclei; (4) Rh nuclei coalesce and further grow into mesoporous Rh nanostructures over the templates; and (5) the templates are removed by a solvent extraction. The as-prepared mesoporous Rh NPs display high surface area with abundant low-coordination atoms, which exhibit great thermal stability. During the electrocatalytic methanol oxidation reaction (MOR), the mesoporous Rh NPs exhibit ∼2.6 times higher activity than that of commercial Rh catalysts (Figure 11D). The Rh NPs also exceed the performance of commercial Rh catalysts for the remediation of nitric oxide (NO) in lean-burn exhaust containing high concentrations of O2 (Figure 11E).In recent years, hollow structures with uniform morphology and good stability have become attractive because of their widespread applications in nanoreactors, adsorption, drug delivery, microelectronics, photonics, and catalysis. 44 In the domain of catalysis, hollow-structured metals with permeable shells represent a new class of efficient catalysts because they offer advantages of high surface area, light density, easy recovery, self-supporting capacity, low cost, and good surface permeability. 45 More importantly, the presence of a concave inner surface in hollow metal nanospheres can exhibit improved catalytic performances relative to that of the NPs exposing only a convex surface, as found on the abovementioned mesoporous metal. Generally, the synthesis of hollow-structured metals can be classified into two categories: soft-templating and hard-templating techniques.The soft-templating strategy is generally used to coat metals onto the surface of “soft” templates by an interfacial reduction reaction. Many endeavors are being devoted to synthesizing hollow metals through a simple vesicle-assisted chemical reduction approach in our laboratory. 20 On the basis of metallic ion-Bu4PBr composite vesicle template, we successfully synthesized a series of Pd-based nanospheres with a hollow chamber, which showed much higher activity than the dense counterpart NPs during liquid-phase hydrogenation or C-C coupling reactions. However, most transition-metal compounds, such as FeCl3, CoCl2, and NiCl2, cannot induce vesicle formation in this way; the hollow-structured metal catalysts are only limited in noble metals. Very recently, we fabricated hollow Ni-Co-B amorphous alloy nanospheres through a vesicle-assisted chemical reduction method. 46 The basis for this synthesis is the use of a Bu4P+/[MX4]2− (M = Ni and Co) composite vesicles templates (Figure 12 A), where spherical vesicle precursors are first formed by the electrostatic interaction between [MX4]2− and Bu4P+. The addition of borohydride induces chemical reduction of the confined [MX4]2− ions into Ni-Co-B clusters, which then develop into particles coating the vesicles to construct a thin shell surrounding the vesicle templates, leading to the hollow spheres composed of uniform NPs (Figure 12B). Coexistence of NiII and CoII species plays an important role in fabricating hollow nanospheric structure because only solid NPs can be obtained in the presence of a mono-metallic precursor. Co-reduction of mixed [NiX4]2− and [CoX4]2− with a Ni/Co molar ratio of 1:1 forms hollow nanospheres in high yield. Previous studies revealed that the chemical preparation of M-B is an autocatalytic reaction, and Ni-Co-B with a Ni/Co molar ratio of 1:1 possesses the highest catalytic reactivity. As a result, co-reduction of a mixture containing the same amount of [NiX4]2− and [CoX4]2− affords a highly active Ni-Co-B catalyst with a stable wall thickness. Such a hollow Ni-Co-B displays a surface-active area of 30 m2 g−g, much higher than 16 m2 g−g of the dense Ni-Co-B. During liquid-phase hydrogenation of 2-ethyl-2-hexenaldehyde (EHEA), this hollow Ni-Co-B catalyst exhibits much higher activity and selectivity than the dense Ni-Co-B catalyst prepared by direct reduction of the mixture of nickel ions and cobalt ions with borohydride (Figure 12C). Additionally, this catalyst is also easily handled in liquid-phase reactions because of its lower density and magnetic property, which allow the reuse for more than seven times with a significant decrease in activity (Figure 12D). This work opens a new avenue for the development of hollow non-noble metal catalysts.The fabrication of hollow metal catalysts with tunable inner and outer diameters is not easy because of the “soft” nature of the structuring units and the difficulty in controlling the phase behavior of surfactants. Therefore, the hard-templating technique is used to deposit a metallic shell onto the surface of “hard” templates (e.g., polymer colloid beads and silica spheres) via a layer-by-layer technique and the subsequent template removal. Such a process shows advantages in transcriptive imprinting of the template morphology and finely tuning the chamber size and shell thickness. However, this method is relatively complicated because multiple steps are often necessary. Strategies based on the mechanisms of galvanic replacement have also been developed to fabricate hollow metal catalysts, which are more facile and cost-effective since the templates could also act as reducing agents to produce metal shells, and thus no additional steps are needed to remove the templates. By using Ni NPs with a tunable size as sacrificial templates, we synthesize hollow Pt-Ni alloy nanospheres with controllable chamber size (7–350 nm) and shell thickness (1.8–22 nm) through a modified galvanic replacement approach (Figures 13A–13D). 47 First, Ni NPs templates are fabricated through chemical reduction of Ni2+ ions with borohydride. The surface of the as-made Ni NPs is wrapped with poly(vinylpyrrolidone) (PVP), which acts as a stabilizer to protect Ni NPs from agglomerating. When PtCl6 2− aqueous solution is added, a portion of PtCl6 2− ions are reduced by the excess borohydride. The produced Pt atoms will preferentially adsorb on the surface of PVP to reduce their surface energy. Meanwhile, partial PtCl6 2− ions can diffuse onto the surface of Ni NPs, followed by immediate reduction to Pt atoms by Ni. Once the Ni NPs are sacrificially dissolved, the produced Ni2+ ions are trapped by the PVP surrounding the Ni NPs and re-reduced to Ni atoms by borohydride. Because Pt and Ni nucleation sites are highly reactive, they diffuse quickly to form homogeneous bimetal in an alloy phase, leading to a shell covering the original Ni core (Figure 13E). During the liquid-phase p-chloronitrobenzene (p-CNB) hydrogenation to p-chloroaniline (p-CAN), such a hollow alloy exhibits much higher activity and selectivity than the solid Pt NPs, together with the excellent durability (Table 2 ). The higher activity could be attributed to the enhanced dispersion degree of Pt active sites (S Pt) and the increased intrinsic activity (TOF). The larger S Pt of Pt-Ni-x(H) compared to Pt(S) is due to the promotional effect of both the chamber structure and the alloying Ni. The increased intrinsic activity could be due to electron-deficient Pt active sites in Pt-Ni alloys, which favors the dissociative adsorption of hydrogen molecules to form H−, leading to the enhanced activity because the nucleophilic attack of H− on the nitrogen atom of the nitro group is a rate-determining step in a p-CNB hydrogenation reaction. In addition, the electron-deficient Pt active sites can “recognize” and preferentially adsorb the electron-enriched nitro group, which can account for the improved selectivity toward p-CAN. Besides, the controllable chamber structure with alloying shell also has a positive influence on the catalytic behavior in the liquid-phase p-CNB hydrogenation to p-CAN (Figure 13F).Tubular nanostructures have well-defined structures in terms of hollow interiors, which stimulate extensive research efforts in recent years because of their unique physical properties and potential applications in advanced electronic or magnetic devices, gas and fluid paths or reservoirs in catalysis, fuel cells, sensors, and separation systems. 48,49 Some catalytic reactions confined within nanotubular materials have been reported owing to the enhanced activities. 50–52 In spite of their importance in catalysis and other nanotechnological fields, only a few reports on their fabrication can be found, possibly because of their extremely difficult processes.Ding et al. synthesized porous Pt-Ni-P composite nanotube arrays (NTAs) with low Pt content through highly efficient template-assisted electrodeposition (Figure 14 A). 53 The as-prepared Pt-Ni-P composite NTAs (Figure 14B) display a unique hollow nanostructure, porous structure, anisotropic nature, and multicomponent effect, which shows high electrochemical activity and long-term stability in methanol electrooxidation (Figure 14C). The synergistic effect allows a homogeneous nanocrystal size distribution and an enormous increase of the electrochemically active surface areas of Pt-Ni-P NTAs, which significantly improves the relative content of Pt (0) and the 5d electron density of Pt in Pt-Ni-P NTAs. The Pt-Ni-P nanotubes display a rough surface and a length of ∼2 μm with an inner diameter and wall thickness of ∼400 and 70 nm, respectively. Thus, the high void volume in Pt-Ni-P NTAs can provide a 3D space for mass transfer of reactant and resultant molecules. These favorable characteristics could sufficiently account for an enormous increase in electrocatalytic activity. The unique hollow tubular structure with porous walls imparts an advantage for the full oxidation of the carbonaceous species generated during methanol electrooxidation, which could effectively reduce the poisoning effect from carbonaceous species (Figure 14D). In addition, well-aligned NTAs can provide a continuous charge-carrier transport pathway without dead ends. By using the similar ZnO-template-assisted electrodeposition method, PdCo NTAs supported on carbon fiber cloth (CFC) (PdCo NTAs/CFC) were also fabricated by Wang et al., which could be used as a high-performance and flexible electrocatalyst in ethanol oxidation for direct ethanol fuel cells (DEFCs). 54 The PdCo NTAs/CFC provide a large surface area and fast electrolyte penetration and diffusion because of the hollow and porous structures. Moreover, because the CFC is highly flexible, the PdCo NTAs/CFC display excellent flexibility. The electrochemical measurements demonstrate that the PdCo NTAs/CFC exhibit significantly improved electrocatalytic activity and durability compared with those of Pd NTAs/CFC and commercial Pd/C catalysts. Most importantly, the PdCo NTAs/CFC exhibit excellent flexibility, high electrocatalytic activity, strong durability, and CO stripping ability, indicating a promising prospect for developing flexible fuel cell devices. As a continuation of the former research, Co nanosheet nanotubes (NSNTs) decorated with TiO2 nanodots (NDS) were supported onto carbon fibers (CFs) (TiO2 NDs/Co NSNTs-CFs), 55 which can promote water adsorption and optimize the free energy of hydrogen adsorption, leading to high catalytic performance toward hydrogen evolution reaction (HER) in alkaline solution.Hierarchically nanostructured materials generally comprise integrated molecular units or their aggregates embedded in or intertwined with other units or aggregates. Learning from nature, the hierarchical nanostructures enable the assembled architectures to obtain unique properties and functionalities. Thus, the hierarchical nanostructures can be considered as advanced materials for their promising applications in various areas including energy storage and conversion, adsorption, catalysis, sensing, and so on. 56–59 The development of such hierarchical nanostructures is an important task in generating advanced nanotechnology to realize better materials performance by a rational combination of multiple components. In this section, we present some examples that highlight the factors related to the hierarchical shape and catalytic properties.The growth of highly anisotropic one-dimensional nanostructures from NPs shows considerable improvement on catalytic, electronic, optoelectronic, and magnetic properties. 60 Because of efficient charge transfer, one-dimensional nanochains are widely used in electrocatalysis. We developed a facile approach to preparing chain-like Co-B amorphous alloy by the chemical reduction of cobalt ions with borohydride in a dodecanethiol-water biphasic system (Figure 15 A). 61 The as-prepared Co-B nanochain comprised uniform spherical NPs with an average size around 29 nm (Figure 15B). These Co-B NPs were connected in one dimension by metal bonding. It was found that dodecanethiol is essential for the formation of Co-B nanochains. Dodecanethiol constructs a biphasic system with an aqueous solution, acts as a stabilizer for Co-B NPs, and also induces the dipoles on Co-B NPs. Because of the resonant enhancing effects from the linearly ordered array of both magnetic moments and electric dipoles, such Co-B nanochain catalysts exhibited stronger ferromagnetic property and higher electrochemical activity (Figure 15C) than the conventionally prepared Co-B NPs catalysts. The ordered array of Co-B NPs in the Co-B nanochain catalyst facilitated the electron transfer, leading to the enhanced electrochemical activity.Very recently, Yuan et al. synthesized a series of Co-based alloy nanochains by a direct current arc-discharge method. 62 The schematic synthesis of Co-Fe alloy nanochains is presented in Figure 16 A. Because of the high arc temperature, the anode evaporated into a homogeneous atomic mixture. Once the mixture cooled down, the Co-Fe alloy was formed; it displayed uniform nanochains with a diameter around 40–50 nm, which could couple to each other and ranged up to several micrometers (Figure 16B). The nanochains with different Co/Fe ratios were synthesized; among them, Co7Fe3 exhibited the optimal performance in an oxygen evolution reaction (OER) with an onset potential of 1.50 V (versus reversible hydrogen electrode [RHE]) and an overpotential of 365 mV at 10 mA cm−m, much higher than that of the Co and Fe nanochains (Figure 16C). The good OER activity of the Co7Fe3 nanochains can be mainly attributed to two aspects. First, the metal can inject electrons into the surface oxide, which manipulates the work function of the oxide and improves its oxygen evolution efficiency. Second, Fe doping can introduce defects into the surface oxide, which could increase the active sites and ECSA. The Co7Fe3 nanochains also exhibit excellent stability with 92.0% current retention after a long-term chronoamperometry test. Cobalt-based alloys with other metals (Ti, Nb, and Mo) have also been synthesized by the same method, which shows promising applications in the OER.Onion-structured materials, generally called multishell spheres, represent those spheres of multiple concentric shells with different diameters. It can be regarded as a nestification of several hollow spheres. 63 Metals in onion-like structures are often synthesized through a hard-templating technique and self-assembly method. The hard-templating technique usually uses a hollow-structured template instead of a solid one and coats metals onto both the inner and the outer surface. This strategy is comprehensible in concept and difficult in practical applications.Composite vesicle systems are fascinating self-assembled structures that can be applied as efficient directors for the rapid synthesis of multishell materials. Currently, we successfully fabricated onion-structured Pd nanospheres by using a simple self-assembly template (unpublished data). The basis for this synthesis is the use of a composite multishell vesicle system that comprises didodecyldimethylammonium bromide, cyclohexane, and water. The region near the surfaces of the vesicle layers can gain a higher concentration of Pd ions. Once chemical reduction of Pd ions with sodium hypophosphite occurs, Pd clusters are produced and develop gradually into several concentric hollow spheres (Figure 17 A), leading to onion-like spheres with an average shell thickness of about 2.2 nm (Figure 17B). The shell layers can be facilely adjusted by changing the amount of cyclohexane. The onion-like Pd nanospheres delivered much higher electrocatalytic activity toward ethanol oxidation for direct alcohol fuel cells than Pd NPs and commercial Pd/C (Figure 17C), mainly attributed to their larger active surface area and the unique multishell configuration that accelerates the electron transfer.Jia et al. developed a low-temperature interface-induced assembly approach to synthesizing onion-like Pt-Cu alloy nanocrystals. 64 A two-phase reaction system comprising HAD and water was designed to form multilamellar micelles. With the assistance of ascorbic acid, Pt-Cu NPs were formed and can be controlled by assembling at a low temperature in the interlayer regions of multilamellar micelles (Figure 18 A). The obtained products consist of superparticles with diameters of 50 ± 10 nm (Figure 18B). High-resolution TEM (HRTEM) (Figure 18C) images reveal that each particle is composed of nested multilayers characteristic of an onion-like structure. Their three-dimensional nanostructure was certified by tilting the sample along the x and y axes from −30° to +30°. The high curvature of each layer makes the onion-like Pt-Cu alloy nanocrystals expose a high density of defects, which acts as a way of releasing strain caused by the bending process. The unique structure with a confined interior as well as the existence of twin defects makes the onion-like Pt-Cu alloy nanocrystals exhibit excellent catalytic properties in the electro-oxidation of methanol and ethanol (Figure 18C) and cycloaddition reactions.Core-shell nanomaterials represent unique spheres containing a middle core and an outer shell. Due to the synergism between two or more components, core-shell nanomaterials exhibit extraordinary properties in many areas, such as electronics, biomedicine, pharmaceuticals, optics, and catalysis. 65–67 For traditional core-shell materials, the core and shell components are compactly attached without any interspaces. 68 In the synthesis of multi-metallic materials from the mixed-metal salt precursors, it is difficult to ensure that different kinds of metal atoms contribute equally to the metal-metal bond formation because of their diverse reaction kinetics. For example, when more than one metal salt precursor is co-reduced in a homogeneous reaction solution, it is difficult to simultaneously control the reduction and nucleation process of different kinds of metals because of the difference in their redox potentials and chemical behaviors. Generally speaking, the expensive noble-metals like Pd and Pt with higher standard reduction potentials are more reactive than non-noble metals such as Fe, Co, and Ni during the co-reduction process. As a result, co-reduction of those mixed-metal salt precursors will be anticipated to lead to noble-metal-enriched cores and non-noble-metal-enriched shell particles. Considering that heterogeneous reactions take place on the surface of catalysts, having a large fraction of expensive noble metals in the core of the catalysts is undesirable. Thus, the design of new core-shell catalysts with non-noble-enriched cores and noble-metal-enriched shells represents a promising way to improve the activity at a low cost. A GRR is a facile and cost-effective approach to preparing noble-metal catalysts by using non-noble metals as the sacrificial templates. 69 Recently, we prepared a core-shell Pd@Co-B with high dispersion of Pd onto Co-B amorphous alloy nanospheres by GRR between Co-B uniform nanospheres with a diameter around 55 nm and Na2PdCl4 (Figure 19 A). 70 During the reduction Pd ions, Pd NPs gradually produced and deposited onto the Co-B core, corresponding to the formation of the Pd outer shell and the size decrease of the Co-B core (see Figure 19B). HRTEM images of Pd@Co-B (Figure 19C) clearly illustrate that the actual incorporation of Pd secondary nucleation on the surface of Co-B. Pd@Co-B with tunable Pd content could be achieved by adjusting the amount of Na2PdCl4 in the reaction mixture. Hydrogenation of EHEA to 2-ethyl-1-hexanol (EHO) is of great industrial importance since EHO is a valuable synthetic alcohol used as a synthon for the manufacture of ester plasticizers, coating materials, adhesives, printing inks, and impregnating agents or as an additive in foods and beverages as a volatile flavor. In industry, two-step hydrogenation of EHEA is necessary to produce EHO because one critical issue associated with that process is the partial hydrogenation that leads to a mixture of EHO, 2-ethyl-hexanal (EHA), and 2-ethyl-2-hexenol (EHEO). In general, the unsaturated alcohol, EHEO, is particularly undesirable because of the great difficulty in separating EHO by distillation. The as-prepared Pd@Co-B catalyst allowed the production of pure EHO via one-step hydrogenation of EHEA in liquid phase, which exhibited extremely higher activity and selectivity than either the Co-B amorphous alloy or the Pd catalyst. Results from the catalytic evaluation demonstrated that both metals in Pd@Co-B play important roles in promoting the reaction. The Pd highly exposed on the surface of Co-B amorphous alloy nanospheres is largely responsible for the hydrogenation of C=C bonds in the EHEA molecule, although Pd is relatively inactive for the hydrogenation of C=O bonds in EHEA. The incorporation of Pd can greatly increase hydrogenation ability associated with Co-B for C=O bonds (Figure 19D). Moreover, the core-shell-structured Pd@Co-B exhibited much greater efficiencies than the classical systems generally used in industry, including Cu-Zn-Al, Cu-Cr, and Cu-Ni (Figure 19E). A synergetic effect between Pd and Co can be demonstrated by hydrogen temperature-programmed desorption (H2-TPD) results, which confirmed that the dispersion of Pd on the surface of the Co-B core could provide a much higher concentration of active hydrogen via hydrogen spillover from Pd to Co. In terms of heterogeneous catalysis, hydrogen spillover is a well-documented phenomenon, which can enhance H2 activation ability and thus improve the hydrogenation activity.Yolk-shell structures can be regarded as a variation of the ordinary core-shell structures in which one or several movable cores are encapsulated inside a hollow shell and can move freely in the void space. They are also termed as movable core-shell or rattle-structured materials. 71 The inside core and the outer shell can be either an identical material or two kinds of different materials. As an interesting family of complexes with new nanoarchitectures, the yolk-shell metal nanostructures have attracted tremendous attention because they bring the advantages of two classes of morphologies of the core shell and hollow together. To create high-surface-area amorphous alloys with tunable chamber structures, we developed a novel hard-soft co-templating method of catalyst synthesis. 72 Several mesoporous M-B (M = Fe, Co, and Ni) amorphous alloys with tunable chambers including yolk-shell or hollow structures were prepared by syringe-squeezing a solution composed of micelles containing Brij-76 [C18H37(OCH2-CH2)10OH] surfactant and metallic ions into borohydride solution to form oil droplets (Figure 20 A), followed by an ecto-entad stepwise reduction of metallic ions with borohydride due to a Kirdendal diffusion process. Reaction temperatures played important roles in determining the morphology of Ni-B amorphous alloys. Only random dispersed Ni-B NPs were obtained at a very low temperature. With an increase in reaction temperature, the morphology changed gradually to nanochains, yolk shell (Figure 20B), and hollow (Figure 20C). Moreover, the Brij-76 surfactant also played a key role in determining the M-B amorphous alloy morphology. On the one hand, it could stabilize the syringe-pinhole templated oil droplets in the KBH4 aqueous solution. On the other hand, it also acted as a soft template to assemble micelles, leading to a mesoporous structure in the M-B NPs. During liquid-phase hydrogenation of p-CNB to p-CAN, the yolk-shell Ni-B amorphous alloy exhibited the highest activity and selectivity. The improved catalytic performances could be mainly attributed to the high surface area and the unique yolk-shell chamber which might enrich hydrogen and act as a microchemical reactor.The lifetime is one of the most important factors for evaluating industrial catalysts. In general, the catalyst lifetime depends on the stability against corrosion, leaching, gathering, phase transformation, structure change and/or collapse, and poisoning, etc. Highly stable metal catalysts can be achieved by encapsulating active metal NPs by a protective matrix or shell. Such an encapsulation structure not only promotes the catalytic performances of metal NPs as a result of the unique collective and synergetic effects but also protects metal NPs from gathering and leaching as well as poisoning during the catalyst synthesis and its application in reactions. 73–76 Porous materials have been demonstrated to be suitable host matrices for the encapsulating metal NPs. The window size of the cavity is a key factor for achieving the efficient catalysts. On one hand, the window entrance can offer the possibility for encapsulating metal NPs and preventing their leaching. On the other hand, the window channels allow the diffusion of reactants and products. Additionally, the environments of the pores also have influences on the host-guest interaction and even the catalytic reaction. In the past few decades, metal NPs supported on metal oxides, zeolites, mesoporous materials, and activated carbons have been widely studied as catalysts. SBA-15 and MCM-41 have been well examined among the silica-based mesoporous materials. 77 However, supported metal catalysts on mesoporous silica prepared by a direct impregnation method usually show a non-uniform distribution of active sites and blockage of the pore channels, leading to the decrease in activity and even selectivity. To solve those problems, we developed a simple synthesis approach to confine Ru-B NPs within the mesopores by ultrasound-assisted incipient wetness infiltration of (NH4)2RuCl6 into SBA-15 channels, followed by reduction with borohydride. 78 During liquid-phase hydrogenation of maltose to maltitol, this catalyst delivered high activity up to seven times higher than that of the unsupported Ru-B catalyst because of the small active sites and cooperative effects from SBA-15 support. It also showed excellent durability owing to the confinement effect of mesoporous silica, which protected the metal active sites from agglomeration and leaching. Besides improving the synthesis method, other efforts have also been made to promote the dispersion of Ru-B NPs and enhance the metal-support interaction, including the change in the silica morphology and the modification of the silica surface with inorganic and/or organic groups. For example, mesoporous silica nanospheres (MSNSs) with an average diameter around a few tens of nanometers are attractive candidates for catalyst carriers because of their large surface area, short pore channels, and regular nanospherical morphologies. Moreover, the surface properties of mesoporous materials can be modified through the region-selective immobilization of diverse organic groups. MCM-41-type MSNSs externally covered by methyl groups (−CH3) but internally grafted by aminopropyl groups (−NH2) were used as a host matrix for loading Ru-B NPs. 79 The −NH2 and the −CH3 groups served synergistically as effective functionalities for highly dispersing Ru-B NPs within the pore channels of the mesoporous host. Such a catalyst exhibited very high activity in liquid-phase d-glucose hydrogenation, much better than industrial Raney Ni and the commercially available Ru/C catalysts.As a new kind of highly ordered porous material with tunable size, shape, and microenvironment of the pores, metal-organic frameworks (MOFs) are promising platforms for preparing supported metal catalysts by embedding metal NPs into MOF pores to limit the migration and aggregation of metal clusters and/or NPs. Very recently, Pd NPs were encapsulated inside the cages of ethylenediamine-grafted MIL-101 (ED-MIL-101), giving high catalytic efficiencies for the racemization of primary amines. 80 The racemization can be combined with a lipase-catalyzed resolution in a one-pot DKR of rac-1-phenylethylamine, leading to an optically pure product (>99% enantiomeric excess [ee]) with an excellent conversion and selectivity up to 99% and 93%, respectively, which are remarkably superior to those of Pd/MIL-101, Pd/MCM-41, and commercially available Pd/C. Such a catalyst can be easily recovered and reused more than eight times without significant loss of activity. The enhanced catalytic performances were mainly attributed to the amine modification of MIL-101, which endows the MOF cage surface with basic properties and also enables efficient confinement of Pd NPs in the cages to inhibit their gathering. Metal NPs encapsulated in channels of mesoporous materials can be further encapsulated by a protective shell to fabricate unique yolk-shell nanoarchitectures, which could be used for those reactions conducted under harsh conditions. Recently, we applied the combination of such a yolk-shell-structured catalyst and enzyme to act as a powerful platform for one-pot biomass conversion via sequential enzyme-catalyzed hydrolysis of biomass materials to glucose and the subsequent metal-catalyzed hydrogenation of glucose to sorbitol. 81,82 Preliminary studies revealed that the enzyme is easily poisoned when contacting Ru-based catalysts. Meanwhile, the metallic Ru active sites would be covered by the enzyme and the colloidal substances originated from dextrin hydrolysis, leading to a rapid deactivation for the subsequent glucose hydrogenation to sorbitol. To solve these problems, Ru-B NPs encapsulated in yolk-shell silica (Ru-B/mSiO2@air@SiO2) (Figure 21 A) and a yolk-shell-structured material composed of a Ru-B/mCarbon core and a mesoporous silica shell (Ru-B/mCarbon@air@mSiO2) (Figure 21B) were prepared. Assisted by amyloglucosidase, these are essentially binary catalytic systems for one-pot conversion of dextrin to sorbitol. (Figures 21C and 21D) More specifically, the permeation-selective outer silica shell inhibited the diffusion of amyloglucosidase with a large molecular size into the chamber to contact Ru-B on the core, which avoided the poisoning effect on each other. Meanwhile, it also prevented the diffusion of dextrin with a big molecular size into the chamber but allowed the diffusion of glucose and product sorbitol inside the chamber owing to their small molecular sizes. As a result, the enzymatic dextrin hydrolysis to glucose occurred in bulk solution, followed by glucose hydrogenation to sorbitol catalyzed by Ru-B on the core, leading to high reaction efficiency. The enzyme integrates into yolk-shell nanostructure containing Ru-B to form a Ru-B/af-mCarbon@air@af-mSiO2 bifunctional biochemical nanoreactor. 83 The synthesis of such a composite is achieved through a stepwise crosslinking method that involves the covalent attachment of yolk-shell-structured catalyst onto amyloglucoamylase with glutaraldehyde and the subsequent coupling of the composite in the presence of modified dextran. The biochemical composite enables the efficient synthesis of sorbitol through one-pot reactions from dextrin, cellobiose, and even cellulose.The incorporation of metal NPs in the framework of porous materials for heterogeneous catalysis may avoid particle aggregation, movement, and leaching. A multicomponent assembly approach was used to cooperatively assemble surfactant, titania, and nickel precursors in a one-pot process. Ni NPs with sizes ranging from 1 to 6 nm were homogenously embedded within the framework of mesoporous TiO2. 84 During gas-phase hydrodechlorination of the chlorobenzene reaction, the as-prepared 0.23%-Ni/TiO2-DS catalyst delivered a TOF up to 1.5 times greater than that of the Ni/TiO2 catalyst prepared by conventional incipient-wetness impregnation method. Meanwhile, such a catalyst also exhibited much improved stability in a continuous reaction. An enhanced hydrogen spillover effect is believed to play an important role in promoting a hydrodechlorination reaction as a result of the intimate interfacial contact between Ni NPs and the TiO2 support. Meanwhile, 3 nm Pd NPs encapsulated homogenously within 10-nm-thick porous silica shells were synthesized by conducting silica polymerization around oleylamine-capped Pd NPs in a water-in-oil microemulsion system, followed by removing oleylamine molecules via calcination. During the CO oxidation reaction, the core-shell-structured Pd@SiO2 delivered a TOF up to 33 times greater than that of the Pd catalyst prepared by a conventional immobilization method and also showed excellent durability. Detailed studies demonstrated that core-shell configuration played a key role in promoting catalytic performance. The Pd cores with a small size of around 3 nm could efficiently activate CO molecules by weakening the strength of CO adsorption, and the core-shell structure could inhibit Pd gathering during reactions. Meanwhile, the porous silica shells allow the reactants to penetrate into the core-shell-structured Pd@SiO2 and thus increase their accessibility with the Pd cores. Furthermore, the porous silica shell also protected Pd NPs from agglomeration, leading to enhanced thermal stability. 85 By a similar reverse microemulsion method, core-shell-structured M@SiO2 (M = Ni, Co, and NiCo) catalysts were also synthesized. 86 Each SiO2 nanosphere contained a single Ni-, Co-, or NiCo-alloy NP as the core (Figures 22A–22F). During the dry reforming of CH4 with CO2 (DRM) reaction, NiCo@SiO2 delivered high activity and selectivity, superior over either the Ni@SiO2 or the Co@SiO2. The CO2 and CH4 (1:1) could be absolutely converted into CO and H2 with a molar ratio around 1:1 (Figure 22G). More importantly, those catalysts displayed very good durability at a high reaction temperature. For example, the Ni@SiO2 could be used continuously for more 1,000 h without a significant efficiency decrease in the DRM reaction at 800°C (Figure 22H). Kinetic studies revealed that the catalyst with a small metal NP size exhibited higher activity and inhibited carbon deposition, which would poison the active sites. The encapsulation of metal NPs by a SiO2 shell could effectively inhibit the agglomeration of active sites, corresponding to high activity and the long lifetime. An approach to coating Al2O3-supported Ni NPs with a porous Al2O3 thin film through atomic layer deposition was developed. 87 During the DRM to synthesis gas at 800°C, the as-prepared Al2O3/Ni/Al2O3 sandwiched catalyst could effectively protect Ni NPs from gathering owing to the double strong interactions of the Ni active sites with the γ-Al2O3 support and the Al2O3 film. Such a sandwiched Al2O3/Ni/Al2O3 catalyst with 80 layers of Al2O3 thin films exhibited the highest activity. Both CO2 and CH4 conversions reached nearly 100% with absolute selectivities toward CO and H2. More importantly, this catalyst displayed excellent stability and could be used for more than 400 h in the DRM reaction at 800°C without significant deactivation.Metal catalysts are always attractive because of their wide applications in chemical industries and many other areas. Their reactivities and selectivities, as well as stabilities, can be tuned by controlling the morphology because the exposed surfaces of the metals have distinct crystallographic planes depending on the shape. This review provides an overview of recent advances in the synthesis of metallic materials and their fascinating catalytic performances in many reactions. Versatile chemical reduction approaches combined with desired structure templates can be conducted in different systems to fabricate metal catalysts with well-controlled shapes and morphologies, leading to excellent catalytic properties.Industrial application of metal catalysts greatly relies on their stability. In an effort to favor the stability of metal catalysts, the applied reaction type should be considered carefully. For example, under some harsh reaction conditions, such as elevated temperature or pressure, the porous metallic materials usually display an irreversible shrinkage caused by pore collapse. As a result, the porous metallic materials are unstable relative to metal NPs loaded on supports for some reactions, especially strongly exothermic reactions. As electrocatalysts, however, the porous metallic materials are always stable compared with metal NPs supported on carbon. The poor stability of metal NPs supported on carbon is mainly ascribed to the corrosion of carbon supports, which further results in migration, aggregation, and Ostwald ripening of metal NPs owing to their high surface energy and isotropic zero-dimensional structural features. Because of the inherent anisotropic morphology and unique structure, the nanoarchitectured metal catalysts in the absence of carbon supports possess superior durability when used as electrocatalysts even in harsh media.We expect that with further development of chemical reaction engineering and nanomaterial synthesis technologies, more and more new metal catalysts will be developed, which will offer more opportunities for their industrial applications. Obviously, fundamental research is necessary for such new materials to uncover the key factors that greatly regulate their nanostructures and finally determine their catalytic performances. One of the most challenging problems is the underlying mechanism responsible for the trajectories to form metal catalysts with controllable morphology, which should be helpful to achieve the goal of rationally designing catalysts. On the other hand, further efforts must be made to achieve the large-scale assembly of these nanoscale catalysts to realize their practical applications. Moreover, we anticipate that a variety of unique nanostructured metals to allow molecular-level fine-tuning of catalytic performances will emerge soon as a new progress in metal catalyst design. In particular, the development of methodology to control the nanostructures of metal catalysts for further improvement of catalytic activity and selectivity and to fabricate novel efficient catalysts will be the keystone for the future industries.We would like to thank the programs supported by the National Natural Science Foundation of China (21761142011), Singapore National Research Foundation (NRF2017NRF-NSFC001-007), Ministry of Education of China (PCSIRT_IRT_16R49), and Shanghai Government (18JC1412900, 15520711300, and 18DZ2254200).

Metal catalysts have been widely employed in chemical production, medicine manufacture, organic synthesis, and environmental cleaning, etc. Catalyst design plays a key role in enhancing efficiencies including activity, selectivity, and durability. Both theoretical predictions and experimental research has demonstrated that besides the composition, the morphology and/or the porous structure play key roles in determining catalytic performances. This review highlights the recent progress in the controlled synthesis of new metal catalysts with hierarchical structures mainly based on our research. First, we describe the metal catalysts with unique morphologies. Second, we show the metal catalysts with different porous structures. Finally, we summarize the metal catalysts with hierarchical structures. The catalytic performances of different catalysts are also included, and their correspondence to the catalyst structure is explored. This review might supply guidance for designing new and powerful metal catalysts for industrial applications.

Hydrodesulfurization (HDS) has become an increasingly important research topic over the past decades due to the growing demand for transportation fuels [1–7]. Environmental regulations in many countries worldwide demand the production of clean fuels with an ultra-low sulfur content of 10 ppm or less [8–10]. As the traditional HDS catalysts, alumina-based Ni(Co)–W(Mo) catalysts are widely used in industrial processes due to their satisfying activity and low cost [11–16]. However, in order to meet the requirements of deep HDS conversion, new efficient catalysts with high activity and selectivity should be developed to remove the refractory sulfur-containing compounds, e.g., DBT and 4,6-DMDBT, from transportation fuels [17–21].Compared to other transition metals, noble metals and the corresponding alloys (Pt or PtPd alloy) show better hydrogenation properties in the DBT and 4,6-DMDBT HDS reactions [22–26]. Although noble metal catalysts display high catalytic performance in the hydrotreating process, they are relatively expensive and susceptible to sulfur poisoning even at very low sulfur contents feedstocks, which limit their practical applications in HDS reactions [27,28]. Therefore, the research of noble catalysts with sulfur resistance properties arouses more interests of scientists in different research fields. Many preparation strategies, such as constructing a core–shell structure or alloying with a second metal (Pd, Ru or Ir, etc.), have been performed to enhance the sulfur tolerance of noble metal catalysts [29–31]. For example, Pt noble metal can be encapsulated in the pores (around 0.5 nm) of zeolite only to allow the migration and dissociation of small hydrogen molecules (0.289 nm) and exclude the large sulfur-containing compounds [32–35]. This method protected the noble metal catalyst from sulfur poisoning and provided activated hydrogen for the HDS reaction [36]. Unfortunately, a fatal disadvantage of poisoning caused by the smaller hydrogen sulfide molecule (H2S, 0.362 nm) poisoning cannot be prevented [37]. Therefore, a critical problem to be solved is to isolate the H2S from the noble metal center, to achieve excellent sulfur-resistance performance in the presence of H2S. Besides, the noble metal catalyst with a high loading of noble metals will increase the production cost of the catalyst compared with those of the transition metal sulfide catalysts. As for noble metal catalysts, the approach to realize high HDS activity but cheap price would be the focus to explore.Herein, a new dendritic MoS2/Pt@TD catalyst consisting of sulfur-resistance Pt@TS-1 and MoS2 active phases was prepared. The pore-modified Pt@TS-1 seed with appropriate micropore sizes (between 0.289 nm and 0.362 nm) not only created activated H protons over the Pt noble metal sites anchored inside the internal channels of micropores, but also incentivized H+ spilling over to the surface of MoS2 active phases located in the external channels, which prohibited the direct contact between Pt noble metal and the sulfur-containing compounds (including H2S), consequently avoiding the poisoning of Pt active sites. Moreover, titanium element with various chemical states in the Pt@TS-1 can generate more d-electron, which is be beneficial to the creation of more S vacancies of MoS2 species. The dendritic MoS2/Pt@TD catalyst possessed the combining characteristics of noble metals of excellent hydrogenation performance at low temperatures and transition metals of low-cost. Thus, the as-prepared novel dendritic MoS2/Pt@TD catalyst showed superior activities and high sulfur-resistance stabilities for DBT and 4,6-DMDBT HDS.The synthesis procedure of Pt@TS-1 seed (molar ratios of SiO2/TiO2 = 25) is as follows. 40 g of tetraethyl orthosilicate (TEOS), 2.64 g tetrabutyl titanate, and 46.72 g TPAOH (25 wt%) were added to 8.68 g of deionized water. The mixture was stirred to dissolve for 1 h in an ice water bath, then switched to a water bath at 70 °C for 3 h. 48 g of isopropanol was added to the mix and stirred for 1 h. The resulting solution was labeled as solution A. 0.10 g of NaOH and 0.12 g 3-mercaptopropyltrimethoxysilane were added to 2.0 g of deionized water. 3.08 mL of chloroplatinic acid (H2PtCl6, 100 mmol L−1) was added to the mixture and stirred for 20 min. The resulting solution was labeled as solution B. Then the solution B was added to the solution A slowly, and continued to stir at 70 °C for 30 min. The resulting solution C was transferred to a 200 mL autoclave at 170 °C for four days. Finally, Pt@TS-1 seed was synthesized by centrifugation, washing, and drying.0.60 g triethanolamine (TEA) was dissolved in 220 mL deionized water and stirred at 80 °C for 30 min. Then 3.34 g of cetyltrimethylammonium bromide (CTAB) and 1.48 g sodium salicylate (NaSal) were added and kept stirring for 1 h. 35.2 g TEOS (28.4 wt% SiO2) was added dropwise and agitated for 30 min, then 2.0 g of the as-synthesized Pt@TS-1 seed was blended into the mixture solution D and kept continual stirring for 1.5 h. The resulting mixture E was transferred to a 500 mL autoclave at 100 °C for 4 h. Finally, Pt@TD catalyst was prepared after washing with water and ethanol, filtrating, drying at 80 °C for 12 h and calcining at 300 °C for 2 h. Pure TS-1/DMSNs (TD) composite was prepared using the same method without H2PtCl6.The micropore of Pt@TD catalyst was modified by using the deposition–precipitation method. 3.0 g of Pt@TD catalyst and a certain amount (0/0.126/0.396/0.816 g) of tetrabutyl titanate was added to 90 mL of ethanol. The mixture F was completely dissolved for 40 min by ultrasonator. Then the mixture was stirred slowly at 45 °C until all the ethanol was completely evaporated. Finally, a series of pore-modified Pt@TD catalysts were synthesized by adjusting the TiO2 amounts (0 wt%/1 wt%/3 wt%/6 wt%) incorporated into the system, and the products were noted as Pt@TD-xTi, which x represented the mass fractions of TiO2.Mo/Pt@TD-xTi catalysts with 12 wt% MoO3 and 0.1 wt% Pt were synthesized via incipient wetness impregnation method using ammonium molybdate ((NH4)6Mo7O24·4H2O). The as-prepared catalysts were denoted as Mo/Pt@TD-0%Ti, Mo/Pt@TD-1%Ti, Mo/Pt@TD-3%Ti and Mo/Pt@TD-6%Ti, respectively. The corresponding MoS2/Pt@TD-xTi catalysts were obtained after presulfidation of Mo/Pt@TD-xTi catalysts using 3 wt% CS2 in hexane. The reference catalysts of MoPt/TD and Mo/TD with 12 wt% MoO3 and 0.1 wt% Pt were prepared through two-step incipient wetness impregnation of TD support using (NH4)6Mo7O24·4H2O and H2PtCl6 (100 mmol L−1).X-ray diffraction (XRD) patterns were measured in wide-angle between 5°–50° and small-angle between 1.3°–6° at 40 kV and 40 mA in Cu Kα radiation. Nitrogen adsorption/desorption isotherms were acquired at −196 °C on a Micromeritics TriStar Ⅱ 2020 instrument. Scanning electron microscopy (SEM) was taken on a Hitachi SU8010 apparatus. Transmission electron microscope (TEM) images were recorded on an FEI F20 apparatus. CO temperature-programmed desorption (CO-TPD) and H2-TPD were carried out on a quartz reactor with a quadrupole mass spectrometer (OmniSTAR TM). X-ray photoelectron spectroscopy (XPS) was collected on a Thermo Fisher K-Alpha spectrometer.The activity of MoS2/Pt@TD-xTi series catalysts and the reference catalysts were investigated within a fix-bed reactor for DBT or 4,6-DMDBT HDS. 1.0 g Mo/Pt@TD-xTi catalysts precursors were loaded in the center of the reactor. Before the HDS reaction, the precursor was presulfided at 340 °C for 4 h under the conditions of 4.0 MPa, volumetric ratio of H2/Oil of 200 (v/v) and weight hourly space velocity (WHSV) of 8.0 h−1. After the presulfidation was completed, fresh H2 and liquid feed (DBT or 4,6-DMDBT) were switched into the reactor. DBT or 4,6-DMDBT HDS experiments were conducted at 340 °C, 4.0 MPa, 200 (v/v) and 10–100 h−1.Hence, the HDS conversions were determined by the following equation [38]: (4) HDS (%) = (S f  − S p )/S f  × 100% where S f is the initial concentration of liquid feed before reaction (ppm) and S p is the residual concentration of DBT or 4,6-DMDBT after reaction (ppm).The HDS of DBT or 4,6-DMDBT is assumed to follow the pseudo-first-order reaction kinetics, and the rate constant k HDS (mol g−1 h−1) can be presented as [39]: (5) k HDS = F m ln ( 1 1 − τ ) where F denotes the feeding rate of DBT or 4,6-DMDBT (mol h−1), m refers to the weight of catalysts (g), and τ refers to the HDS conversion of (%).Five characteristic peaks of Pt@TD-xTi series catalysts appear at the ranges of 7.8–8.7° and 22–25° in the XRD patterns as shown in Fig. 1 , which correspond to the (011), (020), (051), (303), (313), and (532) planes of TS-1 zeolite [40]. No discernible peak relating to Pt metal is identified, which is ascribed to the low loading and high dispersion of Pt. It is evident that peak intensities of Pt@TD-xTi series catalysts decrease in comparison of Pt@TS-1 seed, suggesting that Pt@TS-1 seeds are embedded into the skeletal structure of DMSNs.To better understand the Pt distributions in Pt@TD-6%Ti catalyst, HAADF-STEM, and EDS mappings were carried out to investigate the morphology and metal dispersion of catalysts, and the relevant photos are presented in Fig. 2 . The Pt@TD-6%Ti catalyst (Ultralow 0.1 wt% Pt content) shows no aggregation of bulk Pt particles, indicating that noble metal Pt particles are evenly distributed in the Pt@TD-6%Ti catalyst. In addition, Ti elements are also evenly distributed in the Pt@TD-6%Ti catalyst, demonstrating that Pt@TS-1 seeds are embedded into the skeletal structure of DMSNs.N2 adsorption–desorption isotherms and pore size distributions of the series Mo/Pt@TD-x%Ti catalysts, MoPt/TD and pure TD composite are presented in Fig. 3 . All the isotherm curves of the series Mo/Pt@TD-x%Ti catalysts, MoPt/TD and the pure TD composite display type-IV curves with H2 hysteresis loops. The hysteresis loops of the series Mo/Pt@TD-x%Ti catalysts and MoPt/TD became narrow after the addition of Pt and Mo active metals into TD composite. Moreover, the pore size distributions of the series Mo/Pt@TD-x%Ti catalysts and MoPt/TD are relatively discrete compared with pure TD composite.The texture properties of the series Mo/Pt@TD-x%Ti catalysts, MoPt/TD and the pure TD composite are presented in Table 1 . As shown in Table 1, the surface areas and pore volumes decrease from Mo/Pt@TD-0%Ti (407 m2 g−1, 1.09 cm3 g−1) to Mo/Pt@TD-1%Ti (399 m2 g−1, 1.07 cm3 g−1), then reduce to Mo/Pt@TD-6%Ti (367 m2 g−1, 0.95 cm3 g−1) as the TiO2 amounts increase. Notably, MoPt/TD catalyst shows the lowest surface area, pore volume and the smallest pore size among these catalysts. The sufficient surface area and pore size can provide a favorable environment for excellent dispersion of Pt and Mo active metal species’ and a fast diffusion for the reactants and products.The SEM images of the series Mo/Pt@TD-x%Ti catalysts and Pt/TS-1 seed are shown in Fig. 4 . All the series Mo/Pt@TD-x%Ti catalysts display similar uniform wrinkled morphology, indicating that deposition of TiO2 has little impact on the morphology of the series Mo/Pt@TD-x%Ti catalysts. The wrinkled surfaces of the series Mo/Pt@TD-x%Ti catalysts are tremendous advantages to the enhancement of the surface area. Besides, the series Mo/Pt@TD-x%Ti catalysts present no independent phases of Pt@TS-1 seeds, manifesting that Pt@TS-1 seeds are embedded into the skeletal structure of DMSNs successfully.The TEM images of the series Mo/Pt@TD-x%Ti catalysts and the pure TD composite are shown in Fig. 5 . All the series Mo/Pt@TD-x%Ti catalysts exhibit similar open dendritic center-radial pore channels as the pure TD composite, demonstrating that the dendritic pore structures are well maintained after the additions of Pt and Mo active metals into TD composite. The open dendritic pore structure facilitates the diffusion of reactants and products in the pore channels, which finally contribute to the improvement of catalytic performance of the HDS reaction.The H2-TPD data can be used to investigate the H2 adsorption capacity of the catalyst. The H2 adsorption capacity of the catalyst increases with the increase of the adsorption peak intensity. The adsorption peak area represents the adsorption amount of H2 molecules of the catalyst. The H2-TPD curves of the series Mo/Pt@TD-x%Ti catalysts are displayed in Fig. 6 . In addition, MoPt/TD and Mo/TD catalyst were selected as reference catalysts for the H2-TPD characterization. From the H2-TPD curve of Mo/TD catalyst, it can be seen that the H2 adsorption peak of Mo/TD catalyst is very weak, which indicates that adsorption of H2 by the Mo metal and TD composite is very weak. The H2 adsorption peak of MoPt/TD catalyst is also relatively weak, indicating that Pt particles on the MoPt/TD catalyst are too large and Pt metal dispersion is not even, resulting in the exposure of less H2 adsorption site. From the H2-TPD curves of the series Mo/Pt@TD-x%Ti catalysts, it can be seen that with the increase of TiO2 modification amount, the H2 adsorption amount decreases slightly but not significantly, which indicates that the modified micropores of Mo/Pt@TD-6%Ti can still allow the migration and dissociation of small H2 molecules.CO molecule (0.369 nm) and the H2S molecule (byproduct of the HDS reaction, 0.362 nm) show the similar kinetic diameters [41], indicating that if the CO molecule cannot enter into the micropores of catalyst successfully, then H2S molecule will also be denied entry to the micropores of the catalyst, consequently no access to the active sites of Pt clusters. The CO-TPD data can be used to investigate the CO adsorption capacity of the catalyst. The CO-TPD curves of the series Mo/Pt@TD-x%Ti catalysts are displayed in Fig. 7 . In addition, MoPt/TD and Mo/TD catalysts were selected as reference catalysts for the CO-TPD characterization. From the CO-TPD curve of Mo/TD catalyst, it can be seen that the CO adsorption peak of Mo/TD catalyst is very weak, which indicates that the adsorption of CO by the Mo metal and TD composite is very weak. The CO adsorption peak of MoPt/TD catalyst is also relatively weak, which may be because Pt particles on the MoPt/TD catalyst are too large and Pt metal dispersion is not even, resulting in the exposure of less CO adsorption site. From the CO-TPD curves of the series Mo/Pt@TD-x%Ti catalysts, it can be seen that with the increase of TiO2 modification amount, the CO adsorption amount decreases significantly, which indicates that most micropores of Mo/Pt@TD-6%Ti cannot allow the migration of H2S molecules from the products to the Pt site.It can be seen from the H2-TPD and CO-TPD characterization results that the Mo/Pt@TD-6%Ti catalyst can effectively prohibit the sulfur-containing compounds transferring into the inside of micropores due to the confined small pores after the Ti modification, thus avoiding the direct contact between the sulfur-containing compounds and Pt metals. At the same time, H2 molecules can diffuse in and out of the micropore channels freely on the Mo/Pt@TD-6%Ti catalyst and hydrogen dissociation occurs on the surface of Pt sites, of which activated hydrogen spills over and transmit to MoS2 active sites.To investigate and further understand the chemical surface state of MoS2 species on the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts, XPS measurements were performed and the relevant spectra are presented in Fig. 8 . The XPS fitting criterion is similar to the previous paper [42]. Table 2 displays the Mo 3d statistical results of the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts. As shown in Table 2, the sulfidation degrees of the series MoS2/Pt@TD-x%Ti catalysts have a little change compared with MoS2–Pt/TD catalyst, indicated that the existing state of Pt noble metal has a minor effect on the reduction and sulfidation of Mo transition metal.The HDS activities of DBT over the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts with different WHSVs are presented in Fig. 9 . The DBT HDS conversions decrease with the increasing WHSVs. The DBT HDS conversions over the series catalysts follow the sequence of MoS2–Pt/TD < MoS2/Pt@TD-0%Ti < MoS2/Pt@TD-1%Ti < MoS2/Pt@TD-3%Ti < MoS2/Pt@TD-6%Ti in the WHSVs ranges of 10–100 h−1. MoS2/Pt@TD-6%Ti catalyst displays the best DBT HDS conversion than other MoS2/Pt@TD-x%Ti catalysts and MoS2–Pt/TD catalyst.Further study on the product distributions shows that DBT reacts along direct desulfurization route (DDS) and hydrogenation route (HYD), and the corresponding HDS reaction mechanisms are shown in Fig. S1 (Supporting Information). Table 3 shows the DBT HDS product distributions of the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts. The HYD/DDS ratios of DBT over the series catalysts increase as the sequence of MoS2–Pt/TD (0.52) < MoS2/Pt@TD-0%Ti (0.69) < MoS2/Pt@TD-1%Ti (0.85) < MoS2/Pt@TD-3%Ti (1.04) < MoS2/Pt@TD-6%Ti (1.43), which is consistent with the trend of catalytic activity (k HDS). This result confirms that DBT mainly proceeds HDS via the HYD route, implying the HYD route made a considerable contribution to the total activity of DBT HDS over MoS2/Pt@TD-6%Ti catalyst.The HDS activities of 4,6-DMDBT over the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts with different WHSVs are exhibited in Fig. 10 . The 4,6-DMDBT HDS conversions decrease with increasing WHSVs values. The 4,6-DMDBT HDS conversions over the series catalysts increase in the sequence of MoS2–Pt/TD < MoS2/Pt@TD-0%Ti < MoS2/Pt@TD-1%Ti < MoS2/Pt@TD-3%Ti < MoS2/Pt@TD-6%Ti during the WHSVs of 10–100 h−1. MoS2/Pt@TD-6%Ti catalyst shows the highest 4,6-DMDBT HDS conversion than other MoS2/Pt@TD-x%Ti catalysts and MoS2–Pt/TD catalyst.In order to investigate the stability and sulfur-resistance performance of the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts, long-period (100 h) 4,6-DMDBT HDS reactive experiments over MoS2/Pt@TD-6%Ti and MoS2–Pt/TD catalysts were complemented. It can be seen from Fig. 11 , the 4,6-DMDBT HDS conversions over MoS2/Pt@TD-6%Ti catalyst were maintained at 92.1% at 100 h, In contrast, the 4,6-DMDBT HDS conversions over MoS2–Pt/TD catalyst decrease significantly, indicating that the Pt-confinement MoS2/Pt@TD-6%Ti catalyst possessed better sulfur-resistance performance and 4,6-DMDBT HDS catalytic stabilities.Further study on the product distributions shows that 4,6-DMDBT reacts along DDS, HYD, and isomerization (ISO) routes, the corresponding HDS reaction mechanisms are shown in Fig. S2 (Supporting Information). Table 4 exhibits the 4,6-DMDBT HDS product distributions of the series MoS2/Pt@TD-x%Ti and MoS2–Pt/TD catalysts. The ISO selectivities of the series catalysts increase in the sequence of MoS2–Pt/TD (17%) < MoS2/Pt@TD-0%Ti (24%) < MoS2/Pt@TD-1%Ti (29%) < MoS2/Pt@TD-3%Ti (46%) < MoS2/Pt@TD-6%Ti (54%), which is in agreement with the order of catalytic activity (k HDS). This result confirms that 4,6-DMDBT mainly proceeds the ISO route in the HDS reaction process, proving that the ISO route contributes more to the total activity of 4,6-DMDBT HDS over MoS2/Pt@TD-6%Ti catalyst.Dendritic Pt-confinement MoS2/Pt@TD-6%Ti catalyst with excellent sulfur-resistance performance and HDS catalytic stabilities were successfully prepared. MoS2/Pt@TD-6%Ti catalyst displays much better DBT and 4,6-DMDBT HDS activities than those of other MoS2/Pt@TD-x%Ti and the MoS2–Pt/TD reference catalyst. Advanced characterization and activity evaluation were conducted to interrogate the structure–activity relationship and the results are discussed as follows:Firstly, the modified micropore of Mo/Pt@TD-6%Ti can still allow the migration and dissociation of small H2 molecules (Fig. 6). H2 undergoes dissociation on Pt sites to produce activated hydrogen protons. Then H+ spills over to the MoS2 active sites on the mesoporous surface, which can enhance the HYD route of DBT HDS and the ISO route of 4,6-DMDBT HDS (Tables 3 and 4).Secondly, the modified micropores of Mo/Pt@TD-6%Ti catalyst possesses 0.289–0.362 nm in diameter, which can effectively confine Pt metal and prohibit the sulfur-containing compounds diffusing into the inside of the micropores (Fig. 7), thus avoiding the direct contact between the sulfur-containing compounds and Pt metal. Thus, Mo/Pt@TD-6%Ti catalyst can effectively prevent the sulfur poisoning of noble metal Pt, thus improving the sulfur-resistance performance and HDS catalytic stabilities (Fig. 11).Thirdly, the higher DBT and 4,6-DMDBT HDS activities of the MoS2/Pt@TD-6%Ti catalyst are the results of the synergistic effect from the strong H2 dissociation ability of noble metal Pt and the excellent desulfurization activity of transition metal sulfide MoS2.Fourthly, Pt noble metal’s existing state shows little effect on the reduction and the sulfidation of Mo transition metal (Fig. 8). The series MoS2/Pt@TD-x%Ti catalysts and MoS2–Pt/TD catalyst present almost the same sulfidation degrees, indicating that the higher DBT and 4,6-DMDBT HDS activities of MoS2/Pt@TD-x%Ti catalyst were mainly due to the stronger H2 dissociation ability and the excellent sulfur-resistance performance of pore-confinement Pt metal.Fifthly, the uniform wrinkled surfaces (Fig. 4) and the open dendritic pore structures (Fig. 5) of Mo/Pt@TD-6%Ti catalyst can improve the accessibility of DBT and 4,6-DMDBT reactants to MoS2 active sites and reduce the diffusion resistance of reactants and products in the DBT and 4,6-DMDBT HDS reaction process.Sixthly, Ti species of Pt@TS-1 seeds can provide spillover of d-electrons for the transition metal oxide Mo species on the surface of the mesopore, which can create more sulfur vacancies and facilitate the reduction and sulfidation of the transition metal oxide Mo species.The dendritic Pt-confinement MoS2/Pt@TD-6%Ti catalyst shows excellent sulfur-resistance performance and HDS catalytic stabilities in the DBT and 4,6-DMDBT HDS reactions. The MoS2/Pt@TD-6%Ti catalyst not only exhibits excellent DBT and 4,6-DMDBT HDS activity but also reduces the production cost. This new concept of combining H2 dissociation performance of noble metal catalysts with the desulfurization ability of transition metal sulfide MoS2 may open a unique perspective of the development of noble metal catalysts for the industrial HDS process of feedstocks with high sulfur contents.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 National Natural Science Foundation of China (No. 21808079, 21878330 and 21676298), Key Research and Development Program of Shandong Province (No. 2019GSF109115), the National Science and Technology Major Project, the CNPC Key Research Project (2016E-0707), the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award (No. OSR-2019-CPF-4103.2) and the Project of National Key R&D Program of China (2019YFC1907700).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.2020.10.012.

Metal confinement catalyst MoS2/Pt@TD-6%Ti (TD, TS-1/Dendritic mesoporous silica nanoparticles composite) in dendritic hierarchical pore structures was synthesized and showed excellent sulfur-resistance performance and stabilities in catalytic hydrodesulfurization reactions of probe sulfide molecules. The MoS2/Pt@TD-6%Ti catalyst combines the concepts of Pt-confinement effect and hydrogen spillover of Pt noble metal. The modified micropores of Mo/Pt@TD-6%Ti only allow the migration and dissociation of small H2 molecules (0.289 nm), and effectively keep the sulfur-containing compounds (e.g. H2S, 0.362 nm) outside. Thus, the MoS2/Pt@TD-6%Ti catalyst exhibits higher DBT and 4,6-DMDBT HDS activities because of the synergistic effect of the strong H2 dissociation ability of Pt and desulfurization ability of MoS2 with a lower catalyst cost. This new concept combining H2 dissociation performance of noble metal catalyst with the desulfurization ability of transition metal sulfide MoS2 can protect the noble metal catalyst avoiding deactivation and poison, and finally guarantee the higher activities for DBT and 4,6-DMDBT HDS.

The authors do not have permission to share data.Addressing climate changes and environmental sustainability while meeting the ever growing energy demand is one of the greatest challenges which our society needs [1,2]. Hydrogen has been recognized as one of the green fuels because of its high gravimetric energy density with zero emission [3,4]. To produce high quality hydrogen, electrocatalytic water splitting is one of clean and sustainable approaches [2–4]. Hydrogen evolution reaction (HER) is a cathodic half reaction of water splitting that reduces protons in acid or water molecules in alkaline to generate gaseous hydrogen, respectively [5,6].In particular, alkaline water electrolysis has improved stability over acidic electrolytes. Additionally, earth-abundant transition metals can be utilized as electrocatalysts which can further reduce the manufacturing cost [7]. HER mechanism in an alkaline electrolyte is generally described as following pathways of Eqs. (1)–(3) [8,9]. (1) H2O + e− → Had* + OH− (Volmer reaction) (2) H2O + e− + Had*→ H2 + OH− (Heyrovsky reaction) (3) 2Had* → H2 + OH− (Tafel reaction) Alkaline hydrogen evolution reaction is hindered by sluggish kinetics as a result of the accumulation of energy barriers of multiple elemental reactions [5,7,10]. Engineering the catalyst to alter the rate determined step is still a great challenge [6,7].Platinum (Pt) is known as the golden standard of HER electrocatalysts with satisfied reaction yield. However, vast commercial applications are limited by the prohibitive cost and insufficient reserves of Pt [4,11]. Pure nickel (Ni) has been considered as an alternative to Pt because of its low-cost, and abundant quantity [12]. However, Ni exhibits insufficient electrocatalytic activity and progressive deactivation toward HER which is caused by the weak ability to desorb OH- species from the surface and the formation of nickel hydride species [8,13,14]. To further enhance the HER performance, various Ni based alloys have been widely investigated [6,8,12,15]. The HER activity of Ni based binary alloys have been ranked in the following order: Ni-Mo > Ni-Zn > Ni-Co > Ni-W > Ni-Fe > Ni-Cr [16]. Based on this trend and synergistic effects of the two components, great effort has been devoted to the design and testing of Ni-Mo alloys [6]. Ni-Mo also has higher corrosion resistance in alkaline electrolyte, good electrical conductivity, and excellent thermal stability.Until now, many works designed new electrocatalysts based on different hypotheses and theoretical calculations. However, most works failed to demonstrate the enhancement due to the complexity of electrochemical reactions on heterogeneous surfaces.In this review, we focused on the latest development of Ni-Mo alloys for alkaline hydrogen gas evolution. We first summarize the reaction mechanism with key activity descriptors affecting the HER performance of Ni-Mo alloys in terms of both thermodynamics and kinetics. Then, we highlighted various approaches to improve the HER catalytic activity and stability including ligand and strain effects. By in-depth analysis of the prior works, it is our intention to deconvolute the complex nature of HER and pave the way to further enhance the performance. Finally, we present future perspectives and challenges are presented.In order to improve the electrocatalytic properties, it is essential to understand the causes of high thermodynamic barrier and sluggish kinetic [9]. The electrochemical conversion of water into hydrogen gas involves three elementary reaction steps: i.e., 1) water adsorption, 2) water dissociation and 3) hydrogen generation where Eads (water adsorption energy), Eact (activation energy of water dissociation), ΔGH* (hydrogen adsorption free energy which describes the binding strength of H* on the catalyst surface) are used as descriptors to evaluate each reaction step [7].The process of the adsorbed hydrogen into molecular hydrogen, i.e., the Heyrovsky or Tafel step, is the rate determining step which means that ΔGH*, is a main descriptor [17]. Fig. 1(a) shows that the result of the simple kinetic model plotted as ΔGH* and ΔEH depending on single crystals calculated as proposed by Nørskov and co-workers [18]. When ΔGH* is close to thermoneutral, the HER electrocatalyst has well-balanced hydrogen bonding and releasing properties as well as maximum exchange current density [7,17]. Thus, Pt is well known as the best catalyst with a ΔGH* value of − 0.09 eV (i.e., −0.33 eV of ΔEH) [11,18]. On the other hands, Ni forms too strong hydrogen bonds in accordance with a ΔGH* value of − 0.27 eV (i.e., −0.51 eV of ΔEH) [19]. (Fig. 1(a)) This is related to the weak desorption ability of OH- species on Ni surface, which leads to an essentially irreversible reaction by hindering the desorption reaction and further water dissociation by forming oxides and hydroxide phases [8,13,14].The electronic synergistic effect between Ni and adjacent heteroatoms, i.e., Mo, is generally described as leading to much better surface adsorption properties. The Ni-Mo system improves HER properties due to the excellent water dissociation ability of Ni atoms and the superior adsorption properties towards hydrogen of Mo atoms, based on hydrogen spillover process in which hydrogen species adsorbed on Ni surface transfer to Mo surface (i.e., H migration) [14,15,20–25]. Thus, Mo is considered as the promoter of H2 dissociation by the efficient filling of the d band [26]. According to the Engel–Brewer valence bond theory, the synergistic effect in HER of the material is expected when transition metals with empty or less-filled d orbitals (e.g., Mo) are alloyed with those having more-filled d orbitals (e.g., Ni) [27,28]. Highfield et al. reported the synergy effect in HER of Ni-Mo binary solid solutions formed by physical mixtures [25]. And it experimentally showed that Ni corrodes rather than evolves H2, especially in acidic electrolytes, and the Ni component associated with Mo appears more stable in HER process. It was presumed that the hydrogen trapping role of Mo protects from serious deactivation by impeding formation of Ni hydride, which influences the surface reconstruction and inhibits the further adsorption of H2O [25,26].Up to now, there have been significant efforts to synthesize Ni-Mo alloy with excellent electrocatalytic properties by varying composition to tune the electronic structure to maximize the synergy effect. In addition to the electronic structure changes, the crystal structure of Ni-Mo alloy alters with composition. For example, the addition of Mo atoms into the initial crystal structure of face centered cubic (fcc) Ni induces the local lattice expansion at low Mo content. However, the crystal structure alters to other forms when it exceeds the critical content [25]. Brewer-Engel predictions showed a multicomponent phase diagram of Mo with 3d transition series metals (i.e., Ni, Co, Fe, Mn and Cr) consisting of continuous isoelectronic curved lines (Fig. 1(b)) [20,28]. And each horizontal line between 3d metals (right side of transition series) and Mo (left side) indicate that the representative phase of alloy was changed when the atomic concentration of Mo exceeded a certain range. It has been noted the interdependence between crystal structure and electronic configuration [28]. In addition, it was predicted that the Ni3Mo phase could be the most stable alloy in the Ni-Mo system due to improved d-orbital overlap for hydrogen adherence and transference [28].The recent results accompanying the experimental evaluation of Ni-Mo electrocatalysts generally have been reported as the Mo incorporated Ni, Ni4Mo, Ni3Mo and NiMo phases and mixtures of several phases as good electrocatalysts. Pašti et al. experimentally investigated that Ni-Mo alloys synthesized by arc melting method exhibit the electrocatalytic activities corresponding to Mo contents [29]. The result indicated that the multiphase structure composed of NiMo and Ni4Mo phases shows highest HER activity. ( Fig. 2(a) and (b)) Jakšić et al. also studies the composition effect on HER activity by synthesizing Ni-Mo alloy powders using sol-gel method [30]. The volcano shaped activity-composition relationship indicated that the best catalyst has the structure dominated by Ni4Mo. (Fig. 2(c) and (d)) In addition, Wang et al. reported that Ni-Mo alloy/MoO3−x nanocomposites and Ni10Mo exhibited better HER activity than Ni4Mo and Ni3Mo phases [31]. (Fig. 2(e) and (f)) The calculated value of ΔGH* for Ni10Mo was − 0.27 eV which is smaller than Ni4Mo (−0.39 eV) and of Ni3Mo (−0.59 eV). This was mainly caused by the interaction between H and the composition-dependent Ni-Mo surface, as well as the amount of MoO3 sites with increasing annealing temperature. Geng et al. reported colloidal synthesis of Ni-Mo alloy nanoparticles with various compositions where Ni0.4Mo0.6 nanoparticles exhibited the highest HER performance [32]. (Fig. 2(g) and (h)) As seen from these reported data, the optimal composition of Ni-Mo which exhibit the best HER performance significantly differed. Therefore, it is difficult to simply correlation HER activity to the crystalline features and the electronic synergistic effect, but later other additional parameters need to be considered.ΔGH* based on volcano plot has emerged as a physical descriptor to understand HER. However, it does not capture the effect of other parameters such as interfacial interaction and interfacial reorganization at the interface between electrolyte and electrode[33]. It is worth discussing the interfacial engineering that assists water splitting into H+ and OH- with low barrier energy. Real-time spectroscopies and computation chemistry have been utilized to understand the underlying mechanism [33] where alkaline HER is less investigated than acidic HER [36].In general, the rate of HER in alkaline solution is approximately two orders of magnitude lower than in acid media due to slower the HO-H dissociation reaction [31,37,38]. The high affinity toward OH- could cause it to occupy active sites by the final products and block the consequent reaction, which can eventually affect higher HER overpotential [39]. Thus, the electron-coupled water dissociation, i.e. the Volmer step (H2O + e- → H* + OH-) has a strong influence on the HER kinetics in an alkaline solution. As important descriptors, water adsorption energy (Eads) and activation energy of water dissociation (Eact) can be used to investigate the ease of reaction at each step in the process of converting water molecules into hydrogen molecules [7].Until now, several groups studied the reaction kinetics of Ni-Mo electrocatalysts using DFT calculations followed by experimental verification. Zhang et al. synthesized Ni4Mo/MoO2 cuboids on nickel foam and investigated HER in alkaline solution [24]. The HER performance of these catalysts were comparable to that of Pt and superior to other earth-abundant electrocatalysts due to lower energy barrier for the Volmer step. The ΔG(H2O*) on Ni4Mo is significantly decreased to 0.39 eV, which is not only lower than the values of 0.91 and 0.65 eV for pure Ni and Mo respectively, but even lower than Pt (i.e., 0.44 eV) [24]. ( Fig. 3(a)-(d)) A smaller activation energy represents a faster water dissociation process. Shen et al. developed Ni4Mo electrocatalyst supported by graphene nanosheet. They contributed the superior HER characteristic of Ni4Mo alloy to a lower energy barrier (i.e., 0.39 eV) [15]. (Fig. 3(e)-(h)) Thus, the water dissociation process is also greatly facilitated on the Ni4Mo surface, which speeds up the sluggish HER kinetics under alkaline conditions. Li et al. recently reported the potential of Ni-Mo alloys as bifunctional hydrogen oxidation reaction (HOR) and HER electrocatalysts in alkaline solution [40]. Although HOR mechanism in alkaline solution is still under debate, they suggested that Ni4Mo alloy exhibited better energy profiles for hydrogen adsorption (ΔGH*, −0.09 eV) and the hydroxyl adsorption (ΔGOH*, −1.01 eV) than pure Ni (−0.30 eV and −0.09 eV, respectively). And the energy barrier for water formation was calculated to be 1.03 eV on Ni compared to 0.95 eV on Ni4Mo alloy [40]. Based on both theoretical and experimental investigations, they showcased great potential for Ni4Mo alloy as HOR and HER electrocatalyst in alkaline media. More recently, Gao and Yu et al. reported that Ni4Mo nanoparticles exhibited notable HOR reactivity, which has higher catalytic performance than that of commercial Pt/C [41]. In addition, Ni4Mo nanoparticles exhibited remarkable tolerance against surface poisoning by carbon monoxide (CO) gas. They attributed it to lower CO adsorption energy (∼ −1.45 eV) compared to Pt (∼ −1.68 eV). Additionally, preferential OH adsorption on the surface of Ni4Mo assists the oxidation of the adsorbed CO [41].Yang et al. further discussed the HER kinetics of Ni4Mo phase in terms of interfacial engineering. Ni4Mo nanodots on MoOx nanosheets were electrochemically deposited and their performance was evaluated [34]. The results showed that the Tafel slope of Ni4Mo/MoOx (64 mV/decade) is much smaller than Ni4Mo alloy (140 mV/ decade). Ni4Mo/MoOx has lower charge transfer resistance (3.7 Ω) than Ni4Mo alloy (6.9 Ω) at an overpotential (η) of 130 mV. It means that the interface consisting of metal/metal oxide affects the quicker water dissociation process by lowering the barrier and easy charge transfer. In addition, when Ni4Mo/MoOx was supported on Cu foam as a substrate, the η was 16 mV which is much lower than Ni4Mo/MoOx on Ni foam (∼ 50 mV). It showed that Cu as a substrate enabled amorphous MoOx to be more electron-rich at the interfaces, and it could facilitate water adsorption by the strong interaction with electron-deficient H atoms in water molecules [34,35]. And it is noted that the underlying layer in which Ni-Mo alloy forms an interface affects the HER performance. Ni(OH)2 is well-known as a promoter of water dissociation [42]. Yao et al. studied nanohybrid electrocatalysts consisting of NiMo alloy and Ni(OH)2 nanosheets on carbon cloth surfaces [43]. This electrode exhibited the η of 132 mV at 10 mV/cm2 and Tafel slope of 134.1 mV/decade, which was higher performance than NiMo alloy on carbon cloth (η = 146 mV and Tafel slope = 144.3 mV/decade). ( Fig. 4(a) and (b)) It was assigned by the hybrid structure of NiMo alloy and Ni(OH)2 nanosheets, which resulted in enhanced HER properties by improving the water dissociation rate. On the other hand, heterogeneous structures with nanoscale interfaces have been studied to better understand the effect of HER occurring at the atomic scale rather than the micron scale hybrid structures. Markovic et al. described that the oxophilic group (i.e., M(OH)2) on the surface of Pt could lead to an easier water dissociation [35]. It experimentally showed that the nanometer-scale Ni(OH)2 clusters on the Pt surface significantly enhance the HER activity, despite a 35% reduction in Pt exposed to the electrolyte. It suggested that the water molecule absorbs the Ni(OH)2/Pt interface, and then dissociates by O atom interacting with Ni(OH)2 cluster and H atom interacting with Pt [35]. And when Li+ cations are involved in the electrolyte, the anchoring of Li+ cations to the oxophilic group leads to their strong interaction with H2O and OHad in the electrolyte. (Fig. 4(c)) Subsequently, Markovic et al. reported that even on the surface of various transition metals, Ni(OH)2 nanoclusters have an effect on enhancing HER activity. Especially, Ni(OH)2/Ni electrodes have four times higher electrocatalytic activity in alkaline electrolytes than the Ni electrode without Ni(OH)2. (Fig. 4(d)) And this approach has been proposed to be extended to Ni-based catalysts [44]. Therefore, it is meaningful to develop a heterogeneous catalyst containing Ni-Mo alloy which creates optimal interface between metal and metal oxide/hydroxide in order to lower dissociation kinetic energy barrier [10].Ni-Mo alloy as an electrocatalyst for HER has been fabricated by diverse synthetic methods. The electrochemical deposition is used to uniformly synthesize materials on the surface of a support such as Ni foam or carbon foam, and NiMo or NiMoO phases can be formed depending on process condition. An et al. designed the Ni4Mo/MoOx nanointerfaces by controlling the precursors, pH and applied current density for electrochemical deposition, and Dung et al. studied the effect of complexing agents (i.e., ammonium and citrate) depending on pH and concentration of metal ion precursors which could affect the deposit mechanism [34,45]. And the hydrothermal and solvothermal methods are utilized to synthesize radially formed nanorods with large surface area on a support. Generally, NiMoO4 nanorods were grown by the reactions, and through subsequent annealing in the reducing H2/Ar atmosphere, the Ni4Mo nanocrystals were anchored on the MoO2 or MoO3−x phases. Zhang et al. showed superior electrocatalytic activity by ease electron transfer due to the high electrical conductivity of MoO2, and Chen et al. exhibited excellent catalytic performance by dual active components of oxygen-deficient MoO3−x as well as Ni4Mo [24,46]. In addition, the colloidal method is useful for synthesizing NiMo alloy nanopowders with a large specific surface area. The Mo content in NiMo alloys is easily controlled by processing conditions of the colloidal method, despite the limited solubility of Mo atoms in Ni lattices [32]. And Zhang et al. showed extraordinary performance for both the HER and OER with the optimized catalyst of Mo0.6Ni0.4 nanoparticles [32]. Furthermore, arc metaling is mainly used to synthesize intermetallic compounds of NiMo. Rößner et al. fabricated well-characterized surfaces of the single-phase intermetallic compounds Ni7Mo7, Ni3Mo, and Ni4Mo to investigate their intrinsic performance [47]. Therefore, NiMo catalysts can be synthesized based on various methods depending on the purpose and direction of utilization, and it is very necessary to design materials with high activity and high durability through various strategies discussed below.The surface science on electrocatalysts have been studied for enhancing HER activity. The chemisorption properties of the alloy system can be predominantly affected by the primary two effects which are chemical composition (ligand) and physical structure (strain) [48,49]. The individual effects contribute to the shifts in the metal’s d-band center which means to affect the strength of H* adsorption on the surface of a given material [11]. When the d-band center is close to the Fermi level, it tends to correspond to a strong H* adsorption [11]. Furthermore, it was suggested that it also influences the activation barrier for the bond breaking of molecules [11,19,50]. Thus, ligand and strain effects have a significant impact on thermodynamics and kinetics associated with the surface catalytic reaction of alloys, and would be considered for the rational design of electrocatalysts.Ligand effect occurs due to the formation of heteroatom bonds within first two monolayers, i.e., alloy system, and leads to the change in the electronic structure [11,51,52]. Until now, there have been few studies on the detailed HER mechanism of the Ni-Mo alloy system in relation to ligand effect excluding strain effect. Recently, Tian et al. studied the Ni-Mo alloy system of the cluster of NimMon (m + n = 5) based on the DFT method. It described the mechanism involving the initial adsorption of H2O and subsequent hydrogen evolution according to the clusters which are Mo, NiMo4, Ni2Mo3, Ni3Mo2, Ni4Mo and Ni [53]. It was suggested that Mo is the active site for H2O because of the electron deficiency of Mo by charge transfer from Mo to Ni, and Ni is involved in the adsorption of the split H of H2O after O-H bond breaking [53]. The adsorption energy of H2O on all designed clusters is about 20 kcal/mol, slightly different depending on the NiMo ratio. ( Fig. 5(a)) Among them, the Ni4Mo cluster has the smallest reaction barrier to OH bond breaking, i.e. 3.45 kcal/mol [53]. Furthermore, the reaction of Ni4Mo cluster spontaneously occurs under reduction conditions because the reaction barrier is small, i.e. − 0.27 kcal/mol. (Fig. 5(b)) It was demonstrated by the theoretical calculations how the bond formation between Ni and Mo heteroatoms affects the Volmer reaction step for HER performance, and excellent catalytic properties can be caused by lowering the high energy barrier of water activation.Strain effect occurs due to the alteration of the average length of metal-metal bonds in less than six monolayers, and potentially tunes the electronic structure. Guo et al. reviewed the strain-controlled multi-metallic electrocatalysts for the sake of achieving efficient energy conversion [54]. When the surface atoms of late transition metals (LTM, more-filled d orbitals) are subjected to tensile strain, it causes a stronger interaction with adsorbates [52,54]. For early transition metals (ETM, less-filled d orbitals) undergoing tensile strain, it lowers the adsorption energy [52,54]. On the other hand, compression strain has the opposite effect. The interaction is weaker for LTM and stronger for ETM [52,54]. In accordance with these arguments, it was attempted to induce strain effect using interfacial structure engineering (i.e. the core/shell structure, metallic overlays on substrates, de-alloyed shell) and structural defects (i.e. grain boundaries, multi-twinning) [11,51,52]. It is already known that the strain due to lattice expansion occurs as the Mo content in Ni increases, and the maximum solubility of Mo in Ni is less than 22–25 at% [55]. If it is more than that, it can be regarded as a deformed crystalline structure by the strain energy and causes the change of crystal structure depending on Mo contents [55,56]. Łągiewka et al. experimentally confirmed that each single phase of Ni-Mo alloys was synthesized in the range 0.5–23 at% of Mo contents via electrochemical deposition and presented the strain values through Rietveld refinement and Williamson-Hall’s method [56]. It showed that the lattice strain tended to increase with the Mo contents, and from 20 at% or more, it increased relatively further. And from the alloy with 10 at% Mo, the crystallite size decreased clearly. ( Fig. 6(a)) However, there has been no study on how the strain effect of Ni-Mo alloy affects HER according to the Mo content without ligand effect until now because Ni-Mo alloy system may simultaneously inherit the strain effect due to structural lattice expansion of the Ni unit cell by Mo insertion and the ligand effect due to the bonding between Ni as LTM and Mo as ETM. Thus, since the modulation of electronic structure as a dominant factor in HER activity is caused not only by strain effect but also by ligand effect, it is still a challenge to discover the mechanism by experimentally distinguishing the mutually concomitant two effects.Recently, Yang et al. studied the influence of the HER given of the elastic strain externally to Ni as a method to isolate strain effect from ligand effect [57]. It experimentally showed that compressive strain of Ni film causes higher current and lower overpotential, resulting in increased HER activity. (Fig. 6(b) and (c)) Conversely, the tensile strain leads to decreasing the catalytic performance. And based on DFT calculation, the ΔGH of Ni(111) were − 0.126 eV at about − 1.4% compressive strain and − 0.141 eV without strain, respectively. (Fig. 6(d)) It indicates that the increasing catalytic activity of Ni requires a weakening of H binding energy [57]. It seems that externally applied elastic strain can be considered as one of methods to understand the dominant strain effect in the Ni-Mo system.Chen et al. studied the electronic structure of Ni monolayers according to the supporting other transition metals based on DFT calculation [48]. When Ni monolayer on Pt (111) has a tensile strain (>10%), the d-band width of Ni monolayer is 1.35 eV, which is narrower than 1.89 eV pure Ni(111). It results in stronger adsorption energy on this surface due to the dominant strain effect [48]. Contrary to this, if the supporting metal of Ni monolayer is W(110), the d-band width is broader to 2.00 eV and lowers in adsorption energy. This is because the ligand effect works predominantly, even under similar tensile strain. In the case of Ru(0001) as the supporting metal, the d-band width is 1.76 eV, which is similar with Ni because of the balance between the narrowing of the d-band width due to strain effect and the broadening of it due to ligand effect [48]. It may indicate that the combined effect of the underlying support material on the Ni-Mo alloy catalyst should also be considered for controlling HER performance. Regarding the studies of Yang and Chen groups, the externally applied elastic strain and suitable underlying support materials seem to be considered one of the systematic fine-tuning methods of electronic structure on the Ni-Mo system for optimizing hydrogen adsorption energy and water dissociation energy.Although significant efforts were invested to determine the catalytic properties of Ni-Mo, limited investigations were performed regarding long term stability. For practical applications, it is essential to understand the fundamental aspects of material transformation during electrocatalysis to maximize stability. Until now, they have been reported that Mo might be dissolved in the form of highly soluble molybdate ions in the alkaline electrolyte, which will resulted in lowering electrocatalytic performance [47,58].Schalenbach et al. studied the intrinsic activity and stability of metallurgically prepared Ni-Mo alloy samples in alkaline electrolyte [58]. Bulk specimens were mechanically polished to enable clear distinctions from other factors such as interactions with the support, exposed surface area and morphology [47]. They analyzed Mo content in the electrolyte above anodic potential of − 0.15 V vs. RHE using an electrochemical flow cell coupled to a mass spectrometer. It reported that the alloys with higher Mo content resulted in greater Mo leaching and penetration depth. Selective Mo dissolution resulted in the formation of porous Ni structure such as hydroxide, oxy-hydroxides and oxide. Alternation of composition due to Mo leaching show lowering of electrocatalytic activity of Ni-Mo alloys compared to pure Ni electrode [58]. On the basis these results, they suggested that the high activity of Ni-Mo alloys might be attributed to the high surface area formed by Mo leaching, not inherent [58]. Armbrüster et al. investigated the stability of single-phase intermetallic Ni-Mo compounds using ex-situ XPS, roughness factors, electrochemically active surface area and corrosion current densities in KOH electrolytes [47]. They found that Mo in KOH electrolyte is leached from the intermetallic compound, which is more prone to form a higher Ni(OH)2 content on the near-surface region, resulting in lower surface specific activity [47]. They suggested that the crystal structure does not have effect on the catalytic activity in alkaline media and activity changes are a result of changing electroactive surface areas and the phase composition according to the degree of Mo leaching by the durability test conditions [47].Weckhuysen et al. utilized the density functional theory (DFT) to investigate the dissolution of Mo. They hypothesized that Mo is dissolved through the following reaction in alkaline medium;. (4) Mo + 2 OH- + 2H2O → MoO4 2- + 3H2 Which resulted in the creation of Mo vacancy to form segregated Ni phase. They also suggested that the rate of Mo leaching depends on the type of alkali cation, the reducing potential applied during electrolysis, and the substrate underlying the material [59,60]. For example, it was found that the alkali cations (i.e., K+, Na+ and Li+) infiltrate the surface of Ni-Mo alloys to form Mo vacancy where the Mo leaching rate was different due to the difference in the porosity formation through which the electrolyte infiltrated depending on the type of alkali cation [59]. Based to other cations, LiOH at pH 13 promoted the longevity of electrocatalyst by minimizing the Mo dissolution. Furthermore, it was shown that Mo leaching rate accelerated at high reduction overpotentials due to an increase in local pH. Although Mo leaching is depended on the applied potential, the leaching rate did not increased linearly with increase in applied potential [60]. Based on these observation, they hypothesized that multiple processes may occur that interfere with the oxidation of Mo to Mo6+ (in the form of MoO4 2-). Lastly, the catalytic performance of NiMo was highly dependent on its substrate [60]. The catalytic activity decreased in the substrate sequence of Cu > Ni > Stainless Steel >Ti, whereas the stability of Mo differed from this order. The exact cause is unclear, but in terms of stability of 1 M KOH, Ni-Mo deposited on a stainless steel substrate exhibited the lowest Mo leaching at low overpotential [60]. Overall, the electrocatalyst stability tends to be complicated and it is difficult to deconvolute the causes, so the Mo leaching mechanism is still under investigation. Therefore, further studies are needed to understand the destabilization mechanism and the major factors influencing the degradation of Ni-Mo electrocatalyst.Although long-term stability of electrocatalyst is essential for commercialization, only few works investigated the long-term stability of HER catalysts. For examples, Zhang et al. conducted that carbon plasma treatment improved catalyst stability by encapsulating the entire nanowires with a thin layer of graphitic carbon [61]. They explained the carbon shell effectively protects the active materials from dissolution in KOH solution through the XPS spectra of Ni and Mo before and after stability test as well as electrochemical stability test [61]. Peng et al. improved the stability of water electrolysis by coating on NiMo electrodes with chromium as a protective layer [62]. The chromium coating acted as a physical barrier to allow small molecules (e.g., H2) to pass through, while inhibiting the diffusion of oxygen by dissolved oxygen that could affect active site decomposition of the metal catalyst [62]. Recently, Zhai et al. reviewed latest progress on the long-term stability of HER electrocatalysts, and explained that dissolution of metal atoms, Osterwald ripening, agglomeration, particle detachment, active site poisoning and local corrosion could be the main reasons of catalyst deactivation [63]. And they suggested that the construction of binary transition metal (TM)-based compound materials (alloys, sulfide, phosphide, selenides, oxide, hydroxide, boride, and nitrides) is one strategy for improving electrocatalytic stability and activity.Facet engineering allows fine-tuning physicochemical properties [9]. The exposed active sites of Ni-Mo alloys for hydrogen evolution are also affected by structural and compositional engineering and cause to promote specific activity. Recently, Wang et.al described the H adsorption energies of all possible sites on Ni4Mo according to DFT calculation to approach a fundamental basis for HER activity origin [64]. The typical planes correspond to the detected Ni4Mo diffraction patterns of tetragonal structure, such as (101), (110), (121), (310), and (312), and have at least one strong preferred H adsorption position. Based on theoretical aspect, this indicates that all surface can exhibit catalytic capability and the most suitable adsorption site for H is on the facet (110) with − 0.272 eV of ΔGH* value. Zhang et al. experimentally demonstrated that increasing the fraction of high-index (331) facets of MoNi promotes faster HER kinetics [65]. The high-index MoNi facets were formed during topological transformation of Ni4MoO to MoNi. Compared to its pure counterpart, MoNi dramatically decreased the energy barrier to 0.56 eV in water dissociation step and 0.49 eV in hydrogen combination step, accelerating the sluggish alkaline HER kinetics. These resulted in a remarkable electrocatalytic performance with a Tafel slope of 33 mV/dec with excellent stability (i.e., upto 70 days). More recently, Lee et al. reported the H adsorption energies of Ni3Mo alloy with the low-index surfaces, which is considered to exhibit high catalytic activity. The Ni3Mo(101) showed ΔGH* of − 0.183 eV, which is closer to that of Pt(111) [66]. And the facet of (101) has 0.504 eV as the lowest energy barrier of water dissociation compared to the other facets [66] ( Fig. 7).These papers are meaningful by explaining the facet effect of the Ni-Mo alloy structure, which has been rarely addressed so far. However, the structural instability under high surface energy and the co-exposed crystal facets expressing unsatisfactory catalytic activity are still inherent problems when considering the facet engineering [9,67]. Therefore, in-depth research related to this is also a challenge.It is important to develop a material with a large catalytic surface area while facilitating electron transfer. Since the number of active sites is affected by the surface area of a given electrode, the nanostructured Ni-Mo alloys have been developed as catalysts. In this regard, the catalytic electrode must be able to promote electron transfer in order to make the most of its surface area. Gray et al. developed Ni-Mo alloy nanopowder and exhibited the overpotential of 70 mV at 20 mA/cm2 under a low loading of 1 mg/cm2 in 2 M KOH [68]. This electrode was fabricated by casting nanopowders suspended in isopropanol solvent onto a substrate without a binder, and showed a stable catalytic reaction for 100 h. The current density at constant overpotential (100 and 200 mV) increased with mass loading, indicating that a large amount of nanopowder could be utilized as active sites. ( Fig. 8(a)) Under a mass loading of more than a certain amount (10 mg/cm2 in this study), it may be difficult to maintain HER activity because nanoparticle films have a poor adhesion to the substrate and the transport of reactant species is weakened through porous films of increasing thickness [68]. Thus, easy electron transfer and effective charge transfer into the active surface should be considered simultaneously for enhancing catalytic activity. Recently, conductive carbon support has been studied with the aim of increasing electron conductivity for facial charge-transfer process and lead to favorable adsorption of intermediates by electronic interactions with metal [8]. McKone et al. studied the effect of resistive interface on HER performance, and ameliorate activity of Ni-Mo nanopowders by thermal hydrogen annealing process or adding carbon black as a conductive support [69]. (Fig. 8(b) and (c)) It was noted that a few nanometer-thin oxide layers on the catalyst surface could have an influence on catalytic activity as a result of electrical resistivity rather than being limited by kinetics. Thus, carbon-containing Ni-Mo nanopowders lowered charge transfer resistance and showed higher HER activity than that of thermal annealing processes. Sun et al. synthesized ultrathin 2D Ni-Mo alloy nanosheets directly on a conductive substrate to ensure robust contacts, which showed HER performance with the overpotential of 35 mV at the constant current of 10 mA/cm2 and the Tafel slope of 45 mV/decade in 1 M KOH solution [70]. (Fig. 8(d)) The charge transfer resistance was very low (0.8 Ω), which facilitated electron transfer. In addition, it exhibited smaller bubble adhesion force (∼ 2 μN) than other electrodes, meaning that the vertically aligned assemble structure could lead to much faster bubble release during HER. It showed that the excellent HER performance was attributed to fast mass transfer and easy electron transfer induced by the geometry of the catalyst grown directly on the conductive substrate. For a higher surface area for accessibility of water molecules, Ni-Mo alloys have been synthesized on the surface of 3D-structured conductive materials as the supports, including nanoform, nanorod, nanoporosity and hierarchical structure [46,71–77]. 3D architecture with its large surface area can help to manifest the superior catalytic properties inherent in the Ni-Mo alloy system by accelerating gas release.Interface engineering with surface decoration can modulate the adsorption/desorption energies of each species (e.g. H2O, OH-, H atom, and H2) to make more active catalysts [10]. In the kinetics section of this review, metal hydroxides on metal surfaces, in particular Ni(OH)2, have already been discussed as a promoter to lower the energy barrier of water dissociation. Jiang et al. developed the nanohybrid catalyst by integrating Ni4Mo nanoparticles with V2O3 network and observed that this surface was converted to Ni(OH)2 in the presence of water [78]. And it was evaluated in a neutral pH media for the purpose of utilizing abundant waste water or seawater sources [78]. Even in neutral electrolyte, Ni4Mo-V2O3 nanohybrids showed much better HER performance with an overpotential of 39.3 mV at 10 mA/cm2 and Tafel slope of 65.7 mV/decade, when compared to the Ni4Mo electrode (η = 60.5 mV at 10 mA/cm2 and Tafel slope = 103 mV/decade). ( Fig. 9(a)-(c)) The superior HER performance was attributed to the heterogeneous interface by the presence of oxophilic V2O3, which enhances water adsorption and promotes water dissociation into OH to form Ni(OH)2 [78]. Based on DFT calculation, the Ni4Mo-V2O3 exhibited a higher water adsorption energy (−0.51 eV) and a lower kinetic energy barrier for water dissociation (0.42 eV) than those of Ni4Mo surface, i.e., − 0.41 eV of G(H2O*) and 0.65 eV of ΔG(H2O). Liang et al. electrochemically synthesized NiMoO4-NiO-Ni composite films and evaluated their HER and OER performance in NaCl-containing alkaline electrolyte [79]. These composite films showed excellent catalytic properties compared to its pure counterparts with exceptional durability. They found that NiO promotes initial water dissociation and facilitates the conversion from H+ to H2 [79]. Further, the HER activities on these composite films were unaffected by the present of NaCl. Rather, the presence of Na+ and Cl- ions increased the solution conductivity which resulted in the reduction of IR drop. The NiMoO4-NiO-Ni composite films also exhibited good OER catalytic and corrosion properties. This work showcased that metal and metal oxide composite films can simultaneously promote HER and OER performance, and may offer the potential to utilize various water resources such as rainwater, seawater and wastewater as H2 feedstock. Furthermore, Qu et al. tuned the interfacial properties of NiMo nanoparticles by decorating small molecules of amine [80]. NiMo-EDA (ethylenediamine) required a much smaller overpotential of 72 mV at 10 mA/cm2 compared to 340 mV of un-modified NiMo. And Tafel slope of NiMo-EDA was 89 mV/decade, which was much lower than 135 mV/decade of NiMo. (Fig. 9(d)-(f)) The diamine-modulated interface of NiMo facilitated the HER process and enabled more efficient charge transfer [80]. It may cause to accelerate HER activity by an electron-rich surface due to the electron donation from the amino group to NiMo.Non-metal elements (e.g., C, N, S, O, P, B, etc.) are attempted to be incorporated into Ni-Mo system since non-metallic elements could altered the adsorption free energy of reaction intermediates and assist the fast water dissociation [9,81,82]. Transition metal phosphides (TMPs), transition metal nitride (TMNs), transition metal chalcogenides (TMCs), and transition metal carbides have been studied to further enhance the properties [83]. Most works reported that non-mental elements incorporated metals have better HER performance in acidic electrolytes. However, these materials still have a few challenges including catalytic durability due to acid fog inhibition, corrosion, and surface reconstruction [67,83,84].There are many strategies for optimizing the catalytic performance by heterostructure engineering, phase transition and doping [67,85,86]. Ren et al. studied Ni2(1−X)Mo2XP porous nanowire arrays as transition metal phosphides and showed outstanding HER activity in alkaline solution [81]. Ni2(1−X)Mo2XP catalyst yielded current density of 10 and 100 mA/cm2 at the overpotentials of 72 and 162 mV, respectively, and exhibited a Tafel slope of 46.4 mV/decade and long-term stability for over 160 h. This result outperformed Ni2P nanostructure with the η of 167 at 10 mA/cm2 and a Tafel slope of 89.6 mV/decade. Ni2(1−X)Mo2XP catalyst obtained 1077 mA/cm2 at overpotential of 300 mV, which revealed that stable hydrogen generation at high current density outperformed 566 mA/cm2 of commercial Pt wire. ( Fig. 10(a) and (b)) According to DFT calculations, Ni2(1−X)Mo2XP catalyst possessed low H2O activation energy (0.56 eV) and free energy for H adsorption (−0.08 eV) close to Pt, which imply easy water adsorption and optimized hydrogen adsorption/desorption capability. (Fig. 10(c) and (d)) Specifically, this computational result meant that Mo exposed surface on Ni2(1−X)Mo2XP catalyst has superior catalytic activity than that of Ni exposed surface on Ni2(1−X)Mo2XP catalyst as well as the pure Ni2P phase. Thus, it indicates that Ni2(1−X)Mo2XP catalysts adjusts electronic structure of pristine material by Mo substitution of Ni in Ni2P phase and the HER properties can be varied by the exposed atoms to the surface [81].Li et al. reported Ni-Mo-N catalyst composed metallic Ni and NiMo4N5 nanocrystals as transition metal nitrides. And it showed better HER performance of Ni-Mo-N catalyst compared to Ni-Mo catalyst in both alkaline and acid electrolytes [87]. In the alkaline electrolyte, Ni-Mo-N catalyst required η of 43 mV at current density 20 mA/cm2 and Tafel slope of 40 mV/decade, which superior to Ni-Mo catalyst (η = 86 mV and Tafel slope = 74 mV/decade). (Fig. 10(e)) Under the acid electrolyte, Ni-Mo-N catalyst exhibited η of 53 mV at 20 mA/cm2 and the Tafel slope of 39 mV/decade, outperforming Ni-Mo catalyst (η = 79 mV and Tafel slope = 61 mV/decade). (Fig. 10(f)) These excellent properties remained stable throughout the day in both electrolytes. (Fig. 10(g) and (h)) Sasaki et al. synthesized NiMoNX nanosheets on a carbon support (NiMoNX/C) and evaluated it in 0.1 M HClO4 solution [88]. NiMoNX/C catalyst revealed an onset potential of − 78 mV and Tafel slope of 35.9 mV/decade, which is much better catalytic performance compared to NiMo/C and MoN/C catalysts. In addition, NiMoNX/C catalyst maintained long-term durability, and the NiMo/C catalyst showed remarkable decrease in HER activity. (Fig. 10(i)-(k)) The results of the Li and Sasaki group's work provided that the introduction of nitrides into NiMo bimetallic structures improves catalytic performance and durability, which is evidence of improved corrosion resistance of Ni-Mo in acidic electrolyte. Recently, other efforts are being made to seek superior HER catalysts for neutral water splitting in order to utilize abundant water resources as H2 feedstock. The main challenges associated with HER in neutral solution is low HER kinetics similar to alkaline solution. Zhang et al. synthesized NiMoN nanowires array and investigated HER activity in neutral electrolyte with pH 6.8 consisting of 0.5 M Na2SO4 + 0.25 M KH2PO4 + 0.25 M K2HPO4 [89]. NiMoN nanowires exhibited the η of 46 mV at 10 mA/cm2 and a Tafel slope of 78 mV/dec. In addition, Ni nanoparticle decorate NiMoN nanowires showed lower η of 37 mV at 10 mA/cm2 with the Tafel slope of 51 mV/dec [89]. Based on these results, they calculated that the kinetic barrier for Ni/NiMoN (1.08 eV) is lower than NiMoN (1.69 eV) [89].Yu et al. synthesized Ni-doped Ni-Mo based sulfide (N-NiMoS) as chalcogenide systems and showed a good HER performance [90]. N-NiMoS catalyst possessed the η of 68 mV at 10 mA/cm2 and a Tafel slope of 86 mV/decade, which are much better performance than Ni-NiS catalyst, and displayed excellent long-term stability for 1000 h. Instead of single metal sulfides, bimetal Ni-Mo based sulfides exhibited abundant interfaces derived from nickel sulfides (NiS and NiS2) and molybdenum sulfides (MoS2), which can serve as excellent active sites [90]. In addition, it was confirmed that N doping into NiMoS affects the improvement of electron transport and electronic conductivity by tuning electronic properties. Fu et al. designed heterostructure composed of amorphous NiMoS and crystalline Ni(OH)2 and demonstrated meaningful HER performance over all pH ranges [91]. According to solutions with different pH ranges (acidic, neutral, alkaline, and seawater solutions), the reported η were 138, 198, 180 and 371 mV at 10 mA/cm2 and the Tafel slopes were 80, 81, 118 and 162 mV/decade, respectively. It was suggested that the enhancement of HER activities in all pH ranges were mainly attributed to amorphous NiMoS, which act as more surface defective sites than its crystalline counterparts and allows for fast electron transfer [91]. And Ni(OH)2 as a promoter for H2O cleavage accelerated the water dissociation kinetics in non-acidic electrolytes, which could improve the rate-determining step for HER. Thus, it means that the TMC-based catalytic activity, which largely depends on the amount of active site and the electrical conductivity of the electrocatalyst, can be improved by interfacial engineering with the design of the bimetallic compound [40,91]. It is essential to continue in-depth study in non-metal incorporated electrocatalysts, since it may lead to the development of sustainable, non-toxic, and large-scale processable electrocatalysts.Externally field-assisted electrocatalysis such as an electric field, a magnetic field, a light field, an ultrasonic field, or an electric field are being studied as strategies to improve catalyst performance, unlike the development of catalysts themselves [92,93]. External fields can control the local environment at the interface between the catalyst and electrolyte [93]. Elias and Chitharanjan Hegde studied the effect of magnetic field-enhanced HER on Ni-W alloy [94]. It was validated by increasing magnetic field strength in the range of 0–0.4 T, which clearly reduced overpotential and Tafel slope. ( Fig. 11(a) and (b)) The improved activity can be attributed to the magnetohydrodynamic (MHD) force-induced convection and H2 bubble release. In other words, the adhesion of hydrogen micro bubbles on the surface of catalysts can give rise to additional potential barriers for charge transport at the interface between catalyst and electrolyte. Thus, magnetic fields can accelerate electrolyte convection for easy elimination of hydrogen gas and alleviate polarization resistance [92,93]. Hourng et al. studied MHD phenomenon dependent on the electrodes by magnetic field [95]. Paramagnetic Pt, ferromagnetic Ni, and diamagnetic graphite were utilized as catalytic electrodes. When the magnetic field was applied, the current density increased by 14.6% for Ni and 10% for Pt, respectively, and the graphite changes were almost minor. (Fig. 11(c) and (d)) It showed that magnetic fields improve electrochemical hydrogen evolution and the ferromagnetic materials in particular can be preferred as good electrocatalysts under magnetic fields. Besides MHD, external fields have the possibilities to tune the inherent activity of catalysts [92]. However, the effects of various external fields on the Ni-Mo alloy system are rarely reported, a comprehensive understanding and further investigation of these strategies are required to obtain the field-boosted electrochemical reaction kinetics.For producing hydrogen as a clean fuel, intense efforts are being devoted to technological advances in water splitting in alkaline electrolytes. It is crucial to develop catalysts with essential requirements such as high performance, low cost, abundant resources and large-scale production, to replace Pt-based catalytic materials. In this regard, Ni, known as a major industrial metal, was also considered a promising candidate for alkaline HER. To compensate for Ni's insufficient catalytic activity and progressive deactivation, studies on Ni-based alloy materials are represented, of which Ni-Mo alloy is known to possess superior activity towards HER catalysis in alkaline electrolytes. The electrocatalytic performance of Ni-Mo based materials discussed in this review is summarized in Table 1 and simply plotted in Fig. 12 according to substrate type, surface material, structural difference, synthesis method and electrolyte considered important in catalysis system.In this review, we summarized the progress of Ni-Mo alloys from before to recently based on theoretical calculations and experimental evaluations to understand the mechanism affecting the improvement of HER performance. The hydrogen adsorption free energy (ΔGH*) and the activation energy for water dissociation (Eact) are presented as main activity descriptors to show the reaction pathway of Ni-Mo alloy based on thermodynamics and kinetics. The compositional feature of the Ni-Mo alloy system causes lattice expansion in the Ni unit cell, and changes in the cubic crystal structure of Ni above a certain amount. These behaviors lead to an influence on ΔGH* and tend to vary Eact by tuning electronic structure of Ni-Mo alloy, and finally affect the HER activity of Ni-Mo alloys. In addition, the interface formation between Ni-Mo alloys and other materials as underlying layers or between Ni-Mo alloys and water dissociation promoters as overlying layers gives rise to accelerate the slow HER kinetics under alkaline conditions, leading to improved HER activity. On the other hands, Mo can be selectively dissolved in the form of highly soluble molybdate ions in an alkaline electrolyte to from a porous Ni hydroxide, which eventually causes lowering the electrocatalytic performance. To minimize the Mo leaching during electrolysis, factors such as the type of alkali cation, the reduction potential applied during electrolysis, and the substrate underlying the material need to be optimized.Furthermore, we presented effective strategies for enhancing HER performance of Ni-Mo alloys as shown in Fig. 13, including i) optimization of ligand and strain effects, iii) stability, iv) facet engineering, v) large catalytic surface area with easy electron transfer, vi) interface engineering with surface decoration, vii) incorporation of nonmetal elements, and viii) externally field-assisted electrocatalysis. These strategies are aim at: (a) modulating adsorption/desorption energies each species, (b) control of environmental factors hampering Mo leaching, (c) tuning physicochemical properties of exposed active sites, (d) facilitating electron transfer and charge transfer into all actives sites on a large surface area, (e) leading abundant active sites and facile gas releasing, and (f) controlling the local environment at the interface between the catalyst and electrolyte. Thus, these efforts lead to enhancing intrinsic HER activity and improving stability and durability of pristine catalysts. In addition, it can provide the possibility to utilize various water resources as H2 feedstock regardless of pH ranges and expandability into bifunctional electrocatalyst. Although this review has dealt with the specific alloy as an electrocatalyst in detail, it will be helpful to provide a comprehensive understanding of how the intrinsic and extrinsic activities in other alloying catalysts change based on theoretical studies and various experimental approaches and how they relate to the final HER performance.Up to now, the remarkable progress of Ni-Mo alloyed electrocatalysts for alkaline HER has been studied. However, there are still issues to consider because the experimentally verified catalytic activities are not sufficient to satisfy the expected results based on various experimental approaches and computational science for enhancing HER properties. And the low electrochemical stability of electrocatalyst and low feasibility of large-scale production of lab-developed materials are a major constraint on commercial survival. Other catalyst materials also encountered this challenge.To begin with, it is necessary to consider how the correlation accumulated from the substrate to the surface affects the active surface of the final synthesized electrocatalyst. Although effective approaches are studied to improve HER properties, the demonstrated results are influenced by material and structure of the underlying layer of active sites. And the environment, such as strain effect and external applied field, given to the interface between the underlying layer and the electrocatalyst’s active surface also affects HER performance because of modulating the electronic structure of the active surface.Second, the overlying materials on the active surface of the electrocatalyst require further investigation. The water dissociation promoters on the active surface play a major role in enhancing slow HER kinetics in alkaline electrolyte. And protecting metal catalysts from corrosion on the active surface is a crucial issue for HER stability. Until now, studies on Ni(OH)2 for promoting water dissociation and carbon materials for protecting corrosion reaction have been conducted as the overlying materials on the active surface, but more candidate materials and related studies are needed to solve the fundamental cause.Third, continuous multi-disciplinary efforts are needed to understand the direct correlation between theoretical calculations and electrochemical stability, and to uncover controversial catalytic reaction mechanisms based on in-situ analysis. It also requires an overall discussion of the inherent uncertainties in the DFT calculation is necessary. Recently, Liberto et al. raised current problems related to predicting catalytic activity, such as accuracy of calculations, ignoring important contributions in the model used, and the physical meaning of descriptors and reproducibility [102,103]. And they explained that there are no tools to predict whether it will be possible to synthesize a new catalyst, even if a catalyst is designed. Therefore, it is known that the electronic structure theory aimed at predicting the catalytic activity of a material can be an important guideline for experiments or described experiments results more logically, but the rational design of catalysts based on theoretical calculations requires in-depth discussion and attention.Fourth, there is a need for a feasible synthesis method that can produce a catalytic material developed in a laboratory on a large scale. Although interesting HER catalysts with outstanding performance have been reported so far, simple synthetic methodologies should be developed for large-scale production and ease to process.Finally, efforts on standard and systematic evaluation methods are required to accurately compare the developed electrocatalyst’s performance. It is necessary to establish the essential evaluation methods that represent HER activity and the stability of electrocatalysts, especially their ability to generate stable hydrogen at high current densities for commercial viability. In addition, the criteria for factors influencing HER performance such as Pt as a reference electrode and electrolyte as reference solution during measurement should be discussed to facilitate comparative evaluation with each other.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.SHP acknowledges the support from Characterization platform for advanced materials funded by Korea Research Institute of Standards and Science (KRISS – 2021 – GP2021-0011). DT and NVM acknowledge the financial support from the Hydrogen Energy Innovation Technology Development Program of the National Research Foundation of Korea (NRF-K) funded by the Korean government (Ministry of Science and ICT (MSIT)) (NRF-2019M3E6A1064020).

Hydrogen evolution reaction (HER) from alkaline electrolytes is one of most promising methods for producing hydrogen. The remaining obstacles include the development of high performance and earth-abundant electrocatalysts which can be cost-effectively fabricated in large-scale. Ni-Mo based materials are one of potential candidates which might meet the most needs. In this review, we summarize the latest progress in Ni-Mo based HER catalysts in alkaline electrolytes from theoretical calculation to experimental results. This work also summarizes several different strategies to enhance the HER rate. Finally, the future perspective of the next generation electrocatalysts is discussed.

Sea-level rise and climate change are among the severe effects of the global energy crisis and greenhouse gas emissions. Therefore, research on renewable energy is fundamental and urgent [1–3]. Among new energy systems, hydrogen energy is considered a promising secondary energy source with a higher calorific value. H2 can be stored in high-pressure tanks in gas and liquid phases. It can also be reversibly absorbed and released by solid hydrogen storage materials. Moreover, as an excellent energy storage medium, it could realize high-efficiency conversion and reuse between hydrogen and electricity through water electrolysis and fuel cell technologies. Furthermore, H2 has a strong grid connection potential. It can convert wind energy, photovoltaic solar energy, and other renewable energy power, which have the characteristics of seasonal intermittent and regional limitation, into hydrogen energy, thus optimizing energy storage and transportation [4].Therefore, it has recently attracted much attention as a potential renewable energy source [5]. However, up to now, nearly all H2 is produced from fossil fuels, mainly using natural gas and coal as raw materials. To achieve the Paris Agreement's goals, H2 production should be effectively transformed from traditional fossil-fuel-based ways to the renewable-energy-based scenario [6]. In this context, H2 production by water electrolysis using renewable energy has been promising because of its environmental friendliness, abundant water resources, and high hydrogen purity (≈99.999%) [7–9]. It could also integrate other renewable energies into the grid and create a substantial downstream market [6].According to electrolyte and operational temperature, water electrolysis technologies could be classified as alkaline water electrolysis, proton exchange membrane water electrolysis, anion exchange membrane water electrolysis, and solid oxide water electrolysis. Alkaline water electrolysis is considered a suitable technology for the large-scale synthesis of hydrogen because of the usage of non-precious metal catalysts [10,11]. It is well-established as the most commercially viable and applicable technology with numerous vendors [12,13].The overall water electrolysis is classified into two half slow-rate electrochemical reactions, including the HER and OER [14,15]. Compared with non-precious metal electrocatalysts, precious metals-based materials such as Ir/Ru compounds for OER and Pt for HER are generally engaged as electrocatalysts for acid–media reactions to promote sluggish electrochemical activities [16,17]. However, using such relatively scarce and precious catalysts is not advisable because it significantly raises the cost of H2 manufacturing and severely impedes the expansion of productivity on an industrial scale in the future, thus necessitating the development of non-precious metal alternatives with high performance [18,19]. Currently, reported non-precious metal electrocatalysts mainly include sulfides, selenides, phosphides, carbides, nitrides, etc. [13,20–38] Transitional metal phosphides are potential candidates for HER due to the hydrogenase-like catalytic mechanism. However, hydrogen bonds strongly to nickel-based phosphides, which necessitates optimizing the free energy and electronic structures by rational strategies such as heteroatom-doping and heterostructure construction [37–41]. Transition metal selenides possess low intrinsic electrical resistivity. The 3 d orbitals of transition metal selenides bond with metal atoms because the energy level is close to 3s and 3p orbitals, with higher metallicity. The main limitation of TMSs is the deficient surface active sites. Several strategies have been proposed to increase active sites and modify catalytic activity [33,42–45]. Transition-metal nitrides have emerged as an attractive class of HER catalysts due to their high electrical conductivity and noble-metal-like characteristics, while the performance presents obvious variation. The mechanism for interpreting variation remains unclear, notably the modulation principle of metal centers [46,47]. Recently, great progress has been made on transition metal sulfides. Among them, layered transition metal sulfides such as molybdenum disulfide (MoS2) and tungsten disulfide (WS2) have been studied intensively due to the near-zero energy barrier for HER on edges [34–36,48]. However, the low density of active sites, poor electrical transport, and inefficient electrical contact may hinder further development. Among transition metal electrocatalysts, nickel sulfides, including NiS, NiS2, and Ni3S2, have been identified as ideal candidates for OER and HER catalysts due to the low price, abundant resources, simplicity of preparation, noble metal-like electrocatalytic properties, and electronic configuration [49,50]. It has been reported that Ni3S2 exhibits better HER and OER electrocatalytic performance than NiS and NiS2 [51–53]. Both experimental results and theoretical calculations show that the good catalytic properties of Ni3S2 could be ascribed to its inherent metallic conductivity, the abundance of active sites, and the optimal Gibbs free energy for catalyst-H∗ for HER. Furthermore, because of the more fraction of Ni, catalytically active species of nickel (oxy)hydroxides are formed on the surface of Ni3S2, resulting in improved OER performance [52,54–56].Up to now, there have been many works to review the advances of electrocatalysts for water electrolysis. However, systematic summaries of Ni3S2-based materials are few. Herein, the review of recent advances in Ni3S2-based electrocatalysts is presented. It begins by briefly introducing the mechanisms of HER and OER in alkaline solutions and the performance evaluation parameters of electrocatalysts. Then, the synthesis technologies and performance improvement strategies are summarized. The applications of Ni3S2-based electrocatalysts for industrialized alkaline water electrolysis are discussed afterward, especially for the bubble behavior control and surface wettability construction strategies. Finally, an outlook on future challenges and opportunities for Ni3S2-based electrocatalysts is presented, and potential future directions are also proposed.The HER process in alkaline solutions follows the Volmer–Heyrovsky or Volmer–Tafel step, as shown in equations (1)–(3).Volmer step: (1) H 2 O + e − → OH − + H ∗ Heyrovsky step: (2) H 2 O + e − + H ∗ → H 2 + OH − Tafel step: (3) 2 H ∗ → H 2 The Volmer step explains the splitting of water molecules and hydrogen adsorption. The subsequent stage involves the generation of hydrogen by the Heyrovsky process and/or the Tafel process. The HER mechanism comprises the Volmer process followed by a parallel Heyrovsky and Tafel step at low overpotentials, and the reaction follows the Volmer–Heyrovsky pathway at high overpotentials [57,58].Compared with the HER in acidic solutions, the HER is more challenging to achieve at the low overpotential in alkaline solutions. In acidic solutions, only the hydrogen bonding energies determine the reaction rate, while in alkaline solutions, the reaction mechanism is decided by water adsorption and dissociation, hydrogen adsorption/desorption, and hydroxyl ions affinity to the electrocatalyst surface (poisoned catalysts by occupying active sites) [11,57,59–61].Water absorption is the first step of HER. Compared to the situation in acid media, where the number of protons is much more than that in the alkaline media, the additional water decomposition step is indispensable, leading to another energy barrier. Beak et al. synthesized a ruthenium-based nanoparticle catalyst dispersing within a nitrogenated holey two-dimensional carbon structure (Ru@C2N) [62]. This catalyst improves the binding energy of Ru–H2O significantly, which accelerates the Volmer reaction rate. Moreover, because of the shortage of protons, water dissociation is an indispensable step of HER in alkaline media, slowing down the reaction rate. According to the Sabatier Principle, a good balance between the adsorption and desorption on the electrocatalyst surface is the optimal condition for forming a suitable intermediate bond strength (not too strong and not too weak). The relations between the HER current density and metal-H (Me–H) bond strength could be described using a “volcano” diagram. Fig. 1 g depicts the situation of pure metals, where the bonding energy of Ni is closer to that of precious metals, indicating that Ni is a suitable candidate for water electrolysis. Owing to the intrinsic Ni–Ni metal network, Ni3S2-based materials exhibit metallic-like properties, which implies that they have good electrocatalytic activity.Finally, the aqueous OH− is also essential in influencing the kinetics, even though its role is still debated. Some studies suggest that OH− and Had adsorption is competitive on one single active site, so the Volmer step's kinetics could be improved by creating dual active sites to host OH− and Had, respectively [63]. Nevertheless, others think the adsorption of OH− does not participate in the alkaline HER process, nor does it affect HER's kinetics. Recent studies suggest that the ‘dual active sites’ improve from hydrogen bonding's optimal energy bonding [64]. Therefore, those doubts can be solved after conducting further in-depth studies with a detailed consideration of intermediates' interactions.Compared with HER, the OER in alkaline solutions has a different and complex process because it involves four electrons transfer. It is regarded as the bottleneck constraining the energy efficiency of alkaline water electrolysis due to its sluggish reaction dynamics. All proposed OER mechanisms start from the same imperative step of hydroxide coordination to the active site, followed by varying proposed steps. Two broadly acknowledged OER processes include the classic absorbed evolution mechanism (AEM) and the novel lattice oxygen-mediated mechanism (LOM).In alkaline solutions, the AEM process comprises numerous electron-proton coupled steps in which OH− is oxidized into molecules of oxygen and water [65]. The reaction pathway could be characterized as follows: (4) OH − + ∗ → HO ∗ + e − (5) HO ∗ + OH − → O ∗ + H 2 O + e − (6) O ∗ + OH − → HOO ∗ + e − (7) HOO ∗ + OH − → ∗ + O 2 + H 2 O + e − where ∗ denotes active sites, O∗, OH∗, and HOO∗ represent adsorbed intermediates on active sites [66]. Firstly, OH− originated from the electrolyte is absorbed by an active site, followed by M−O generation with the removals of a water molecule. Then, an M−O is converted to an M-OOH intermediate by coupling a hydroxyl anion under one-electron oxidation. After that, an M-OOH gets coupled with a hydroxyl anion, generating an O2 molecule and an initial active site under one-electron oxidation. The other approach is the linking of M−O to generate O2, but this path has a higher thermodynamic barrier than that of the pathway involved in the reaction (6) and reaction (7). In addition, previous studies have also detected the existence of the M-OOH intermediate, expediting the approach by the steps from (4) to (7) as the more general OER pathway [67,68]. Significant attempts have been made to link OER activity with a particular descriptor to understand the OER active site and further forecast active OER electrocatalysts. In 1955, the M−OH bonding energy was first found to be related to the OER activity, where the lowers in an almost linear relationship with the M−OH bonding energy rise [69,70]. With the development of molecular orbital theory, the 3 d electron number of the B site is claimed to have a linear relationship with the catalytic ability for OER in ABO3-type perovskite oxides [71]. The enhanced catalytic ability could be attributed to decreased M–OH bond strength with increasing d-band electrons. Recently, much progress has been made in correlating the OER activity with descriptors. The volcano-type relationship between eg electrons and OER activity is proposed on the perovskite oxides, which is different from the previous linear description of d band electron number [72]. Oxides with eg occupancy close to unity could attach moderately with oxygen, resulting in the ideal OER activity [73].In addition to the AEM, the LOM has been considered the new pathway for OER in alkaline solutions. It is possible that direct O–O bonding with the reversible production of oxygen vacancy (LOM) may become favorable as covalency increases. This is because the capacity of metal cations to bind with oxygen will become weaker as covalency increases. Some studies have reported that the LOM is directly observed to enhance the OER activity by in-situ characterization techniques and density functional theory (DFT) calculations [74–77]. The lattice oxygen could participate in the OER reaction, resulting in the generation of oxygen molecules via the production of oxygen vacancies, as shown in Fig. 1e. For highly covalent oxides, the OER on oxygen sites could be initiated when the Fermi level is pinned to the top of the O 2p band, resulting in electronic states near the Fermi level with significant O 2p characteristics. Furthermore, it is observed that OER activities increase with rising pH for perovskites with metal–oxygen covalency, demonstrating that the activation of lattice oxygen redox processes is linked to non-constrained proton-electron transfer steps [74,78,79].Before delving into the advances in Ni3S2-based electrocatalysts in alkaline water electrolysis, it would be advantageous to go over the screening parameters and highlight what one should expect while screening HER and OER electrocatalysts. In this part, some vital performance evaluating parameters that have been well-accepted are summarized.The extra potential (beyond the thermodynamic requirement) needed to drive a reaction at a certain rate is defined as overpotential [83]. Overpotential could be presented as equation (8) (8) V o p = V e q + η a + | η c | + I R V op is the overall potential, and V eq is the equilibrium potential (1.23 V for water electrolysis). The η a and η c are OER and HER overpotentials, respectively. IR is the overpotential caused by Ohmic impedance. The overpotential of OER serves as the primary barrier due to the four-electron transfer process. Overpotentials at specific current densities are frequently utilized as quantitative parameters for evaluating the catalytic ability of electrocatalysts [11,84–86]. The overpotential required under the current density of 10 mA cm−2, which is the current density to match the 12.3% efficiency for photoelectrochemical water electrolysis, has been considered an important indicator to evaluate different catalysts. Still, it cannot determine the intrinsic activity [87,88]. In cases of electrocatalysts for commercial applications, higher current densities, such as 200 and 400 mA cm−2 could also be taken into account [12,89]. For Ni3S2-based catalysts, recent studies have adopted the overpotential at a higher specific current density, such as 200 mA cm−2, 500 mA cm−2, and 1000 mA cm−2, to evaluate the commercial application ability [2,90–92].The areal, mass, and specific activities are perspectives of the thermodynamic activity parameters. The areal activity is calculated by taking the geometrical surface area of the electrocatalysts, the mass activity is ascertained by taking the mass of the loaded catalyst and normalizing it, and the specific activity is obtained by taking either electrochemically active surface area (ECSA) or the Brunauer–Emmett–Teller (BET) for normalization. Mass activity is often evaluated on comparable material systems and is heavily influenced by the active area of the catalyst. Catalysts with larger surfaces frequently have higher mass activity. Due to the much lower cost of 3 d transition metals, the mass activity is less essential than the other characteristics. Among these parameters, the specific activity is the most reliable one when using ECSA rather than the BET to normalize [93]. The gas adsorption/desorption sites utilized to calculate the BET may not be electrochemically active.The Tafel slope, j0, and TOF are three kinetic activity characteristics [91]. The well-known Butler–Volmer equations imply that lower Tafel slopes and higher j0 are desirable, especially for alkaline water electrolysis at high current density [94]. When the overpotential is neither very small nor not very large, it is often simplified using the Tafel equation. The Tafel equation could be expressed as equation (9) (9) η = a + b × log(j) Where the b is the Tafel slope. The j0 is calculated by extending the Tafel plot's linear fit to its intersection with the corresponding logarithmic current density at the electrocatalytic study's reversible potential. The TOF measures how many products each active site in the reaction produces each second. For water electrocatalysis, TOF values are always reported as a function of the overpotential. There are several equations for TOF calculation. The most commonly used one is shown as equation (10) [95]. (10) TOF = j × N A / ( F × n × Γ ) Where the j, NA, F, n, and Γ indicate current density, the Avogadro constant, the Faraday constant, the number of electrons transported to generate one molecule of the product, and the precious number of active sites, respectively. The n is equal to 2 for HER and is equal to 4 for OER. Some researchers would also calculate the TOF of OER by concurrent oxygen reduction reaction (ORR) at the ring electrode utilizing a revolving ring disk electrode assembly [96].The Tafel slope is a primary evaluation perspective to evaluate the inherent kinetics and determine the mechanism of electrocatalysts [97]. The Tafel slope of 29, 38, or 116 mV dec−1 indicates that the HER rate-limiting step is Tafel, Heyrovsky, or Volmer step, respectively [98,99]. The kinetics occurring in the interface could be ascertained from the Tafel slope. Several approaches to acquiring the Tafel slope have been described in past studies [100,101]. The Tafel slope and j0 could provide many insights into the kinetic behavior of the electrochemical process. TOF is always tricky to measure accurately because the number of active sites participating in reactions is difficult to ascertain, and the assumption made during the calculation primarily affects the results [91,94]. Several methods used to calculate the TOF have been adopted in experiments, but they all have drawbacks. [95] Recently, a systematic path for calculation has been proposed to improve the accuracy of the calculation of TOF values. The relatively accurate calculation of TOF values should first determine the exact number of active sites and then exclude the error originating from other current normalization methods, including geometric area normalization and mass normalization. After that, errors for catalysts with FE <100% should also be excluded [102].FE is the efficiency of transferring electrons transported by the external circuit to the electroactive species to generate products [85]. Two methods are widely employed to determine the FE of HER and OER reactions. The first is based on the rotating ring disc electrode (RRDE). This instrumental technique only determines the intrinsic of OER catalysts. The equation calculating the FE is shown as equation (10) FE = (IRnD) / (IDnRNCL) (10) n R and n D represent the number of electrons transferred at the ring and disc. I R and I D represent the currents at the ring and disc, respectively. N CL represents the collection efficiency of the RRDE. The other method calculates the generated gas (H2 and O2) by the chronoamperometric or chronopotentiometry analysis. Subsequently, the gas amount may be determined using the water gas displacement technique, gas chromatography, and the spectroscopic technique [103]. If the calculated and collected volumes correspond, the catalyst's FE for the specific gas evolution process is 100 percent, showing that it is selective for electrochemical reactions. It assures that the supplied energy would not be wasted on side reactions. Other electrochemical events would also not contribute to the gas evolution current [89]. For Ni3S2-based electrodes, electrocatalysts are expected to have a FE of 100%. If the proportion of FE falls between 90% and 100%, the OER is deemed to be satisfactory [91,104].The Ni3S2 possesses a rhombohedral crystalline structure. Each nickel atom is in a pseudo-tetrahedral site in a sulfur lattice that is about body-centered cubic. Ni3S2 units are connected with a short distance of Ni–S (2.29 Å) and Ni–Ni (2.53 Å), which shows apparent metal–metal bond interactions. According to the calculation results, the dominant interactions and contributions of Ni3S2 are Ni(d), with little contribution from S(S) orbitals. Ni3S2 exhibits metallic behavior, with the majority of its orbitals crossing the Fermi level, as shown in Fig. 2 d and e [90,105,106]. This metallic property could be explained accorded to previous reports: when all the S atoms in the Ni3S2 bulk are removed, the remaining Ni–Ni structure could be regarded as a crystal, and 4 Ni–Ni bond modes appear for each Ni atom in the structure. This forms a continuous Ni–Ni bond network throughout the crystal. Therefore, this structure is basically more conducive to the transport of electrons, presenting good metallic properties [49,107]. Besides, its low cost makes Ni3S2-based materials more competitive as alternative electrocatalysts for alkaline water electrolysis. In addition, the high electron density around metal sites could enhance OER activity. The stability of Ni3S2 is high under alkaline solutions because of the formation of an oxide/hydroxide layer on the Ni3S2 surface. Therefore, Ni3S2-based electrodes act as good HER electrocatalysts and good OER pre-catalysts [108].Ni3S2 is considered as a pre-catalysts rather than the true catalyst during operation in alkaline, and nickel (oxy)hydroxides are always considered as the real-time catalysts during OER [109–112]. This transformation process is because Ni3S2 would undergo surface reconstruction. From the point of view of solid-state chemistry, the thermodynamic stability of Ni3S2 is not very satisfactory [113]. This property makes the surface of Ni3S2 susceptible to redox reactions in the redox environment. The easy oxidation of the surface of Ni3S2 to the corresponding oxides or hydroxides is therefore expected, particularly in the oxidative conditions of OER. It is reported that this surface reconstruction process could result in a core–shell structure, while other studies suggest it undergoes complete oxidation/anion exchange with hydroxide/oxide during OER [55,56,114–117]. Recently, further understanding of the Ni3S2 surface self-reconstruction process has been achieved by in-situ investigation methods [115,118]. Systematic in-situ/ex-situ experimental investigations and DFT results indicate that the combination of progressive S leaching and OH− adsorption promotes the structural evolution of Ni3S2 into NiOOH/Ni3S2 and the residual SO4 2− could benefit the OER by boosting the adsorption of intermediates [54,119].During the development of Ni3S2-based electrocatalysts, it has attracted much attention for materials preparation research. Experimentally, it is eager to synthesize Ni3S2-based materials with good performance and low cost. Especially for industrial applications, the techniques suitable for large-scale manufacture are one of the research hotspots. In this part, several methods employed to synthesize Ni3S2-based electrocatalysts and the characteristics associated with each technique are introduced.Hydrothermal and solvothermal synthesis methods are widely accepted due to low cost, simple synthesis procedure, and wide application ranges [120,121]. It could obtain electrocatalysts with high purity, good crystallinity, and regular morphology by applying these methods [122]. The core is to select suitable precursors and reaction conditions (temperature, pressure, reaction time, pH, etc.). Previous studies suggest the reaction temperature is the main factor in controlling the structure and morphology of Ni3S2-based catalysts [123]. It is a practical way to adjust catalysts by changing the ratio of reactants and reaction conditions [124].Hydrothermal and solvothermal synthesis methods are time-consuming processes despite their widespread use. Both approaches are hampered by the difficulty of managing the active chemical loading and the shape of the product. In comparison, electrodeposition is a cheap, efficient, and convenient method that synthesizes materials on a nanometer scale under ambient conditions, and it could be used for interface engineering [3,125]. It requires the applied electricity to drive the deposition on conductive substrates. Electrodeposition could also fabricate self-standing electrodes with the deposits rigidly attached to the substrate [126]. Like the hydrothermal method, almost all metal-based materials could be adjusted using varying precursor solutions and controlling the deposition time. As shown in Fig. 3 a and b, electrodes with variable mass loads on nickel foam (NF) labeled NiFeS-1/NF, NiFeS-2/NF, and Ni3S2/NF are fabricated by controlling deposition time from 2.5 min to 5 min. The optimal product shows overpotentials of 231 mV for OER and 180 mV for HER and excellent stability for more than 200 h in alkaline solutions [127]. In this regard, it is highly desired to expand the electrodeposition approach's scope for developing unique micro- and nanostructured non-precious metal-based electrocatalysts [28].The sulfur powders are used as a precursor to react with the nickel substrate to form Ni3S2 directly in the deposition process at the atomic level. The direct growth of Ni3S2 produces a self-supporting electrode that does not require additional binders, which could enhance ion/electron transport and promote better electrolyte channelization [128].It is a facile method to fabricate various composite electrodes by vapor deposition. The substrate not only acts as the nickel source but could accelerate the catalyst growth [129]. Besides, it is also a practical way to create novel catalysts by combining different synthesis methods. For example, an active Ni3S2 is deposited in situ on the active material by combining hydrothermal and chemical vapor deposition techniques [130]. The synthetic process could tightly connect the NF and sulfide, which promotes electron transportation between the Ni3S2 and conductive substrates.Other methods have been applied to prepare Ni3S2-based electrocatalysts with novel nanostructures, including etching deposition and anion-exchanging strategies. A facile method to synthesize efficient catalysts for OER reactions derived from nickel-iron Prussian blue analogs (PBAs) is proposed through the comprehensive utilization of synthetic technologies, as shown in Fig. 3d [131]. Ni(OH)2@NiFe PBA/NF nanoarrays are fabricated firstly by chemical etching, followed by chemical etching/anion exchange. The mass/charge transport and gas release would also be modified by the direct growth of porous nanosheets and three dimensions (3D) interconnected structures. In addition to the etch strategies, the sulfurization process is also a practical way. For example, 3D NiCo2S4 and Ni3S2 nanosheets on NF are synthesized via hydrothermal and thermal sulfurization processes, shown in Fig. 3c [132]. The electrocatalyst's good catalytic performance is attributed to the massive catalytic sites, the better bonding strength, and facilitated electron transport.It is necessary to do further research to increase the electrochemical active surface area and improve the stability of Ni3S2 [49,134]. Thus, several strategies are proposed to address these issues. It generally consists of four categories: 1) electronic structure engineering, including defect engineering and heteroatoms doping; 2) adjusting surface geometry of catalysts; 3) constructing nanostructures; 4) using the conductive and 3D substrates. In the following section, the above strategies are discussed in detail.The Ni3S2 has outstanding conductivity, which makes Ni3S2 function as a critical component to improve conductivity in an electrocatalyst. To enhance the activity, generally, Ni3S2 would be combined with other materials showing high active performance to form a heterogeneous structure. Besides, some electrodes directly use Ni3S2 as an active material for OER reaction due to its unsatisfactory HER activity [135]. For Ni3S2-based materials, electronic structure engineering is always applied for active materials rather than Ni3S2. Combining these materials well to make them work synergistically is also a critical point (see Table 1 ).Constructing heterostructures is an effective way to introduce the interface effect to catalysts. The structured coupling interfaces would remarkably enhance catalytic performances because they generate electronic interactions [136].MoS2 is a promising material for electroactivity due to undercoordinated Mo–S edges [137]. However, its poor conductivity hinders performance improvement. Therefore, integrating the active MoS2 and conductive Ni3S2 is a valuable method for designing effective electrocatalysts. Abundant interfaces could be created by decorating the MoS2 nanosheets on the surface of Ni3S2 nanoparticles. Interfaces between Ni3S2 and MoS2 would facilitate the HER process, and interfaces between the in-situ formed oxidation product of Ni3S2 and MoS2 would facilitate the OER process [138].For heterostructures, synergy is the highlight of this strategy. The fast charge transfer needs to be guaranteed, and the interfaces should be manifested. As shown in Fig. 3a–j, a MoS2/Ni3S2 co-axial catalyst on NF (denoted MoS2/Ni3S2 NE-NF) is reported, which requires overpotentials of 182 and 200 mV at 500 and 1000 mA cm−2, respectively [92]. The co-axial structure could provide great active sites and stimulate the charge transfer along the axial direction. Additionally, the electrical coupling among MoS2, Ni3S2, and NF facilitates electron transfer, thus improving the free Gibbs energy and accelerating the HER reaction's kinetics. Compared with Ni7S6 and Ni3S2–NF catalysts, the better performance of MoS2/Ni3S2 suggests the heterostructure could facilitate the kinetics of the electrocatalytic reaction, which is consistent with previous reports [139–142]. Besides the MoS2/Ni3S2 heterostructures, Ni3S2–Co9S8 heterostructure nanowires on conductive NF could also act as an efficient catalyst for the OER, as shown in Fig. 4 k-r [124]. The defects in the metallic Co9S8 and Ni3S2 interfaces increase the number of catalytic sites and generate electrical interactions. Similarly, a novel MoS2/Co9S8/Ni3S2/Ni electrocatalyst for OER and HER reactions is reported by combining Co9S8 and MoS2, as shown in Fig. 5 [143]. Both experiments and calculation results show that the boosted activity is mainly ascribed to charge transfer at interfaces between the Co9S8 and MoS2. Unlike the MoS2 material, an advantage of Co9S8 is that it possesses better conductivity, which enhances the electrocatalytic activity of the hybrid electrocatalyst.Another effective strategy to construct a heterostructure is to combine hydroxides/(oxy)hydroxides with Ni3S2. For HER, it is reported the well-constructed Ni(OH)2/Ni3S2 nanoforests show promising catalytic performance for HER with overpotentials of 193 mV and 238 mV at the current density of 500 and 1000 mA cm−2, respectively, and maintains the durability for more than 1000 h at 320 mA cm−2 without obvious degradation [144]. The constructed heterointerfaces could generate numerous catalytic sites and defects [47]. In addition, by using hetero-interface engineering with the Ni(OH)2 cocatalyst, the surface atomic configuration of Ni3S2 could be changed in a way that speeds up the Volmer step and OH− adsorption during the HER without blocking the OER [144–146]. For OER, the highly effective and malleable catalysts could be obtained by the combination of NiFe(OH)x and the Ni3S2 with hetero-interfaces [147]. Recently, a novel heterointerface-engineered NiFe(OH)x/Ni3S2 electrocatalyst has been fabricated by the controllable sulfurization method. The obtained catalyst with good catalytic performance shows an overpotential of 310 mV at the current density of 2000 mA cm−2 with the Tafel slope of only 20.8 mV dec−1. In addition, this catalyst could steady work for more than 100 h without significant degradation at the current density of 1000 mA cm−2. The outstanding performance could be attributed to the heterointerface synergy between NiFe(OH)x and Ni3S2. The heterointerface leads to a parallel catalytic mechanism for OER. Namely, intermediates (HO∗ and O∗) could be absorbed by NiFe(OH)x and Ni3S2, respectively, and then incorporate to HOO∗ at the interface. This path could break the scaling relationship of OER (ΔGHOO∗ − ΔGHO∗ ∼ 3.2 eV), which improves the catalytic performance very well, particularly at high current densities [147,148].Among strategies for improving performance, element doping is universal and could regulate the coordination of valence states and the chemical environment. In general, the introduction of heteroatoms could increase catalytic sites and superior conductivity. Iron (Fe)-doped, zinc (Zn)-doped, tungsten (W)-doped, and nitrogen (N)-doped Ni3S2 hybrid structures have been reported, presenting the performance improvement in the HER and OER reactions.Recent studies have shown that incorporating Fe ions could significantly improve Ni3S2-based electrocatalyst performance by tuning the local electronic structure. For instance, a reported Fe-doped Ni3S2 electrocatalyst could exhibit outstanding performance under large current density [150]. However, the preparation process is complex.Recent studies have shown that incorporating Fe ions could significantly improve Ni3S2-based electrocatalyst performance by tuning the local electronic structure. For instance, a reported Fe-doped Ni3S2 electrocatalyst could exhibit outstanding performance under large current density [150]. However, the preparation process is complex.Recently, novel Fe-doped Ni3S2 electrodes conducted by simplified preparation methods have been reported [151,152]. Apparent advantages of doping Fe include: (1) The doping Fe would result in a disordered state, thereby increasing the density of states (DOS), providing more conductive paths, and facilitating electron transfer. (2) Fe could act as active sites, resulting in boosted H2O adsorption and easier O2 desorption.In addition, it is reported that Fe affects the morphology of Ni3S2-based electrocatalysts, as shown in Fig. 6 a–d. The experiment results suggest that adding only 2.1 (at.)% of Fe (Ⅲ) could make a difference to the catalytic ability and control the evolution of the morphology of Ni3S2 from nanorods to nanosheets, doubling electrochemical surface area [153]. Therefore, it is a practical and effective method to enhance the catalytic ability of Ni3S2-based catalysts, especially for the OER reaction, by doping the proper amount of Fe.Cu, as another earth–abundant transition metal, possesses excellent electronic conductivity. Moreover, Cu and Ni are adjacent in the element periodic table, which means they have similar electronegativity and atomic radius. This similarity provides the basis for doping Cu into Ni3S2. Doping Cu element would optimize energy barriers of intermediate steps and enhance the conductivity. Several studies have used Cu as the doping element to improve the catalytic performance of Ni3S2-based electrocatalysts [154–156]. For example, a novel Cu-doped Ni3S2-based electrocatalyst is reported for the HER reaction [157]. Copper nanodots (NDs) are deposited on the Ni3S2 surface supported by chemical reduction in carbon fibers (CFs). The introduction of Cu generates the electronic interaction between Cu and Ni3S2, which positively charges Cu NDs and negatively charges Ni3S2. Cu could promote water adsorption and the Ni3S2 could weaken S-Hads bonds, so the Volmer and Heyrovsky steps are announced, as shown in Fig. 6e–k. Besides, the Cu dopants could change morphology, increase active site strength, and improve intrinsic activity by optimizing the energy barrier, which is consistent with the study mentioned above [158].It is reported that V dopants could be used as a hopeful constituent to optimize the catalytic properties of Ni3S2-based catalysts for water electrolysis [159]. For the HER reaction, it is indicated that the interaction between the V and S could optimize the hydrogen adsorption Gibbs free energy, thus manifesting better catalytic activity [160,161]. Introducing V leads to a large number of free carriers around the Fermi level, which accelerates the charge transfer [160]. For the OER reaction, doping V could introduce active sites and create synergy with Ni3S2 [159]. For instance, a V-doped Ni3S2 nanorod is fabricated and decorated by NiFe-layered double hydroxide (LDH) nanosheets on the NF [162]. The Ni3S2 functions as conductive cores, and the doped V improves conductivity. The defective NiFe LDH nanosheets provide plenty of active sites and 3D core–shell nanostructures. The obtained catalyst exhibits overpotentials of 209 and 286 mV at 10 and 100 mA cm−2, respectively. The good performance is attributed to several strategies, including heteroatom doping, defect engineering, and constructed hierarchical nanostructure.Anion doping (F, P, and N doping, etc.) is an ideal avenue to modify the electronic structure of advanced Ni3S2-based electrocatalysts. This strategy could also increase active sites and improve conductivity [163]. Compared to transition-metal cation doping, anion doping could simplify active site identification [164,165]. A N-doped Ni3S2 material, denoted N–Ni3S2/NF 3D electrode, is put forward by a one-step calcination route [16]. The electrocatalyst requires overpotentials of 330 and 110 mV for OER and HER at 100 mA cm−2 and 10 mA cm−2 in 1 mol L-1 KOH solutions. The decoration of N optimizes the electronic structure, changes the morphology of Ni3S2, and provides appropriate HER Gibbs free energy and water adsorption energy [166]. A similar effect could be made by rationally doping the F element in Ni3S2 because the electronegativity different between F and S is bigger than that between N and S, leading to a stronger interaction between the anion and transition-metal cation [167,168].In addition, it is reported that a phosphorus-doped Ni3S2/NF electrode needs a relatively low overpotential of 306 mV at the current density of 100 mA cm−2 for OER and 137 mV at the current density of 10 mA cm−2 for HER [169]. Doping P could also improve the catalytic performance of HER due to the modified electronic structure, more active sites, and improved electrical conductivity [170].Many studies suggest that W doping would modify the hydrogen adsorption energy of intermediates and improve conductivity [171–173]. Inspired by this, a W-doped Ni3S2 nanoparticle catalyst that presents HER overpotentials (67 mV at 10 mA cm−2 and 330 mV at 552 mA cm−2) is fabricated and exhibits merely an increase of approximately 15 mV for overpotential after 40 h [174]. Moreover, the effect of doping Zn is not disappointing. A Zn-doped Ni3S2 nanosheet catalyst denoted Zn–Ni3S2/NF presents a high catalytic ability and good stability for OER reaction [175]. It requires an overpotential of 330 mV at 100 mA cm−2 and maintains activity for 20 h without apparent degradation.Despite many reports having applied various strategies to improve Ni3S2-based electrocatalysts, relatively few studies have focused on the Ni3S2 nanostructures with exposed active facets. The understanding of the surface structure is not deep enough, leading to the unclarity of the location of the catalytic sites that promote reactions on the surface.A stable Ni3S2 nanosheet array electrocatalyst with ( 2 1 ¯ 0 ) -exposed high-index facets were reported for the first time in 2015 [49]. The catalyst exhibits excellent catalytic performance for both OER and HER reactions, which attributes to the synergistic effects between 3D nanostructure and ( 2 1 ¯ 0 ) high-index facets. Compared with low-index facets, such as (001), ( 2 1 ¯ 0 ) facets have more optimized surface structure for electrocatalysis, which could be substantiated by the fact that the catalytic activity of ( 2 1 ¯ 0 ) -Ni3S2/NF is better than that of (001)-exposed catalysts. It is indicated that the S sites and six-coordinated Ni5 sites at the terrace of ( 2 1 ¯ 0 ) facets have a less steric effect, as shown in Fig. 7 a and b. In addition to exposing the ( 2 1 ¯ 0 ) facets (003) facets of Ni3S2 are reported to contribute to HER activity due to active Ni3-triangle active sites, as shown in Fig. 7c–i. The low-coordinated Ni3-triangles are more favorable for O–H bond breakage than other sites. They are the only sites for OH− adsorption [105]. Thus, the unique Ni3-triangles make (003) facets beneficial to improve water dissociation for HER reaction.Disorder engineering serves as an effective method to fabricate amorphous structures and has attracted much attention. It affects the lattice distortion of materials and thus contributes to the activity via improving ionic conductivity and increasing vacancy number.A hierarchical (Fe–Ni)Cox-OH/Ni3S2 electrocatalyst for OER and HER reactions is assembled [176]. This catalyst needs overpotentials of 91 and 145 mV at 100 and 1400 mA cm−2, respectively, with outstanding durability for 100 h at 200 mA cm−2 in 1 mol L-1 KOH solution. The Gibbs free energy for intermediates tends to be more moderate by breaking the long-range order to form the S-vacant amorphous phase. In general, there are two types of Ni3S2-based catalysts for which unstructured engineering is applied. As mentioned above, one is the Ni3S2 material itself, and the other is materials coupled with Ni3S2. For example, the synergistic effect between the coated amorphous α-MoS2 and Ni3S2 skeleton is reported [177]. Results indicate that most compounds of Mo present long-range disordered, which leads to the distortion of the crystalline lattice of Ni3S2. The amorphous structure could generate many active sites, thus exhibiting higher catalytic activity.However, there is a paradox between the active site number and the conductivity. Excess active sites would impress the conductivity and further constrain the electrocatalytic process. An optimum electrocatalyst should exhibit a short-range order and long-range disorder structure to balance active sites and conductivity. Inspired by this, the relationship between the disorder degree and OER reaction activity is deeply explored via fabricating novel Fe-doped Ni3S2 electrocatalysts [178]. By regulating the molar concentration in hydrothermal synthesis, a series of Fe-doped Ni3S2 electrocatalysts with different disorder degrees are obtained. The best active one exhibits an overpotential of 295 mV at 10 mA cm−2 in alkaline solutions. The results indicate that a great degree of disorder might restrict OER activity. It is necessary to select a suitable component ratio to obtain the facile extent of the amorphous phase.One-dimensional nanostructures have gained a lot of interest as electrocatalysts due to good mass and charge transmission and efficient bubbles removal [179–181]. Thus, one-dimensional Ni3S2-based electrocatalysts have been widely investigated, allowing exquisite composition, morphology, heterostructures, and reactivity controls[139,160,182,183]. The typical structure mainly includes nanotubes, nanowires, and nanorods.Ni3S2 nanotube arrays on NF are fabricated by the catalysis of thermally reduced graphene for the HER [184]. The resulting electrode exhibits an overpotential of 157 mV at 10 mA cm-2. The average pore size is 6.9 nm, which increases the surface area and the number of active sites. Besides, the nanotube structure could bolster electrolyte diffusion, making active sites more accessible.The nanowire is also widely used in electrocatalysts [2,185]. A core/shell electrocatalyst based on Ni3S2 nanowires and N-doped carbon layers, denoted as Ni3S2@NGCLs/NF, exhibits overpotentials of 271 mV and 134 mV at 10 mA cm-2 in alkaline solutions for OER and HER reactions, respectively [183] As shown in Fig. 8 b–g, one-dimensional Ni3S2 nanowires could enhance the inherent activity and provide more active sites. Moreover, the activity for HER and OER reactions could be improved by compounding other materials to facilitate charge transfer to the nanowire structure.Unlike conventional electrocatalysts with limited active sites, 2D structures could increase specific surface area to expose more active sites. For example, a 2D nanosheet heterogeneous electrocatalyst donated as Ni(OH)2/Ni3S2, as shown in Fig. 8i–k, exhibits the synergistic effect [146]. Due to the strong OH− adsorption ability, water could be effectively cleaved into H∗ and OH−. Then, the increased H∗ intermediates would absorb on the Ni3S2 side and recombine to H2 because hydrogen generation's energy barrier is lower than Ni(OH)2 [186]. The mechanism study exhibits the significant effect of Ni(OH)2 to accelerate the water dissociation. However, the van der Waals attraction between 2D structures makes them aggregate during the preparation process, limiting the accessible regions and degrading electrochemical performance [187]. The development of 3D structures might effectively overcome this issue [188].In general, there are two ways to construct 3D Ni3S2-based electrocatalysts. One is that the active material (i.e., Ni3S2) possesses 3D morphology; the other is combining one-dimensional/two-dimensional Ni3S2 with a 3D substrate, including NF, copper foam, and so on [106,189–192].Compared with the one- and two-dimensional Ni3S2 materials, 3D Ni3S2 endows a larger surface area, more active sites, plenty of pore channels, rapid gas bubbles release, and boosted electron transfer ability [188,193–195]. 3D Ni3S2-based electrocatalysts have been regarded as a competitive candidate for alkaline water electrolysis. For instance, the hollow MoOx/Ni3S2 composite microsphere catalysts on NF are synthesized for OER reaction for the first time, as shown in Fig. 8h-n [106]. The fabricated material possesses the hollow and 3D microporous structure with a 0.5–1 μm diameter. The shell thickness is determined to be ≈ 60 nm. This ultrathin nanosheet-assembled hollow architecture provides a large active area and more accessible catalytic sites, which could be verified by the fact that it has roughly 37 times the higher electrochemical active area than Ni3S2/NF. Recently, a similar Ni3S2-based electrocatalyst with a microsphere-like interconnected structure is also reported by other researchers [196]. Besides, the 3D treelike structure could reduce ion and electron transport resistance, facilitate electrolyte penetration, and expose numerous active sites [189]. If the material also has an amorphous phase, the amorphous phase would further enhance the electric conductivity.The conventional method for preparing electrodes is to drop slurry consisting of conductive binder and catalyst onto the electrodes [122,197]. However, this method has two apparent drawbacks. The first one is that the binder will decrease the contact area of the catalyst and electrolytes, thus increasing the ohmic resistance; the second is that attached catalysts tend to peel off at high current density. New electrodes are developed by growing electrocatalysts directly on the conductive substrates to overcome the abovementioned issues. This method no longer needs the binder and could accelerate the electron-transfer rate. Substrates should have good conductivity and make catalytic species expose more sites. Under these requirements, 3D substrates become the most promising candidate for preparing efficient catalysts for water electrolysis.The NF has attracted much attention among different substrates due to its good conductivity and porous 3D structure. The catalyst is evenly distributed on the porous NF, exposing more active sites and avoiding agglomeration. Besides, NF with open architectures is beneficial to release bubbles rapidly and to dispense with binder agency [154]. The NF always serves as the conductive porous substrate and forms a unique 3D coral-like electrocatalyst [32]. Besides, it can be used as the nickel source for the formation of Ni3S2. Analogously, conductive copper foam is also used as a substrate because of the 3D structure with a smooth surface [198].In addition to NF and copper foam, graphene is another promising option for porous support. The good stability and excellent conductivity of graphene could be attributed to its two-dimensional hexagonal stacked carbon nanostructure and π-bond that allows electrons to move freely [199]. Moreover, catalytic performance is also enhanced by increasing the surface area of microporous channels of graphene [200]. Besides graphene, graphene oxide (GO) is also an excellent alternative substrate. Due to many oxygen-containing functional groups, GO has good water dispersibility and is easy to assemble. The as-prepared catalyst exhibits that the GO could effectively hinder the aggregation of Ni3S2 nanoparticles [201]. Therefore, it could be an effective way to construct the composites by growing Ni3S2 active materials on carbon materials [202].Besides the abovementioned 3D substrates, a 3D NiFeCo foam is reported as the substrate to fabricate Ni3S2–FeS–CoS trimetallic sulfide nanosheets catalyst, shown in Fig. 8l [9]. The NiFeCo foam provides the foundation for the growth of Ni3S2–FeS–CoS nanosheets. The catalytic performance of this catalyst is boosted by the large surface area of nanosheets grown from NiFeCo foam and the good conductivity of the substrate.Though the strategies above have been widely adopted to design active and durable Ni3S2-based catalysts at the current density level of tens mA cm−2, it is still imperative to further improve the performance under the high current density (see Table 2). It requires that the catalyst not exceed 300 mV at 500 mA cm−2 for industrial application [203,204]. Some advances have been made in designing nanostructured Ni3S2-based electrocatalysts for industrial alkaline electrolysis. Some reported Ni3S2-based materials present good catalytic stability at 500 mA cm−2 and even 1000 mA cm−2. For example, the NixFe1-x alloy - oxyhydroxide nanowire arrays (donated as NixFe1-x-AHNAs) have the overpotential of merely 248 and 258 mV at 500 mA cm−2 and 1000 mA cm−2, respectively. Notably, the alkaline water electrolyzer using NixFe1-x-AHNAs as the anode and pure Ni as the cathode exhibits a potential of 1.76 V at 100 mA cm−2 [2]. Table 3 summarizes recently reported Ni3S2-based catalysts for alkaline water electrolysis at high current density. Even though impressive materials have been reported, most Ni3S2-based electrocatalysts operate inefficiently and unstably at 0.5 and 1 A cm−2 [90,205]. It is essential to pay attention to electrochemical stability and bubble release behavior to enhance the performance at high current density.Stability is the most important and difficult part of maintaining the performance of Ni3S2-based electrocatalysts at high current density. An electrocatalyst with high stability should keep the composition and structure throughout the reaction, which is important because both of them affect how well the electrocatalyst works. The electrochemical composition stability is discussed in this section, and the discussion about the structure stability is provided in section 6.2.3.Electrochemical stability is a critical point in evaluating the electrocatalyst's performance. It is reported that the durability of Ni3S2 could be enhanced by increasing the Fe content [209–211]. In addition to Fe, other elements, including cerium (Ce) and phosphorus (P), have been reported to be advantageous for creating durable Ni3S2-based electrocatalysts [212,213]. A reported Ce-doped Ni3S2-based electrode shows good retention of 84.7% for 24 h at 60 mA cm−2, which might be because the doping Ce prevents the structure from evident evolution, as shown in Fig. 9 a [214].During the OER process, it is easy to form a thin layer of amorphous oxide/peroxide on the surface, which protects the core material from being oxidized [215–218]. Therefore, the stability could be boosted by building a suitable shell–core structure. For example, NixCo3-xS4 coupled Ni3S2 nanosheet arrays on NF show the potential of 1.53 and 1.80 V at 10 and 100 mA cm−2 with stability for >200 h for overall water electrolysis, as shown in Fig. 9d [191]. The NixCo3-xS4 shell could be transformed into hydroxides during the OER process, protecting Ni3S2 nanosheets from collapsing. A similar effect has also been reported, as shown in Fig. 9h [2]. The shell is densely covered on the core's surface and maintains stability under alkaline conditions, preventing the core from corrosion. In addition, the stability could also be modified by applying rational synthesis methods. It is indicated that catalytic ability and stability could be bolstered by electrodeposition because it combines interface modification and 3D porous structure construction [125,219].The alkaline water electrolysis system is constantly hampered by the adherent bubble layer resulting from dissolved gas bubbles on electrode surfaces. The coverage of bubbles would act as an electrical shield that increases the total resistance and impedes the mass transfer by reducing interfaces between electrodes and electrolytes [220,221]. In addition, the tension caused by the detachment of numerous bubbles would also damage the structure of electrodes. Thus, it is essential to understand the bubbles removal behavior to alleviate the adverse effects.The bubble release process is divided into two stages. Firstly, tiny bubbles are generated on the electrode surface. The size depends on the surface topography; secondly, tiny bubbles leave the electrode surface and combine to form large bubbles. These bubbles could be restricted within the cavity of electrode surfaces, which has adverse effects on electrode performance, especially at high current density [222]. Therefore, rapid bubble release is essential to improve electrode performance. Currently, two methods are mainly proposed to accelerate the bubbles removal process. The first one is to construct the superhydrophilic surface of electrodes. The second one introduces external fields, including magnetic, ultrasonic, and supergravity [222].·Water molecules close to the electrode surface will compete with bubbles attempting to adhere to the surface when the electrode is immersed in aqueous solutions. The surface with a bubble contact angle greater than 150° is the superoaerophobic surface [223,224]. Equation (11) could explain the relationship between the bubble contact angle and water contact angle： (11) θ b  = 180° - θ w Where the θ b represents the bubble connect angle and θ w means the water connect angle [223]. It could be indicated that a superhydrophilic surface is superaerobic under water and vice versa. The relationship between the wettability of substrate in the atmosphere and aqueous solutions is shown in Fig. 10 a [222]. It is indicated that good surface wettability favors mass transfer because it could boost the contact between electrolytes and electrode surfaces [222]. Therefore, hydrophobic and hydrophobic qualities could be concurrently gained by constructing superaerophobic surfaces, which could complement mass transfer and bubbles removal behaviors.It is indicated that rough nanostructures are investigated to modify the bubbles removal behavior of Ni3S2-based electrocatalysts [162,193,195,225–229]. Rapid water entry could crowd tiny bubbles out when generated on surfaces [142,230]. Fig. 10b shows that the hierarchical morphology could change the three-phase contact line (TPCL) from a continuous curve to a discontinuous line segment, which could reduce the contact area between bubbles and solid electrodes and then reduce adhesive force to accelerate the bubbles removal [225,231]. A novel CoSx-Ni3S2 nanosheet catalyst simultaneously possess apparent superhydrophobicity and superaerophobicity, which could expose more active sites and alleviate the “bubble shielding effects” for gas diffusion, as shown in Fig. 10d–j [232]. The as-prepared electrocatalyst exhibits 1.63 V at 100 mA cm−2. Another MoS2/Ni3S2 catalyst with co-axial heterostructure nanowires shows a 0° contact angle of a 1.0 M KOH droplet, demonstrating good wettability [92]. This phenomenon might be attributed to the multilayer and low-crystallinity structure of the outer MoS2 nanosheet [233,234]. It is also indicated that sulfurization time could affect surface wettability, as presented in Fig. 10c [105]. Furthermore, changing chemical composition is also a method to modify surface wettability. The regulation of the amount of doping Fe is a practical way to adjust the wettability. This result again proves the important role of iron in enhancing the performance of Ni3S2-based electrocatalysts [153].Besides constructing a nanoscale surface, it has been shown that various methods, including the magnetic field, supergravity, and ultrasonic field, could accelerate bubble dissociation and release. It is reported that the external magnetic field affects the bubble releaseProcess. The bubble coverage and size are reduced under the external magnetic field because the magnetohydrodynamic (MHD) convection could expedite the bubbles removal behavior [235–237]. The catalytic performance could be boosted by applying a moderate magnetic field since the external magnetic field causes some components to be magnetized into a high spin polarization state, optimizing the energy barrier of oxygen intermediates and the transfer of electrons [238]. Besides the magnetic method, it is indicated the supergravity field and ultrasonic fields could promote the bubbles removal behavior [239]. However, few studies are applying these external fields to Ni3S2-based electrocatalysts. The main limitation of using external fields is that external fields would increase the complexity and cost of the electrolysis system.The main difference between water electrolysis at low current density and high current density is the number of generated bubbles. For an electrocatalyst with good durability, both electrochemical composition and structural stability are required. In most cases, the structural stability of electrocatalysts is related to the bubble release behavior at high current density [240]. Even though the bubble release rate could be accelerated by surface engineering, the catalyst could peel-off by the strain induced by the detachment of bubbles [241]. In addition, bubbles would be made and built up in the cavity if water molecules are taken into the cavity of electrocatalysts, which would stress the structure and lead to the structure oscillation [242,243]. Therefore, the catalyst and substrate should be connected firmly to overcome the interfacial adhesion between the bubble and the catalyst [244]. Some self-stand and binder-free Ni3S2-based electrocatalysts with good structure stability have been fabricated, while the threshold value of the interaction force between the catalyst and substrate should be found [245–247]. The understanding of catalyst/substrate interfacial structure that resists gas bubble strain also needs to be further deepened [240].In recent decades, it has witnessed the rapid development of non-noble metal-based electrocatalysts for alkaline water electrolysis. Among them, Ni3S2-based catalysts are considered promising candidates.Ni3S2-based catalysts could be designed and prepared by elaborate strategies consisting of electronic structure engineering, lattice strain engineering, morphology design, and usage of 3D substrates. Recent studies always adopt a combination of strategies to improve catalytic activity. Among them, constructing the heterostructure, doping heteroatom, and using 3D substrate are commonly adopted methods. For Ni3S2-based catalysts with heterostructure, the combination of MoS2 and Ni3S2 is a classic strategy and could improve the catalytic activity effectively. Besides, incorporating between (oxy)hydroxides and Ni3S2 could also enhance the performance effectively and fabricate catalysts that could meet the catalytic activity requirements of commercial applications. For heteroatom-doped Ni3S2-based catalysts, Fe, V, and anion elements (F, N, P) are widely chosen. However, most of the catalysts are doped by hydrothermal methods, which may hinder the further improvement of performance. Thus more ways of doping are expected. For Ni3S2-based catalysts with 3D substrates, the Ni foam is widely adopted due to its good conductivity and porous 3D structure.From the view of practical applications, electrode bubble release behavior, electrode surface wettability, and electrochemical stability of Ni3S2 at high current density are also vital points for industrial applications. Up to now, challenges still exist which hinder the application of Ni3S2-based catalysts. It is summarized as follows, combing with prospects. 1. The performance should be further optimized. Ni3S2-based electrocatalysts still have disadvantages, such as high overpotentials and inferior durability. Few of them meet the commercial requirements considering their electrochemical activity and stability, especially working at high current density (>500 mA cm−2), high temperature (60–90 °C), and concentrated solutions (6–10 M KOH) [12,248,249]. The stability requirements are relatively more difficult to achieve. Most stability tests of Ni3S2-based electrocatalysts are still tested under low current densities, which obviously cannot meet the needs of future commercialization. 2. The real electrochemical active site is still unclear. For Ni3S2-based electrocatalysts, the thermodynamic instability of Ni3S2 makes electrocatalysts undergo surface reconstruction. There are still debates on whether the in-situ formed (oxy)hydroxide or the electrocatalyst itself acts as the active site. Therefore, it is necessary to clarify the active site or active phase. In addition, the reconstruction process may differ on the lab and commercial scales. To uncover the real catalytic species, in-situ characterization technologies should be further developed to observe the dynamic changes of active sites during the electrochemical reaction. Moreover, the computational simulation needs to be combined to guide further investigations. 3. The variety of nanomaterials composed of Ni3S2 should be widened. For instance, anion-doped Ni3S2-based electrocatalysts are suitable candidates and should get more attention because of the modified electronic structure, increased active sites, and better conductivity. Furthermore, exploring further viable composite catalyst material combinations based on Ni3S2 will advance the development of alkaline water electrolysis. 4. Research on bubble release behaviors needs to be carried out sincerely. The bubbles removal dynamic process is complex in alkaline water solutions, which is related to composition, microstructure, the surface wettability of electrodes and working current density, etc. It is recommended that various electrocatalysts with high intrinsic activity could be combined with substrates with the 3D structure to reduce the adverse impacts on bubbles removal, especially at high current density. 5. The synthesis method at the industrial scale needs further improvement. The synthesis method suitable for industrial application should be mature and scalable, and should be able to provide electrocatalysts with tailor-made shapes and compositions according to industrial applications. Among various catalyst preparation methods, electrodeposition could be an ideal candidate for industrial-scale preparation due to its convenient operation conditions control. However, there is a lack of comprehensive technological procedures and standards, which needs to be focused on in the following research. The performance should be further optimized. Ni3S2-based electrocatalysts still have disadvantages, such as high overpotentials and inferior durability. Few of them meet the commercial requirements considering their electrochemical activity and stability, especially working at high current density (>500 mA cm−2), high temperature (60–90 °C), and concentrated solutions (6–10 M KOH) [12,248,249]. The stability requirements are relatively more difficult to achieve. Most stability tests of Ni3S2-based electrocatalysts are still tested under low current densities, which obviously cannot meet the needs of future commercialization.The real electrochemical active site is still unclear. For Ni3S2-based electrocatalysts, the thermodynamic instability of Ni3S2 makes electrocatalysts undergo surface reconstruction. There are still debates on whether the in-situ formed (oxy)hydroxide or the electrocatalyst itself acts as the active site. Therefore, it is necessary to clarify the active site or active phase. In addition, the reconstruction process may differ on the lab and commercial scales. To uncover the real catalytic species, in-situ characterization technologies should be further developed to observe the dynamic changes of active sites during the electrochemical reaction. Moreover, the computational simulation needs to be combined to guide further investigations.The variety of nanomaterials composed of Ni3S2 should be widened. For instance, anion-doped Ni3S2-based electrocatalysts are suitable candidates and should get more attention because of the modified electronic structure, increased active sites, and better conductivity. Furthermore, exploring further viable composite catalyst material combinations based on Ni3S2 will advance the development of alkaline water electrolysis.Research on bubble release behaviors needs to be carried out sincerely. The bubbles removal dynamic process is complex in alkaline water solutions, which is related to composition, microstructure, the surface wettability of electrodes and working current density, etc. It is recommended that various electrocatalysts with high intrinsic activity could be combined with substrates with the 3D structure to reduce the adverse impacts on bubbles removal, especially at high current density.The synthesis method at the industrial scale needs further improvement. The synthesis method suitable for industrial application should be mature and scalable, and should be able to provide electrocatalysts with tailor-made shapes and compositions according to industrial applications. Among various catalyst preparation methods, electrodeposition could be an ideal candidate for industrial-scale preparation due to its convenient operation conditions control. However, there is a lack of comprehensive technological procedures and standards, which needs to be focused on in the following research.The authors declare no conflict of interest.This work is supported by the National Key Research and Development Program (No. 2022YFB4202200) and the Fundamental Research Funds for the Central Universities.

Green hydrogen (H2) produced by renewable energy powered alkaline water electrolysis is a promising alternative to fossil fuels due to its high energy density with zero-carbon emissions. However, efficient and economic H2 production by alkaline water electrolysis is hindered by the sluggish hydrogen evolution reaction (HER) and oxygen evolution reaction (OER). Therefore, it is imperative to design and fabricate high-active and low-cost non-precious metal catalysts to improve the HER and OER performance, which affects the energy efficiency of alkaline water electrolysis. Ni3S2 with the heazlewoodite structure is a potential electrocatalyst with near-metal conductivity due to the Ni–Ni metal network. Here, the review comprehensively presents the recent progress of Ni3S2-based electrocatalysts for alkaline water electrocatalysis. Herein, the HER and OER mechanisms, performance evaluation criteria, preparation methods, and strategies for performance improvement of Ni3S2-based electrocatalysts are discussed. The challenges and perspectives are also analyzed.

Concerns about the vital global warming and ocean acidification problems caused by CO2 excessive emission (Karl and Trenberth, 2003; Orr et al., 2005) have triggered extensive researches on its large-scale reutilization via effective, economical, and sustainable technologies for a CO2 circular economy (Aresta et al., 2014; Porosoff et al., 2016). However, industrialized CO2 reutilization is just limited to the synthesis of urea and polycarbonate (occupying only 0.5% [Shima et al., 2012; Su et al., 2017] of CO2 emissions), whereas enzymatic and electro-/photo-chemical strategies are hampered by their low CO2-conversion efficiency (Wang et al., 2008; Kondratenko et al., 2013). To achieve the large-scale CO2 reutilization, CO2 hydrogenation with renewable-energy-generated H2 to CO by the reverse water-gas shift (RWGS) reaction is the most techno-economically viable candidate (Porosoff et al., 2016; Kondratenko et al., 2013; Xu and Moulijn, 1996; Porosoff and Chen, 2013; Zhang et al., 2017), thanks not only to its high efficiency, enabling to deal with vast amounts of CO2, but also to the great versatility of syngas (CO + H2, product gas of RWGS reaction) to produce commodity chemicals and fuels (occupying 40% CO2 emissions [Zhang et al., 2017] via mature Fischer-Tropsch and methanol (CH3OH) syntheses [Porosoff et al., 2016; Kondrat et al., 2016]).The RWGS reaction is an equilibrium-limited endothermic reaction (required enthalpy of 41.17 kJ mol−1). According to Le Châtelier's principle, high-temperature (about 400–800°C) thermodynamically favors high CO2 conversion and high CO selectivity, but the undesired methanation also proceeds under the preferred RWGS conditions (Chen et al., 2001; Wu et al., 2015; Gonçalves et al., 2017; Yang et al., 2017). Therefore, a techno-economically available catalyst with outstanding CO2-to-syngas performance is the prerequisite for the large-scale RWGS implementation. To date, homogeneous complexes and heterogeneous solids catalysts have been extensively explored. The homogeneous catalysts show satisfactory activity and selectivity (Federsel et al., 2010), but their difficult recovery from the reaction mixture makes them unattractive. The heterogeneous catalysts are more competitive in terms of ready catalyst-product separation and continuous processes. They mainly include the nanoparticles of precious metals (e.g., Au, Ag, Pt) (Porosoff et al., 2016; Yang et al., 2017) and non-precious metals (e.g., Cu, Ni) (Zhang et al., 2017; Chen et al., 2001; Wu et al., 2015; Gonçalves et al., 2017) dispersed on supports (e.g., SiO2, Al2O3, CeO2, MoCx) (Porosoff et al., 2016; Zhang et al., 2017; Chen et al., 2001; Wu et al., 2015; Gonçalves et al., 2017; Yang et al., 2017). Despite the excellent RWGS activity, the precious-metal catalysts suffer from their limited natural abundance. Cu and Ni catalysts are intensively studied but are not promising owing to either serious sintering (Cu) (Zhang et al., 2017; Chen et al., 2001) or high methanation activity (Ni) (Wu et al., 2015; Gonçalves et al., 2017). Given the chemical inertness of CO2 molecule (Xu and Moulijn, 1996), the heart of RWGS is to exquisitely design and tailor a groundbreaking catalytic material with both high efficiency and low cost, but this represents a grand challenge within the CO2-conversion field.Against all odds, the tantalizing progresses in nano-intermetallic catalysis (Stamenkovic et al., 2007; Studt et al., 2014) open an opportunity for designing and tailoring qualified RWGS catalysts because nano-intermetallic has fascinating prospects in catalysis field, with their tunable components and ratios, variable constructions, and reconfigurable electronic structures, distinctly different from their single metals (Stamenkovic et al., 2007; Armbrüster et al., 2012; Ji et al., 2010). Particularly, their precise atomic ordering structure can provide rational predictions of the effects of geometry and electronic structure on their catalytic properties for required reactions (Wang et al., 2013; Nicholson et al., 2014; Qin et al., 2018). One of the recent pertinent examples is the discovery of a Ni5Ga3 nano-intermetallic, which strikingly shows that the Ni, originally active for CO2 methanation, turns itself suddenly into a qualified CO2-to-CH3OH catalyst after Ga alloying (Studt et al., 2014), because this intermetallic offers the unique Ga-rich sites for CH3OH formation. Encouraged by these big achievements toward nano-intermetallic catalysis, we believe that the nano-intermetallic can pave a road to the rational engineering of more intelligent catalysts gifted with flexibly arranged atomic structures and tailor-made catalytic properties for the RWGS reaction as well as other reactions for CO2 reutilization.Here, we present a nano-intermetallic InNi3C0.5 catalyst that is particularly active, selective, and stable for the RWGS reaction under extremely wide reaction conditions. Such nano-intermetallic is fabricated via carburizing the In-Ni nano-intermetallic in the real RWGS stream and is gifted with dual active sites (i.e., 3Ni-In and 3Ni-C) on the InNi3C0.5(111) surface. The dual sites act in synergy to facilely dissociate CO2* (adsorbed on 3Ni-In sites) into CO* (on 3Ni-C sites) and O* (on 3Ni-In sites), and the O* can favorably react with 3Ni-C offered H* to form H2O. Most notably, the CO* is mainly desorbed into gas phase at and above 400°C but can be highly selectively hydrogenated to form CH3OH below 300°C with a promising CO2-to-CH3OH capacity. Furthermore, this nano-intermetallic can fully hydrogenate dimethyl oxalate (obtainable from oxidative coupling of CO (Fenton and Steinwand, 1974), product of the RWGS) to ethylene glycol (a commodity chemical) with high selectivity (above 96%) and favorable stability.To exquisitely tailor a groundbreaking RWGS catalyst, the elaborate choice of appropriate elements oriented by this reaction should be initially conducted but poses a great challenge because the relevant elements for this reaction traverse most of the periodic table. The first metal that mostly attracts attention is Ni, because Ni-based catalysts are typically used for the RWGS reaction despite CH4 formation (Wu et al., 2015; Gonçalves et al., 2017). Moreover, In is another attractive element, because In-based catalysts are burgeoning in CO2 conversion (Ye et al., 2012; Park et al., 2017; Larrazábal et al., 2016), and, for example, the intermetallic AgIn catalyst is highly efficient for electrochemical reduction of CO2 to CO (Park et al., 2017; Larrazábal et al., 2016). We thus surmise that In-Ni intermetallic could reconstruct geometric-electronic structures of Ni, which might be feasible to switch Ni catalysis in CO2 reduction from CH4 formation to CO formation.A series of pure intermetallics of InNi, InNi2, and InNi3 were successfully synthesized (Figure 1 A) and then were evaluated for the RWGS reaction. Comparison with the conventional Cu-based catalysts (Zhang et al., 2017; Chen et al., 2001) reveals that the intermetallic In-Ni catalysts deliver exciting intrinsic RWGS performances, especially for InNi3 with a high CO formation rate of 1.96 mmol gcat −1 min−1 and a considerably low CH4 selectivity (Figure S1). It is very intriguing to find that after reaction the InNi, InNi2, and InNi3 phases are in situ changed in association with a new phase formation of InNi3C0.5 (Figure 1B, identified in following section). Consistently, the InNi3C0.5 formation is thermodynamically favorable with large ordering energy (such as 2.72 eV for InNi3 carburization with CO, Figure S2), which also portends that the InNi3C0.5 is stable under the RWGS conditions. Notably, only InNi3 could be fully transformed into pure InNi3C0.5 owing to the identical stoichiometric In:Ni ratios of 1:3 and offers the highest RWGS performance, indicating that InNi3C0.5 should be responsible for the RWGS reaction.The above-mentioned results and analyses make us confident that the InNi3C0.5 intermetallic is a superior RWGS catalyst. To make it a practical catalyst, the thin-felt Al2O3/Al-fiber substrate consisting of 10 vol% 60-μm Al2O3/Al-fiber and 90 vol% voidage (Wang et al., 2016) was used to support 9 wt% InNi3C0.5. This strategy permits the engineering of InNi3C0.5 nano-intermetallic at “nano-meso-macro” triple-scale levels of both porosity and structure in one step (Figures 2A–2C, S3A, and S3B), thereby making the catalyst development and reaction engineering (for enhanced heat/mass transfer) go hand in hand (Wang et al., 2016; Li et al., 2015). The InNi3C0.5/Al2O3/Al-fiber catalyst was tested for the RWGS reaction in a tubular fixed-bed reactor. As expected, this catalyst always achieves high CO2 conversions very close to the thermodynamic equilibrium values with above 97% CO selectivity under the wide reaction conditions (Figures 2D–2F). For example, a 53% CO2 conversion is obtainable, quite close to the equilibrium value of 54%, at 540°C and a gas hourly space velocity (GHSV) of 54,000 mL gcat −1 h−1. This catalyst delivers a very high intrinsic activity with a turnover frequency (TOF) of 11.0 CO per active site per second at 540°C (see detailed TOF calculation in Supplemental Information), almost one to two orders of magnitude higher than that seen with most platinum/oxide and non-noble-metal catalysts (Table S1). Furthermore, a kinetic study was carried out over the InNi3C0.5/Al2O3/Al-fiber catalyst, and the apparent activation energy was calculated with the result as shown in Figure S3D. InNi3C0.5/Al2O3/Al-fiber provided a much lower E a (60 kJ/mol) than Cu/ZnO-based catalysts (112 kJ/mol, Schumann et al., 2015), further indicating that this catalyst has a high intrinsic activity. Also encouraging is the exclusive CO selectivity (above 98%) with pressure increasing from 1.0 to 4.0 MPa at 540°C (Figure 2F), despite the fact that CH4 formation is much favorable at high pressure over the conventional Ni-based catalysts (Wu et al., 2015; Gonçalves et al., 2017; Li et al., 2015).Stability is a significant consideration for catalysts in practical applications. Our InNi3C0.5/Al2O3/Al-fiber catalyst is very stable with 52%–53% CO2 conversion and 97%–99% CO selectivity throughout the entire 150 h testing at a GHSV of 54,000 mL gcat −1 h−1 and 540°C (Figure 2G). Even at a high GHSV of 300,000 mL gcat −1 h−1 and 600°C, the InNi3C0.5/Al2O3/Al-fiber catalyst also shows a high stability with no deactivation sign throughout 65 h testing (Figure 2H). In comparison, the Cu/β-Mo2C catalyst maintains 85% of its initial activity after 40 h reaction and the Cu/ZnO/Al2O3 catalyst loses more than 60% of its initial activity within 15 h reaction under the identical reaction conditions (Zhang et al., 2017). It is not surprising that the InNi3C0.5 crystalline phase, surface morphology, and structure of the used catalysts are preserved unchanged (Figures S3E–S3H), consistent with the excellent activity/selectivity maintenance in Figures 2G and 2H. To the best of our knowledge, the InNi3C0.5 intermetallic has never been used before for any application in catalysis, and herein we discover its superior RWGS performance—including CO2 conversion, CO selectivity, and especially high-temperature stability—over the reported state-of-the-art catalysts (Table S1).To definitely identify the crystal structure and composition of the as-formed carbide-intermetallic from In-Ni intermetallics, such pure carbide-intermetallic was synthesized via fully carburizing InNi3, and its X-ray diffraction (XRD) pattern completely coincides with the one of InNi3C0.5 that has an anti-perovskite-type structure (Joint Committee on Powder Diffraction File No. 28-0468; Figure 3 A and Table S2). Moreover, the In:Ni:C molar ratio of the as-synthesized InNi3C0.5 was determined to be 1:2.99:0.49 (see elemental analyses in Supplemental Information), quite close to its stoichiometric ratio. Figure 3B shows its structural model containing eight InNi3 units. For each unit, eight In atoms occupy the eight corners and six Ni atoms occupy the six face centers; four C atoms randomly disperse in these eight body centers, but with the most stable configuration in a regular tetrahedron (Figure S4). The Wulff equilibrium shape of the InNi3C0.5 nanocrystal was further optimized, and its optimum shape exposes fourteen surfaces consisting of eight hexagons and six squares (Figure 3C). The InNi3C0.5(111) is the most stable surface of the hexagonal shapes with the lowest surface free energy (Table S3). Interestingly, high-resolution transmission electron microscopy (TEM) also displays an approximate hexagonal morphology of the real synthetic InNi3C0.5 nanoparticles (Figures 3D, 3E, and S5), and the lattice spacing of 0.218 nm is assignable to the InNi3C0.5(111) surface.In the last decade, significant advances have been achieved in the atomistic-theoretical calculations, enabling us to computationally construct molecular and crystalline structures and to reveal the reaction pathways on the catalyst surface at atomic-molecular level (Nicholson et al., 2014; Qin et al., 2018; Studt et al., 2014; Mao et al., 2017). Therefore, the RWGS reaction mechanism on InNi3C0.5 is first investigated by the density functional theory (DFT) calculations. We selected the most stable InNi3C0.5(111) as the ideal surface and established the dual active sites (h1: Hollow(3Ni-In); h2: Hollow(3Ni-C); Figure 4 A) from nine kinds of possible active sites (see detailed results in Table S4). As shown in Figure 4B, the CO2 molecule is chemically adsorbed via a bending configuration to form CO2* on h1 site, and the H2 molecule spontaneously dissociates into H* that can be adsorbed on both h1 and h2 sites. Electron density distribution for the dual active sites is richer than the others, which makes them more nucleophilic and more favorable for CO2 activation (Figure S6). Therefore, the CO2* facilely dissociates into CO* adsorbed on h2 site and O* adsorbed on h1 site with moderate exothermicity (namely, reaction energy E r, −0.38 eV) and a low activation barrier (E a, 0.32 eV), but with higher E a of CO2* hydrogenation to formate (HCOO*, 0.42 eV) and to carboxyl (COOH*, 0.75 eV, Figure S7 and Table S5). Clearly, the CO2* dissociation to CO* and O* (i.e., redox pathway) is preferred over the formate and carboxyl pathways on the InNi3C0.5(111) surface. Furthermore, the formed O* on h1 site preferably reacts with H* on the neighboring h2 site to produce an OH* group (E a, 0.73 eV), and subsequently, two OH* groups on the dual sites are easily transformed into H2O* (E a, 0.25 eV) that is finally desorbed into the gas phase (E a, 0.35 eV). The dual active sites provide much lower E a than the sole h1 sites for the above-mentioned steps (see detailed results in Table S5), probably the consequences of appropriate adsorption of reaction intermediates in terms of their adsorption strength (Table S4) and the distance between them (the dual active sites have shorter adjacent h1-h2 distance of 3.106 Å than the sole h1 sites with an adjacent h1-h1 distance of 5.345 Å, Figure 4A). In contrast, CO2* dissociation on Cu(111) becomes endothermic (E r, +1.06 eV, thermodynamically unfavorable) and is kinetically unfavorable (E a of 1.55 eV versus 0.32 eV on InNi3C0.5(111), Figures 4B, S8, and S9).The formed CO* either undergoes further hydrogenation to CH4 and/or CH3OH or desorbs into the gas phase. Figure 4C shows that CO* desorption overcomes a slightly higher E a of 1.36 eV at 0 K than the formation of CH4 (CH3*-to-CH4*, 1.27 eV) and CH3OH (CO*-to-HCO*, 1.05 eV), clearly exhibiting a possibility of CH3OH formation (see detailed results and discussion in Figures S10 and S11). It should be noted, however, that CO* desorption is thermodynamically more favorable at elevated temperatures (Figures S12 and S13) owing to the significant entropy contributions (Graciani et al., 2014), and therefore CO* is preferentially desorbed into gas phase rather than hydrogenated into CH3OH at our real RWGS temperature of 420°C –600°C (see experimental results in Figures 2D–2F).To verify the RWGS reaction pathway on InNi3C0.5 from experimental perspective, the in situ Fourier transform infrared (FTIR) spectroscopy analysis was carried out on pure InNi3C0.5 in a continuous H2/CO2/N2 (molar ratio of 66/22/12) flow at ambient pressure. As shown in Figure 5 A, the linear adsorbed CO* species are formed from CO2 dissociation even at 50°C, evidenced by infrared (IR) bands (Martin et al., 2016) at 2132, 2107, 2094, 2077, and 2055 cm−1. Along with the increase in the temperature, the IR band intensity of linear adsorbed CO* becomes slightly stronger from 50°C to 175°C, remains almost unchanged from 200°C to 250°C, and then diminishes until disappearance at 325°C. In addition, two new bands at 1942 and 1824 cm−1 assignable to the bridge-absorbed CO* species (Dou et al., 2017) are observed at 100°C while becoming stronger and stronger along with the temperature. Plentiful gaseous CO starts to be detected only at 300°C, and its formation is favored with the temperature. Neither CH4 (at 3013 cm−1) (Dou et al., 2017) nor formate and carboxyl species (at 1281 and 1360–1600 cm−1) (Dou et al., 2017) are detectable in the whole temperature range studied, coinciding with the DFT-suggested preferable formation of CO over CH4, formate, and carboxyl. It should be also noticed that no adsorbed CO2* species are detectable; a possible explanation is that the CO2 adsorption-dissociation is too fast to be monitored by IR, also coinciding with the DFT-indicated very low E a of only 0.32 eV for CO2* dissociation. These IR spectra undoubtedly validate the DFT results: CO2 can be efficiently converted to CO via redox pathway rather than formate and carboxyl ones.Moreover, DFT calculations on InNi3C0.5(111) surface predict the possibility of CH3OH formation (Figures 4C and S10 and Table S5). CO* is first hydrogenated into HCO* (E a, 1.05 eV), which is easily hydrogenated into CH2O* (E a, 0.32 eV); CH2O* can be continuously hydrogenated into CH2OH* (E a, 0.65 eV) or CH3O* (E a, 0.60 eV); however, CH2OH* is more favorably hydrogenated into CH3OH* (E a, 0.88 eV) over CH3O* to CH3OH* (E a, 1.71 eV). Therefore, we infer that CH3O* should be detectable by IR owing to its high accumulation and that CH3OH can be formed through the CO*-to-HCO*-to-CH2O*-to-CH2OH*-to-CH3OH* pathway (see detailed results and discussion in Figure S10). Indeed, CH3O* with IR band at 1033 cm−1 are detectable at 200°C –325°C (Figure 5B), whereas CH3OH is detected by the on-line mass spectrometry (MS) at 220°C–310°C accompanied by gaseous CO formation above 300°C (Figure 5C). Notably, the absence of CH2O* and CH2OH* in in situ IR spectra is probably a consequence of the low residence time of these species on the surface under atmospheric conditions (Graciani et al., 2014). These IR and MS spectra consistently display that CO* is hydrogenated into CH3OH highly selectively below 300°C, whereas it is dominantly desorbed into gas phase above 300°C.The above-mentioned DFT and FTIR results also make us confident that the InNi3C0.5 nano-intermetallic is a potential catalyst for the CO2 hydrogenation to CH3OH, which becomes more and more competing in recent years. With reaction temperature reduced from 400°C–600°C (for the RWGS reaction) to 300°C and below, the InNi3C0.5/Al2O3/Al-fiber indeed turns itself suddenly into a CO2-to-CH3OH catalyst, being capable of converting 1%–8% CO2 into CH3OH with 60%–98% selectivity (corresponding to the CH3OH space time yield of 70–330 gMeOH kgcat −1 h−1) at 200°C –300°C (Table S6). The preferable CH3OH formation rather than CO formation below 300°C is attributed to the fact that low temperatures thermodynamically favor further hydrogenation of CO* to CH3OH* (Figures S12 and S13). These results exhibit an interesting temperature-dependent selectivity switching for CO2 hydrogenation.Moreover, in the light that CO2 molecule has the carbonyl property and InNi3C0.5 intermetallic can efficiently activate CO2 molecule, we wonder whether this catalyst is favorable for other carbonyl-compounds transformation, such as the hydrogenation of aldehydes/ketones/esters to corresponding alcohols. To avoid the adverse influence of acid groups on the surface of Al2O3, we directly supported the InNi3C0.5 nano-intermetallic onto a thin-sheet Ni-foam substrate with 110 pores per inch (Figure S14, see detailed preparation in Supplemental Information). Indeed, the InNi3C0.5/Ni-foam catalyst presents the satisfying activity and high product selectivity (Tables 1 and S7), providing the general and efficient ability to activate the C=O bond for carbonyl-to-hydroxyl transformation. Notably, ethylene glycol (EG) is an important commodity chemical, used for polyester manufacture, anti-freeze compounds, and solvents (Yue et al., 2012), and the gas-phase hydrogenation of dimethyl oxalate (DMO) to EG (its commercialization is on the way) is an attractive alternative EG synthesis using syngas (Fenton and Steinwand, 1974) derived from non-oil resources (such as coal, natural gas, and biomass) even from CO2 through the RWGS reaction. This foam-structured catalyst is capable of completely converting DMO at a high EG selectivity of 96% with a promising stability (Table 1). Moreover, the InNi3C0.5/Ni-foam also shows favorable RWGS and CO2-to-CH3OH performances that are comparable with those seen with the InNi3C0.5/Al2O3/Al-fiber (Tables S8 and S9).In summary, we have discovered an outstanding nano-intermetallic InNi3C0.5 catalyst system via RWGS-reaction-oriented pre-design combined with atomistic-theoretical calculations and experimental verifications. Practical fiber/foam-structured InNi3C0.5 nano-intermetallic catalysts engineered from nano- to macro-scale in one step have been developed, achieving unprecedented performance in the RWGS reaction and showing potential to catalyze CO2 hydrogenation to CH3OH. Most notably, such nano-intermetallic catalysts are also highly active, highly selective, and highly particularly stable for the DMO-to-EG process (EG synthesis using syngas derived from non-oil resources even from CO2 through the RWGS reaction). We anticipate our essay to be a new point closer toward the ultimate goal of catalysis, namely, designing and tailoring the catalysts atom by atom with precise structure, and our findings might lead to commercial exploitation of such kind of nano-intermetallic catalysts for applications in highly efficient reduction of CO2 to CO as well as carbonyl-to-hydroxyl transformation.The large-scale H2 production should be from the renewable solar, hydraulic, and wind energy.All methods can be found in the accompanying Transparent Methods supplemental file.We acknowledge the financial supports from the National Natural Science Foundation of China (grants 21773069, 21703069, 21703137, 21473057, U1462129, 21273075), the Key Basic Research Project (grant 18JC1412100) and Shanghai Pujiang Program (grant 17PJ1403100) from the Shanghai Municipal Science and Technology Commission, and the National Key Basic Research Program (grant 2011CB201403) from the Ministry of Science and Technology of the People's Republic of China. We thank Prof. Dr. Roel Prins from the ETH Zurich for fruitful discussion.P.C., G.Z., X.-R.S., and Y.L. conceived the idea for the project and designed the experiments; P.C., G.Z., X.-R.S., and Y.L. carried out the interpretation and wrote the manuscript; P.C., J.Z., and J.D. conducted the material synthesis, characterizations, and catalysis tests; P.C. and X.-R.S. performed the structural analysis and modeling; X.-R.S. carried out the DFT calculations; all authors discussed and commented on the manuscript; Y.L. directed the research.Y.L., P.C., J.Z., and G.Z. have a patent application related to this work filed with the Chinese Patent Office on 15 October 2017 (201710956080.1). The authors declare no competing interests.Supplemental Information can be found online at https://doi.org/10.1016/j.isci.2019.07.006. Document S1. Transparent Methods, Figures S1–S14, and Tables S1–S11

CO2 circular economy is urgently calling for the effective large-scale CO2 reutilization technologies. The reverse water-gas shift (RWGS) reaction is the most techno-economically viable candidate for dealing with massive-volume CO2 via downstream mature Fischer-Tropsch and methanol syntheses, but the desired groundbreaking catalyst represents a grand challenge. Here, we report the discovery of a nano-intermetallic InNi3C0.5 catalyst, for example, being particularly active, selective, and stable for the RWGS reaction. The InNi3C0.5(111) surface is dominantly exposed and gifted with dual active sites (3Ni-In and 3Ni-C), which in synergy efficiently dissociate CO2 into CO* (on 3Ni-C) and O* (on 3Ni-In). O* can facilely react with 3Ni-C-offered H* to form H2O. Interestingly, CO* is mainly desorbed at and above 400°C, whereas alternatively hydrogenated to CH3OH highly selectively below 300°C. Moreover, this nano-intermetallic can also fully hydrogenate CO-derived dimethyl oxalate to ethylene glycol (commodity chemical) with high selectivity (above 96%) and favorable stability.

Solid oxide fuel cells (SOFCs) have successfully demonstrated flexibility of the utilization of multiples of fuels ranging from syn-gas, bio-gas, natural gas and other hydrocarbons to pure hydrogen [1,2]. Conventionally, hydrocarbons are externally reformed and reformate serves as fuel for electrochemical oxidation on the cell anode [3]. Direct internal reforming in SOFC, on the other hand, allows the hydrocarbons to be simultaneously reformed and electrochemically oxidized at the anode, resulting in high conversion and efficiency for electrochemical performance improvement, cost reduction and thermal management by combining exothermic oxidations with endothermic reformation reactions [4,5]. Ni-based anode in conventional SOFC provides high electronic conductivity and electrocatalytic activity, but suffers from excessive cooling and coke formation due to rapid endothermic reforming and thermal cracking of hydrocarbons [5,6]. This leads to a steep reduction in temperature especially at the inlet of the cell stack, resulting in a non-uniform temperature distribution along the cell [7,8]. A large temperature gradient along the cell surface may cause high mechanical instability and thermal stress between the anode and the solid electrolyte, leading to inevitable cell fracture and spallation [9]. Besides changes in the mechanical properties of cell and stack components leading to failure, the role of local temperature on subsequent reforming reaction and electrochemical reaction rates, as well as ionic conductivity of the electrolyte are influences and should not be overlooked [10]. To mitigate these challenges, significant efforts have been directed towards the development of electrochemically active anode materials with uniform temperature distribution and high coking resistance [11,12].Hydrocarbons and reformate gas mixtures have been extensively used in commercial internal combustion engines and fuel cells for power generation [13]. Fig. 1 compares the energy densities of a number of sources of energy, where H2 shows the largest mass energy density, but the lowest liquid volumetric energy density due to storage and transportation challenges. Approximately 95% of H2 is currently being produced by steam reforming of methane (SRM) or partial oxidation of methane as well as gasification of coal [14,15]. Since a SOFC typically operates in the 600–1000 °C temperature range, relatively high operating temperature favors SRM on the cell. Additionally, the heat produced from the electrochemical reaction at the anode is used to promote the SRM reaction while the exothermic WGS reaction takes place concurrently to produce CO2 and more H2. Hence, the main advantages of utilizing internal steam reforming in SOFC include lower operational cost and higher thermal efficiency [16,17].The choice of electrocatalyst catalyst and anode configuration plays an important role on the long term stability of fuel cell operation. That is, the catalyst must exhibit high catalytic activity and stability under industrial operating conditions. The state-of-the-art catalyst for SRM utilizes precious metals such as Pt and Rh, which are considerably expensive and scarce. Ni-based catalyst is widely used owing to its comparable reforming performance to that of precious metals [19] as well as cost effectiveness and availability [20]. Despite these advantages, Ni-based catalysts deteriorate very quickly due to Ni sintering and coking [21]. At such high temperatures during SRM, carbon formation can cause rapid catalyst deactivation. The two types of carbon that can form on a catalyst surface are encapsulated and filamentous carbons [22]. The latter, although does not deactivate the catalysts, is highly responsible for mechanical failure and increase pressure drop in the reactor, especially in SOFCs. Equation (1) represents carbon formation by CH4 cracking. Subsequently, equation (2) refers to the Boudouard reaction, which is another possible route to form carbon during SRM. (1) C H 4 → C + 2 H 2 ( Δ H 298 ° = 76 k J / m o l ) (2) 2 C O → C + C O 2 ( Δ H 298 ° = - 172 k J / m o l ) Studies have shown thermodynamically that increasing the S/C ratio can reduce coke deposition, consequently leading to higher conversion [23,24]. It is worth nothing, however, that the introduction of excess steam may lead to higher energy demand and operating cost, as well as lower H2 yield [25].To alleviate the aforementioned challenges faced with Ni anode and reforming catalyst, our approach focused on the development of multi constituent alloys, also known as high-entropy alloys (HEAs), as anode and SRM catalysts. HEAs are promising alloys that combine five or more metal elements to improve the catalytic and mechanical properties [26,27]. One metal of consideration is cobalt (Co). Reports indicate that Co exhibits relatively high affinity for oxygen species, which is beneficial for suppressing carbon formation [28,29]. Besides an effective oxidizing catalyst, it has been observed that Co also promotes WGS reaction to produce more syngas, while simultaneously inhibit the Boudouard reaction responsible for carbon formation [30]. Copper (Cu) is another common metal additive for SRM catalysts. Huang et al. demonstrated that the addition of Cu to Ni catalysts promotes the WGS reaction activity [31]. Despite this, Cu is known as a poor catalyst for C–C and C–H scission, thus slowing the rate of carbon formation [32]. DFT studies have confirmed that the incorporation of Cu results in higher activation energy barrier (E act ) of carbon formation, while still maintaining an acceptable rate of reforming [33]. Hence, Cu is used to slow down the reforming rate, since the highly endothermic reaction could cause rapid cooling and consequently, thermal stress on the SOFC anode [5]. Besides chemical activity and stability, the SRM catalyst must also exhibit good physical stability and durability under industrial operating conditions. The use of Ni as an SOFC anode at high temperatures for long durations may undergo sintering and particle coarsening [34].One plausible approach is to add metal additives with high melting point. Fe has been shown to be thermally stable at high temperatures, which makes it candidate for SOFC anode material [5]. Huang et al. reported that Fe possesses strong resistance against carbon formation during internal ethanol reforming in SOFC [35]. Due to the high affinity of Fe for oxygen species, the surface carbon can be easily oxidized to CO and subsequently CO2 to avoid catalyst deactivation and further promote the WGS reaction. Similarly, manganese (Mn) has shown to be a beneficial oxidation catalyst without the risk of sintering or agglomeration [36]. This is advantageous as oxygen can transfer to the carbonaceous species and frees the surface from carbon deposition. Ouaguenouni et al. prepared a nickel-manganese oxide catalyst that exhibits good activity towards the complete oxidation of methane [37]. The ability of Mn to exist in different oxidation states makes it a good redox couple catalyst for SRM [38]. Table 1 summarizes the catalytic role of each metal in SRM and the corresponding drawbacks. In this study, the five metals discussed above have been consolidated in a solid solution known as the high-entropy alloy (HEA) as means to utilize the advantageous properties of each metal, while keeping thermal stress, endothermic cooling and rate of carbon formation minimal. Contrary to other fuel cell systems, the main challenge with SOFC does not concern with mass transfer or kinetics, but rather long term-stability, for which internal distributed reforming of hydrocarbon plays a key role. Long-term stable cell and stack operation require that the cell experience and possess distributed reforming and endothermic cooling as well as resistance to carbon formation. By controlling the catalytic reaction of the anode, a thermal neutral state can be achieved as a result of both the endothermic steam reforming reaction and the exothermic electrochemical oxidation reactions [39,40]. It has been shown computationally using a 3D CFD model that the reforming rate should be reduced by a factor of 0.01 relative to that of Ni-based anode for a more uniform temperature distribution along the cell [41]. With the development of advanced anode, our objective is to reduce the reforming rate without significantly lowering the electrochemical activity of the cell, so that adequate current density can still be maintained. Thin film studies performed on sputter deposited above alloy compositions indicated the formation of solid solution (R. Bhattacharya, UES Inc. Personal communication). At elevated temperatures and under the SOFC operating conditions, it is envisioned that select alloy constituent can oxidize to form respective oxide based on the local oxygen partial pressure of the fuel. For the support, gadolinium- doped ceria (GDC) was used. CeO2 is widely used as a support for SRM due to its oxygen storage capacity (OSC) to store and release oxygen species [42,43]. Additionally, Ce-based materials present high oxygen ion mobility that promotes carbon removal and hence, long-term stability of the cell [44]. The addition of Gd increases sintering resistance by enhancing the metal-support interaction [45]. HEAs with various metal contents supported on GDC were prepared and tested for SRM. Then, direct internal steam reforming in laboratory scale SOFC button cells were performed to examine the performance of HEA/GDC as a candidate anode. The reforming and electrochemical measurements, resistance to carbon formation were analyzed and compared to those of conventional Ni/YSZ and standard Ni/GDC anode.HEA was prepared using the co-precipitation method by dissolving optimized formulation of nitrate precursors obtained from Fisher Scientific (98% pure nickel (II) nitrate hexahydrate, 99% pure cobalt (II) nitrate hexahydrate, 99% pure copper (II) nitrate trihydrate, 98% iron (III) nitrate nonahydrtae and 98% manganese (II) nitrate tetrahydrate). A total of three different anode catalyst formulations were synthesized and tested for methane reforming, from which the alloy mixture with resistance to carbon formation and stable reforming was selected as the SRM catalyst for further bench-top and fuel cell studies. The resulting optimized formulation of the HEA anode is given in Table 2 . The metal nitrates were dissolved in excess deionized (DI) water, stirred and heated to 90 OC. Then, citric acid (CA) was added as a chelating agent to the mixture using a 1.5:1 CA: metal ratio. Ammonium hydroxide (NH4OH) solution was added dropwise to the metal-chelate solution to adjust the pH value to about 7–9 while stirring. The solution stirred overnight to homogenize the mixture and to evaporate excess water. On the next day, the remaining solution was transferred to an alumina crucible for calcination at the rate of 5 OC/min to 500 OC and held for 6 h to burn off nitrates, organic compounds and other contaminants. The as-obtained HEA powder and commercial 10% GDC (GDC-10 M) obtained from Fuelcell Materials USA were weighed (65:35 wt%) and physically mixed in a mortar and pestle until a homogenous mixture of fine powder was obtained. Table 2 provides the metal composition of each SRM catalyst for this study.N2 adsorption/desorption analysis was conducted using a Micromeritics ASAP 2000 analyzer to determine the sample surface area, pore volume and pore distribution. Before the analysis, about 0.1 g of sample was outgassed for 12 h under vacuum in the degas port. Then, the sample was re-weighed to obtain the new moisture-free mass before starting the analysis. The measurement was carried out at 77 K under N2 flow. The Brunauer-Emmett-Teller (BET) theory was then used to calculate the surface area. H2 chemisorption was performed using the Micromeritics ASAP 2000C software to determine the metal dispersion. Powder X-ray diffraction (XRD) pattern of each sample was collected using a Bruker D8 Advance X-ray diffractometer to identify surface phases. The diffractometer was equipped with a Cu Kα radiation (λ = 0.15406 Å) operating at 40 kV and 40 mA. The XRD patterns were obtained in a 2θ range of 10–90°. The scanned XRD patterns were indexed using the ICDD (International center for Diffraction Data) database. Surface morphology and elemental composition of each sample before and after the SRM experiment were characterized using a FEI Quanta 250 FEG scanning electron microscope (SEM) coupled with energy dispersive E-ray spectroscopy (EDAX). H2 temperature-programmed reduction (TPR), oxidation (TPO) and desorption (TPD) were carried out on an Altamira Instruments AM1-200 unit. About 50 mg of sample was placed between quartz wool supports inside a U-shaped quartz tube. Prior to the TPR and TPD analyses, the sample was first pretreated in 10% O2/He gas at a flow rate of 30 SCCM from 50 to 1000 °C and heating rate of 10 OC/min. TPD was carried out under inert atmosphere in pure Ar flow. Reduction experiments were performed using 50 SCCM of 10% H2/Ar. After reduction, the gas feed was subsequently switched to 50 SCCM of 10% O2/He for TPO study. All TPR, TPO and TPD studies were analyzed using a thermal conductivity detector (TCD). Inductively coupled plasma (ICP) with an optic emission spectroscopy (ICP-EOS) was used to quantify the bulk metal loadings of each catalyst. Post-test samples were also characterized for carbon formation by a Renishaw System 2000 equipped with a 514 nm green laser.The SRM test was performed in the temperature range of 700–800 °C at 1 atm and gas hourly space velocity (GHSV) of 45,000 h−1.100 mg of SRM catalyst was loaded into a fixed-bed quartz tube with an outside diameter (OD) of ½” and a length of 38 cm as shown in Fig. 2 . Both sides of the catalyst were supported by quartz wool. The reactor was then placed into a horizontal tube furnace. Prior to the test, the catalyst was first reduced in a constant 4% H2/N2 gas flow at 700 OC for 2 h. Then, the gas was switched to flow 10 SCCM of CH4 and allowed to mix with 20 SCCM of H2O inside an evaporator heated to 120 °C before entering the catalyst bed. H2O was supplied by an HPLC pump at a flow rate of 0.016 mL/min to maintain a steam-to-carbon ratio (S/C) of 2.20 SCCM of N2 was used as a carrier gas, amounting to a gas hourly space velocity (GHSV) of approximately 45,000 h−1. Exhaust gas was condensed, collected and analyzed by a SRI 8610 gas chromatograph with a helium ionization detector (HID). Upon completion, the reactor and gas lines were purged with N2 gas and then switched back to H2 before cooling the reactor down to room temperature. The post-test samples were carefully removed from the quartz tube and quartz wool, and saved to be analyzed under SEM and Raman spectroscopy for any carbon deposition on the catalyst. The methane conversion (X CH4 ) and hydrogen yield (Y H2 ) were determined using equations (3) and (4), respectively. The rate of CH4 consumption (r CH4 ) normalized to the active metal loading was calculated by equation (5). A time-on-stream (TOS) test was conducted at 600 °C for 30 h to investigate the stability of SRM catalysts towards carbon poisoning. Except the operating temperature, the same operating conditions for the bench-top test was adopted. The TOS post-test samples were saved and analyzed using Raman spectroscopy. Additionally, surface morphology and elemental composition of post-test samples were characterized by SEM. (3) X C H 4 ( % ) = F C H 4 , i n − F C H 4 , o u t F C H 4 , i n × 100 % (4) Y H 2 ( % ) = F H 2 , o u t 2 F C H 4 , i n + F H 2 O , i n × 100 % (5) R CH 4 ( mol CH 4 mol metal - 1 s - 1 ) = F C H 4 , i n − F C H 4 , o u t N m e t a l Where F i is the molar flow rate of species i in mol/s and N metal is the amount of active metal in moles.An electrochemical button cell (HEA/GDC - Ni/ScSZ|ScSZ|LSM/YSZ) was fabricated to examine the electrochemical performance of HEA/GDC as anode material for direct internal SOFC at 750 °C. Ni/ScSZ functional layer (10% Scandia stabilized zirconia purchased from Fuelcell Materials USA) was first deposited on the anode side and then sintered at 1350 °C for 2 h. This was followed by screen-printing LSM/YSZ cathode on the opposite side of the electrolyte and sintering at 1200 °C. The final anode layer of HEA/GDC was screen-printed on top of the anode functional layer and subsequently sintered at 1000 °C. The cell performance was evaluated at 750 °C on an in-house test station shown in Fig. 3 . The cell was sealed using CeramaBond on one end of an alumina tube and the gold meshes were used as current collectors. To create a base line and reduce the anode, humidified hydrogen (9% H2–3% H2O–N2 bal.) was supplied to the anode side at a flow rate of 100 SCCM. The corresponding I vs. T for 3 h is shown in Fig. S1 in the supporting information. Subsequently, the gas was switched to CH4 fuel with steam (S/C = 2) before entering the catalyst bed on the anode side. Unlike the bench-top experiment, a carrier gas was not used in this case to ensure low mass transfer resistance and maximum contact between the active area of the catalyst and methane. The resulting GHSV was similar to that used in the bench-top experiment. Then, the exhaust gas was condensed, collected and analyzed by the GC-HID. Air was fed through the cathode at a flow rate of 150 SCCM. The electrochemical test, similar to the bench top test, was carried out for 30 h. Current density and the electrochemical impedance spectra (EIS) measurements were acquired at a constant voltage of 600 mV using a VMP3 Bio-Logic potentiostat/galvanostat. The frequency ranged from 10 mHz to 200 kHz, with 10 mV perturbation. For a more quantitative insight into the electrochemical phenomena, electrical equivalent circuit (EEC) RΩ(QRHF) (QRLF) was employed using ZSimpWin software to analyze the impedance data. The impedance of two interfaces metal/electrolyte and surface coating/electrolyte were analyzed to represent the two semicircles corresponding to the high and low-frequency arcs, respectively, which relate to gas adsorption–desorption on the electrode surface followed by charge transfer and incorporation of adsorbed gas at the three phase boundary, and the gas concentration polarization loss of the electrode. To investigate a carbon-free cell operation, SEM and Raman spectroscopy on the post-test cell were conducted.In this study, the co-precipitation method was employed to synthesize the HEA anode material. To confirm if a single-phase alloy was formed, room-temperature powder XRD pattern was performed. Fig. 4 (a) presents the XRD patterns of HEA/GDC as well as those of Ni/GDC and Ni/YSZ for comparison. The intensity of the XRD peaks is directly correlated to the crystallinity of the material. As expected, the crystallinity of reduced Ni/YSZ and Ni/GDC is more pronounced than that of the HEA/GDC due to less chemical and heat treatments of the former materials, thereby preserving the integrity of the crystal. Indexing by ICDD reveals that the catalysts were successfully reduced, as evidenced by the absence of oxide peaks. For the HEA/GDC catalyst, the denoted peaks are attributed to mixed metal alloys. This confirms that HEA was successfully synthesized without additional phases of oxides. The other diffraction peaks have also been indexed and confirmed by ICDD to denote the respective metal supports. Fig. 4(b) shows the N2 adsorption/desorption isotherms of the SRM catalysts and the calculated BET surface area and pore volume are tabulated in Table 3 , along with other physicochemical properties. The linear relationship at the beginning of the isotherm, followed by a significant increase in the adsorption of N2 indicates a type II physisorption isotherm, suggesting a nonporous structure. The addition of GDC as support increased the surface area. According to Angeli et al. the presence of CeO2 improves the surface area and active metal dispersion [3]. Subsequently, the substitution of Ni with HEA supported on GDC further increased the surface area to 35 m2/g due to enhanced pore size volume, which may enhance the catalytic properties of SRM.SEM images of as-synthesized Ni/YSZ, Ni/GDC and HEA/GDC are shown in Fig. 5 . The standard Ni/YSZ catalyst containing 45.2 wt% of Ni (Table 3) in difference resolutions is shown in Fig. 5(a–c). It can be seen that the Ni particles are relatively small and close to each other. H2 chemisorption reported a Ni dispersion of 0.327% with a particle size of 310 nm. For Ni/GDC, the structures of NiO and GDC powders were dissimilar and could be easily distinguished from each other as seen in Fig. 5(d–f). The metal dispersion was slightly lower due to the increase in crystal size to 461 nm, which suggests that the increase in surface area could be attributed to the enhanced pore volume, owing to the GDC support. Fig. 5(g–i) show two distinct phases on HEA/GDC, arising from the presence of HEA and the GDC support. From the morphology, it can be seen that the particles tend to sinter and form larger agglomerates. This, however, changes as the HEA/GDC catalyst was reduced at higher temperature as the oxide phase converts into the FCC cubic phase, as shown by the XRD pattern in Fig. 4(a). Subjecting the catalyst to reduction may also result in higher porosity and smaller metal particles, leading to increased surface area.The coking resistance and SRM performance of a catalyst highly depends on interaction between the active metal and the support. Having a strong metal-support interaction in the catalyst suppresses metal sintering at elevated temperature and reduces coke formation, which in turn improves catalyst activity and stability [22,58]. To evaluate the chemical interaction between metal and support, the SRM catalysts were analyzed by H2-TPR as shown in Fig. 6 (a). The reduction peak centered at 375 °C was assigned the reduction peak of NiO to Ni, which resulted from a weak interaction between Ni and the support [59]. During the TPR of HEA/GDC, it was observed that a broad peak emerged at 400 °C due to the reduction of the metal alloy. The reduction peaks of Co3O4 typically appear at approximately 400 °C and 470 °C, following a two-step reduction process to Co0 [50]. Similarly, the two-step reduction of Mn2O3 to Mn3O4 and subsequently, to MnO would result in reduction peaks at 300 °C and 420 °C, respectively [60]. Finally, the reduction of Fe2O3 to Fe also follows a two-step process, although the reduction of Fe3O4 to Fe0 occurs at a much higher temperature of 835 °C [61]. The successful synthesis of HEA brings about a single-phase solid solution through which the compositions of five metals have been optimized. As a result, the properties of HEA are typically more superior than the corresponding metal counterparts. Such is the case in Fig. 6(a) showing that HEA/GDC requires a higher temperature for reduction compared with Ni/YSZ and Ni/GDC. Thus, the HEA/GDC catalyst should exhibit higher sintering and coking resistances. TPO of reduced samples is displayed in Fig. 6(b). Oxygen uptake appears to be minimal or non-existent, which can be explained by the rapid re-oxidation of metals and oxide supports during the switching between reducing and oxidizing gases. Nonetheless, the desorption of O2 takes places at roughly 500 °C for Ni/YSZ and Ni/GDC, and ∼600 °C for HEA catalysts under oxidizing conditions. The higher desorption temperature suggests a stronger interaction between HEA and oxygen. Fig. 6(c) displays the TPD results of samples after being subjected to oxygen pretreatment. As expected, Ni/YSZ and Ni/GDC did not show any desorption of oxygen, while HEA/GDC exhibited a TPD peak at ∼850 °C. This suggests that some HEA constituents such as Fe2O3 or Mn2O3 possess high oxygen storage capacity (OSC) and they have been proven beneficial for carbon removal. Quantification of TPO and TPD peaks of these samples in Table 3 shows that HEA/GDC is capable of adsorbing and desorbing a higher amount of oxygen, owing to the enhanced surface oxygen mobility and oxygen uptake of HEA and the GDC support [62].The catalytic activity of SRM catalysts was examined using a fixed-bed tube reactor at varying operating temperatures of 700, 750 and 800 °C at a GHSV of 45,000 h−1. At each temperature, the reaction was allowed to reach equilibrium, after which methane conversion was calculated and reported in Fig. 7 (a). The main products in the exhaust were H2, CO, CO2 and CH4. Water vapor in the exhaust was condensed before being fed to the GC. All SRM catalysts showed increasing conversion with temperature, with both nickel-based catalysts (Ni/YSZ and Ni/GDC) displaying the highest methane conversion at equilibrium. This is expected as the SRM reaction is endothermic and therefore, thermodynamically favorable at higher temperature. It is also worth mentioning that while partial oxidation of CH4 may occur subsequently with steam reforming, our calculations show that the partial pressure of O2 is too low (i.e. ∼ 10−17 atm) for CH4 to undergo direct oxidation, consistent with other reports [63]. The lowest conversion was reported by HEA/GDC, which increased from 27% at 700 °C to 35% at 750 °C and then to 42% at 800 °C. Subsequently, the increase in conversion is accompanied by an increase in hydrogen yield with temperature, as displayed in Fig. 7(b). Among the HEA catalysts, HEA/GDC reported the highest conversion at each temperature. The H2 yield was calculated by measuring the amount of H2 produced with respect to theoretical amount of H2 produced from maximum conversion of CH4 and H2O. In all cases, the H2 yield increased with temperature with both Ni/YSZ and Ni/GDC displaying slightly higher H2 yield than HEA/GDC. The rate of CH4 consumption was also calculated and compared as shown in Fig. 7(c). The increased temperature enhanced the consumption rate of CH4 for all SRM catalysts. Both Ni/YSZ and Ni/GDC catalysts showed the highest rate of ∼3.5 molCH4 molNi −1 s−1 by 750 °C. The high activity of Ni in SRM has been well-documented in the literature. A strong endothermic reaction may induce large temperature gradient especially when operating the catalyst on a SOFC. Furthermore, a fast reforming reaction as displayed by the Ni-based catalysts may result in thermal stresses and mechanical failures, thus lowering the cell efficiency. The HEA/GDC catalyst, however, showed a lower reforming rate of ∼1 mol molCH4 molHEA −1 s−1 at 750 °C, which may provide a smaller temperature gradient during concurrent reactions of heterogeneous catalysis and electrochemistry. Experimental results show that an optimized formulation of HEA has reduced the reforming rate of methane compared with highly endothermic Ni-based catalysts. Consequently, the cell life and current distribution can be maintained. This is more advantageous than standard Ni/YSZ as an anode material since the latter has shown to experience major drawbacks especially under harsh conditions such as coking, metal agglomeration, thermal stress, mechanical failure and poor redox stability [3]. While higher localized reformation rates may imply faster and higher production of H2, the objective of the alloy anode development is largely to minimize localized cooling and carbon deposition.The catalytic stability of SRM catalysts was investigated at 600 °C and S/C ratio of 1. The TOS experiment was carried out isothermally for 30 h. Using the HSC® Chemistry 10 software, the equilibrium compositions were calculated for a temperature range of 25–1000 °C as shown in Fig. S2(a) in the Supporting Information. At 600 °C, the carbon activity should be at its highest and this temperature is thermodynamically favorable for studying carbon resistance of each SRM catalyst. Fig. 6(d) reports the conversion of CH4 over time. It is evident that the initial conversion rate was high for both Ni/YSZ and Ni/GDC due to enhanced catalytic activity of Ni. However, the conversion gradually decreases over time with Ni/YSZ showing the fastest degradation rate, followed by Ni/GDC. After 30 h of TOS, the final conversions were 54% and 66% for Ni/YSZ and Ni/GDC, respectively. The higher stability of Ni/GDC suggests that the GDC support plays an important role in reducing catalyst deactivation. For the HEA/GDC catalyst, the conversion rate was relatively low compared to initial conversion rates of the standard catalysts, as shown in Fig. 6. Nonetheless, the catalyst maintained a stable run over 30 h of TOS between 15 and 18% conversion, revealing the ability of HEA catalysts to resist deactivation over long periods of operation. To assess the source of catalyst deactivation, post-test catalysts were saved from TOS experiments and were subjected to Raman analysis and SEM imaging.Carbon deposition has been regarded as one of the main reasons for catalyst deactivation during SRM. To identify the nature and structure of these surface carbonaceous species, post-test catalysts from the 30 h TOS test were subjected to SEM imaging. Fig. 8 (a–c) show high resolution SEM images of post-test SRM catalysts after 30 h of TOS experiment. The surface of all samples appear to be free of any carbonaceous species. Post-test Ni/YSZ and Ni/GDC samples in Fig. 8(a and b) did not show any dissimilarities compared to their corresponding pre-test samples. HEA/GDC was observed to be more porous with a uniform distribution of particle size after reduction at 700 °C, consistent with relatively high BET surface area reported in Table 3. The absence of surface carbon on HEA/GDC in Fig. 8(c) may explain the promising stability during the 30 h of TOS. To confirm this and to further investigate the deactivation of the former two Ni-based catalysts, Raman spectroscopy was performed on all post-test samples. As shown in Fig. 8(d), all SRM catalysts showed two distinct characteristic peaks at 1345 and 1595 cm−1. The peak at 1335 cm−1 can be attributed to the D band of carbonaceous species, formed by the vibrations of disordered carbon atoms (amorphous carbon for example), while the peak at 1595 cm−1 has been assigned the G band to represent the presence of ordered and graphitic crystalline structure caused by vibration of the in-plane sp [2]-bonded carbons [64]. Amorphous carbon has been shown to play a significant role in catalyst deactivation via encapsulation of the metal active sites [65]. On the other hand, graphitic carbon with filamentous structure may also form as a result of migration of surface carbon to the bulk metal phase, resulting in nucleation growth of carbon on the other side of the metal particle [21]. While graphitic carbon may not directly affect the activity of the catalyst, uncontrolled growth of carbon whiskers may result in reactor blockage and pressure drop [66]. Additionally, coke formation on the anode material of a SOFC can be detrimental to the long-term stability of the system and could potentially lead to mechanical failure [12]. In Fig. 8(d), Ni/YSZ catalyst showed the highest amount of both amorphous and graphitic carbons on the surface, leading to catalyst deactivation during the 30 h TOS test as demonstrated in Fig. 7(d). Similarly, Ni/GDC exhibited some amorphous and graphitic carbonaceous species, which explains the gradual deactivation of the catalyst. The enhanced stability of HEA/GDC during TOS was due to the high carbon resistance of the catalyst, as both post-test SEM and Raman analyses did not show signs of carbon.The electrocatalytic activity and stability of (HEA/GDC-Ni/ScSZ|ScSZ|LSM/YSZ) for direct internal SOFC have been investigated. The moderate reformation rate and high long-term stability of the anode catalyst may prove beneficial in a SOFC system by preventing mechanical failures due to rapid temperature change and carbon deposition [5,12]. In this study, the cell test was performed at 750 °C, to which the anode was subjected a constant flow of CH4 and steam (S/C = ∼2) and the cathode with air. Under reduced atmosphere and internal reforming condition, the open-circuit voltage (OCV) was measured to be ∼0.9 V at 750 °C due to favorable interfacial interaction between the HEA/GDC anode layer and the ScSz electrolyte later after high-temperature sintering. The OCV plots can be found in Fig. S3. Upon switching the feed to CH4 fuel and imposing a bias of 0.6 V, the I-T electrochemical data was collected for 30 h of SOFC test, as shown in Fig. 9 (a). A slight dip in current density was noticed after 8 h of testing, which could be attributed to lower reformation rate, condensation of steam in cold zones of the inlet and diffusion of transition metals in the anode that could affect the ionic resistance of the cell [67]. This self-activation phenomenon has also been reported elsewhere as a result of surface modification on the anode during steam reforming [68,69]. During this time, the initial reforming rate is extremely low. However, upon reduction and activation, the reforming rate increases, leading to an increase in performance and current density. As soon as the HEA/GDC anode was fully activated after the first 15 h, the reforming rate was enhanced leading to an increase in current density to ∼100 mA/cm2 for the next 15 h. In comparison, the current density of a Ni-based anode was reported to be ∼250 mA/cm2 at the start of the cell operation, but quickly approached 0 mA/cm2 due to carbon formation [63]. The relatively low current density may be due to electrolyte thickness, whose role on electrochemical performance will be explored in future work. Nonetheless, the HEA/GDC anode yielded sufficient current density to maintain a stable and carbon-free operation. This confirms that controlled and distributed reforming also improved the current density distribution in the cell. A more comprehensive electrochemical study involving dual atmosphere cycling to compare the current densities in reducing atmosphere and hydrocarbon-rich atmosphere will be considered in the future. The corresponding Nyquist spectra in Fig. 9(b) acquired at the different times are composed of two depressed semicircles which correspond to the polarization resistance Rp (RHF + RLF) while the high frequency intercept with the real impedance axis corresponds to the purely ohmic resistance (RΩ) of the electrolyte and current collecting wires. As demonstrated in Fig. 9(c), the overall non-ohmic resistance Rp (RHF + RLF) decreases after 15 h indicating higher mass transfer due to increase in hydrocarbon reformation rate. The ohmic resistance remains stable during internal reforming indicating stable and carbon-free cell operation. The exhaust was simultaneously analyzed by a GC-HID and the conversion of CH4 over 30 h of testing is shown in Fig. 9(a). The conversion of CH4 was stable at 20% throughout the whole electrochemical test, which is consistent with the results obtained from the bench top experiments (Fig. 7). This is a good indication that the cell test is stable and scalable for future long-term testing. Once the cell test has been completed, the anode layer of the cell was analyzed using SEM microscopy and Raman spectroscopy for any carbon deposition. High-magnification SEM in Fig. 9(e) shows the presence of carbon on Ni/YSZ anode with a composition of 17.3 atomic%. On the other hand, post-test HEA/GDC anode in Fig. 9(f) shows a clean and carbon-free surface. After 30 h of cell test, not only did the anode layer show remarkable carbon resistance, but good and stable contact were also formed between the anode layer and electrolyte. Fig. 9(d) shows the Raman spectra of two locations of the anode surface, one being the center and the other towards the edge of the anode layer. The absence of D and G bands at 1345 and 1595 cm−1, respectively, suggests that the HEA/GDC anode was free of both amorphous and graphitic-typed carbons. HEA/GDC also exhibits high OSC, as suggested by TPD in Fig. 6(c), which can play an important role in the rapid oxidation of carbon to COx species, thus minimizing carbon poisoning. Fig. 10 shows a schematic depicting the role of coke-resistant HEA/GDC as oxygen vacancies in the anode enhance the mobility and diffusivity of oxygen ions to the anode surface to gasify any deposited carbon.The HEA/GDC catalyst displayed promising potential as an anode material for internal utilization of CH4 in SOFC. The bench top experiment suggested that HEA/GDC exhibits moderate reforming rate and excellent coking resistance under CH4 reforming conditions, owing to the optimized mixture of HEA constituents and the high OSC of the GDC support. HEA/GDC also showed superior operational stability for CH4 conversion over 30 h and post-test analysis of the catalyst did not indicate presence of carbon deposition, while both Ni/YSZ and Ni/GDC catalysts could be seen deactivating over time, despite the high initial CH4 conversion and H2 yield. The activity, stability and carbon-resistance of HEA/GDC as anode were also further investigated in a SOFC cell test. A current density of ∼100 mA/cm2 was achieved at 750 °C. The cell performed successfully over 30 h without any sign of decay. The overall polarization (ohmic and non-ohmic resistances) of the cell was low and stable. The moderate reforming rate of HEA/GDC is important for maintaining uniform temperature distribution and high coking tolerance especially for long-term high temperature SOFC operations, without compromising the electrochemical activity of the cell. Given the promising attributes of HEA/GDC as anode material in this study, successful effort has been made to further improve HEA/GDC for the direct utilization of other hydrocarbons such as methanol and ethanol, which are more easily stored and transported.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 acknowledge financial support from Advanced Research Projects Agency - Energy under contract DE-AR0001012. Technical discussions with Dr. Mike Tucker (Lawrence Berkeley National Laboratory), Dr. Greg Tao (Chemtronergy), Dr. Abdul Jabbar Hussain (Nissan Motors) and Dr. Rabi Bhattacharya (UES Inc.) is acknowledged.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.2022.09.018.

High-entropy alloy (HEA) anode and reforming catalyst, supported on gadolinium-doped ceria (GDC), have been synthesized and evaluated for the steam reforming of methane under SOFC operating conditions using a conventional fixed-bed catalytic reactor. As-synthesized HEA catalysts were subjected to various characterization techniques including N2 adsorption/desorption analysis, SEM, XRD, TPR, TPO and TPD. The catalytic performance was evaluated in a quartz tube reactor over a temperature range of 700–800 °C, pressure of 1 atm, gas hourly space velocity (GHSV) of 45,000 h−1 and steam-to-carbon (S/C) ratio of 2. The conversion and H2 yield were calculated and compared. HEA/GDC exhibited a lower conversion rate than those of Ni/YSZ and Ni/GDC at 700 °C, but showed superior stability without any sign of carbon deposition unlike Ni base catalyst. HEA/GDC was further evaluated as an anode in a SOFC test, which showed high electrochemical stability with a comparable current density obtained on Ni electrode. The SOFC reported low and stable electrode polarization. Post-test analysis of the cell showed the absence of carbon at and within the electrode. It is suggested that HEA/GDC exhibits inherent robustness, good carbon tolerance and stable catalytic activity,` which makes it a potential anode candidate for direct utilization of hydrocarbon fuels in SOFC applications.

Lignocellulosic biomass can be effectively used to reduce carbon footprint. As it is present in abundance, it can also be used for the sustainable production of fuels and chemicals [1]. The valorization of lignocellulose for the production of drop-in fuels and chemicals is a multi-step process that involves the steps of fractionation, depolymerization, and upgrading [2]. The fast pyrolysis can be used to produce depolymerized products from biomass. The method can be effectively used to produce biomass pyrolysis oil (or bio-oil (BO)), an essential feedstock that can potentially replace petroleum-based oil [3]. The quality of BO characterized by low heating value, high acidity, chemical instability, high viscosity, and high oxygen water content [4] can be improved by subjecting the sample to upgrading processes such as hydrotreatment and water-soluble phase separation [5,6]. It is difficult to directly use phenolic compounds and depolymerized lignin derivatives as fuels because labile radical intermediates are formed during the hemolytic cleavage [7]. These can be repolymerized to form oligomers during the recovery of liquid products [8,9]. Catalytic upgrading can be achieved following the hydrodeoxygenation (HDO) to address these issues and improve the extent of valorization of BO for the production of fuels and chemicals. Highly oxygenated compounds can be converted to highly stable products such as alcohols and alkanes during the process. The HDO has been investigated using noble metals (such as Pt, Pd, and Ru) and MoS-based catalysts as these exhibit high activities [4]. Transition metal catalysts can also be used for achieving the economically viable process [10].Among the monomeric compounds prepared by the HDO of BO, the lignin-derived cyclohexanol or alkyl cyclohexanol compounds are widely used for the development of polymers, fuels, and pharmaceuticals. Cyclohexanol is used for synthesizing cyclohexanone [11] which can be used as a precursor during the preparation of adipic acid, a monomer present in various polyamides, including nylon 6,6 [12]. Cyclohexanone is also used for the synthesis of caprolactam, a monomer of polycaprolactam such as nylon 6 [13]. Cyclohexanol is also used to produce cyclohexanolamine [14] and cyclohexyl acetate [15] that can be used in the pharmaceutical industry for the preparation of artificial sweeteners, food additives, and fragrances. Ketone–alcohol oil (KA oil) is composed of cyclohexanol and cyclohexanone and is commercially produced in petrochemical refineries from cyclohexanone following the oxidation using Co and Mn-based catalysts. The product yields are low (conversion: ≤ 11%; selectivity: ≤ 85%) [16]. The conditions that can be used for the effective production of cyclohexanol from biomass has been presented (Table S1). High yields and good selectivity were observed for the products synthesized from phenolic monomers [17,18], dimers [19,20], and lignin oil [21].Supercritical methanol depolymerization and hydrodeoxygenation processes (catalyst: CuMgAlOx; pressure: 20 MPa; temperature: 300 °C) have been conducted for the production of (alkyl)cyclohexanol from lignocellulose or lignin derivatives in batch reactors [22,23] and semi-continuous flowthrough reactors. The average selectivity achieved during the production of C6-C10 cyclic alcohols from maple wood (using a semi-continuous flowthrough reactor in the presence of CuMgAlOx catalyst) [24] was 31%. Insoluble pyrolytic lignin, coke, and gases (such as H2, CO, and CO2) were also produced under severe reaction conditions that hindered the production of cyclic alcohols. The seminal work on producing cyclic alcohols from pinewood-derived pyrolysis oil was performed using Raney Ni as the catalyst at a low reaction temperature of 120 °C. The hydrogen transfer reaction was conducted in the presence of isopropanol. The one-step reaction for the production of cyclic alcohols was highly selective [25]. The results indicated a remarkable hydrogenation activity of the catalysts that could be used to saturate the aromatic rings present in lignin components. We have previously reported two-step upgrading reactions that can be followed for the conversion of sawdust pyrolysis oil [26] and furan condensates to alkane-rich upgraded oil [10] using supported noble and transition metal catalysts in a continuous flow reactor.Inspired by the reported results, we investigated if Ni-based catalysts could be used for the selective hydrodeoxygenation of lignin monomers to synthesize cyclic alcohols following the two-step upgrading reactions involving BO. While Ni catalysts can be used to conduct hydrogenation and hydrodeoxygenation reactions [10], the low affinity of Ni nanoparticles (NPs) toward phenolic monomers can hinder the selective production of deoxygenated chemicals such as phenols and cyclohexanols. It has been reported that TiO2-supported Ni catalysts can promote deoxygenation, exploiting the strong metal–support interaction (SMSI) existing between the coordinatively unsaturated Ti3+ and Ni speices. This can improve the oxophilicity of Ni NPs reacting with biomass-derived oxygenates [27]. SMSI, however, can poison the active Ni sites. This can be attributed to the formation of TiO2 overlayer on the Ni NPs. The formation of the overlayer hinders the H2 dissociation. The entry of biomass-derived oxygenates into the Ni sites is hindered under these conditions [28]. As reported in the literature, SMSI can be tuned to modulate the intrinsic catalytic properties of Ni NPs to achieve the desired HDO activity.We studied the structures of active sites on the catalysts composed of Ni NPs dispersed on TiO2. The strong electrostatic adsorption (SEA) of metal on the TiO2 support and the hydrothermal (HT) treatment process led to the formation of active sites. The conventional impregnation (IM) method was also used to understand the effects of SEA-HT method on the active site forming ability. Bimetallic NiFe supported on TiO2 was fabricated and the Ni and Fe loading amounts were tuned to optimize the catalyst structures. The geometric and electronic structures of catalysts were investigated, and the formation of oxygen vacancies at the metal–metal oxide interface was studied. The hydrodeoxygenation of alkyl methoxypehnols following the processes of demethoxylation and hydrogenation was performed. The investigation of structure–activity indicated the presence of three active sites. The Fe-added Ni alloy NPs formed the metallic sites for the dissociative adsorption of H2 and hindered the hydrogenation of phenyl rings on the Ni active sites. The coordinatively unsaturated Fe species present on the FeOx shell layer adjacent to the NiFe sites and oxygen vacancies at the FeOx-TiO2 interface provided the oxophilic sites for the demethoxylation of guaiacylic monomers. The Ti3+ species present on the reducible TiO2 support promoted the dispersion of NiFe-FeOx core–shell NPs by exploiting the FeOx-TiO2-x/TiO2 anchor sites. The optimum catalysts were identified and they were further used for the production of cyclic alcohols from a mixture of alkyl methoxyphenols prepared from BO (Fig. 1 (B)). The high conversion and the high selectivity to cyclic alcohols were achieved during the conversion of methoxypehnols. Based on these results, we propose the strategy of producing cyclic alcohols from BO that can be used for the preparation of sustainable chemicals, such as nylon, plasticizers, and fuels [29]. Herein, the methods to synthesize improved catalysts (that can be used for HDO) consisting of non-precious metals and containing highly selective metal–metal oxide interfacial sites have been presented. The rational design for the preparation of supported non-precious metal catalysts that can be used for upgrading biomass-derived oxygenates has also been presented.Nickel(II) nitrate hexahydrate (Ni(NO3)2·6H2O, 99.999%), iron(III) nitrate nonahydrate (Fe(NO3)3·9H2O, ≥ 99.95%), titanium(IV) oxide (TiO2, P25, ≥ 99.95%), niobium(V) pentoxide (Nb2O5, 99.9%), cyclohexane (C6H12, ≥ 99%), cyclohexanol (C6H11OH, 99%), cyclohexanone (C6H10O, 99%), 1,2-dimethoxybenzene (C6H4(OCH3)2, 99%), 2-methoxy-4-propylphenol (C10H14O2, 99%), and 2-methoxy-4-(2-propenyl)phenol (C10H12O2, 99%) were purchased from Sigma-Aldrich (St. Louis, Missouri, USA). Ammonium bicarbonate (NH4HCO3, 97%), 2-methoxyphenol (C7H8O2, 99%), n-decane (C10H22, 99%), methanol (CH3OH, 99.9%), and silicon dioxide (SiO2, extra pure) were purchased from DaeJung Chemicals (South Korea). Zirconium (IV) oxide (ZrO2, ≥ 99%), gamma-phase aluminum oxide (γ-Al2O3, 99.98%), n-dodecane (C12H26, 99%), 4-methyl-2-methoxyphenol (C8H10O2, 98%), and 4-ethyl-2-methoxyphenol (C9H12O2, 98%) were purchased from Alfa Aesar (Ward Hill, Massachusetts, USA). Deionized (DI) water was obtained using the EXL® 7S Analysis Water system (Vivagen Co., Ltd., Seongnam, Korea) equipped with a filter (diameter: 0.22 µm). BO was purchased from BTG Bioliquids BV (Enschede, The Netherlands), which was prepared by the fast pyrolysis lignocellulosic biomass. N2 (99.999%), He (99.9999%), Ar (99.999%), 5% (v/v) H2/Ar, 1% (v/v) O2/He, and 5% (v/v) NH3/He, were purchased from the Sinyang Medicine (Anseong, Korea).A series of TiO2-supported Ni catalysts were prepared following the conventional wet IM (Ni/TiO2-IM) and HT synthesis methods (Ni/TiO2-HT, Fe/TiO2-HT, and NiFe/TiO2-HT). The wet IM method was used to prepare the catalyst containing 15 wt% Ni on the TiO2 support, denoted as Ni/TiO2-IM. TiO2 (P25, 1.70 g) was added to Ni(NO3)2·6H2O (1.49 g) dissolved in DI water (50 mL). Following this, the mixture was stirred for 1 h at room temperature. The prepared suspension was transferred to a round bottom flask (250 mL) and the solvent was evaporated at 50 °C using a rotary evaporator. The solid powder was further dried in air at 105 °C over 16 h, following which it was calcined for 2 h under a flow of N2 at 300 °C. It was reduced under a flow of 5% (v/v) H2/Ar at 450 °C for 4 h. Following this, a gray solid catalyst (Ni/TiO2-IM) was obtained. Ni/TiO2-HT was prepared (following the HT method) by depositing Ni (15 wt%) on the surface of TiO2 support. Fe on the TiO2 support was mixed with Ni (15 wt%) to prepare NiFe(x)/TiO2-HT following the HT method. Here, x denotes Fe loading (wt%). Ni(NO3)2·6H2O (1.49 g) and Fe(NO3)3·9H2O (0–1.09 g, depending on x) were dissolved in DI water (20 mL) and transferred to a 100-mL Teflon-lined chamber. TiO2 (P25, 1.40–1.70 g, depending on x) was added to the mixture and the mixture was ultrasonicated for 30 min to prepare a well-distributed suspension. An aqueous solution of NH4HCO3 (1 M, 30 mL) was added drop-wise to the suspension at room temperature under conditions of continuous stirring. The Teflon-lined chamber was sealed and heated at 150 °C for 15 h. The temperature of chamber was slowly brought down to room temperature, and the prepared solid was filtered under conditions of vacuum. The solid was washed with DI water until the pH reached 7. Following this, it was further washed three times using ethanol (50 mL each). The solid powder was further dried in air at 105 °C for 16 h and calcined under a flow of N2 at 300 °C for 2 h. The sample was reduced at 450 °C over a period of 4 h under a flow of 5% (v/v) H2/Ar. Following this, a black solid catalyst (Ni/TiO2-HT or NiFe(x)/TiO2-HT) was obtained. The freshly reduced catalyst was exposed to an atmosphere of 1% (v/v) O2/N2 for 30 min (at 25 °C) to passivate the metal surface to prevent the excessive oxidation of catalyst during storage. TiO2-supported 3 and 15 wt% Fe were prepared following the HT method. The process followed was similar to the process followed during the fabrication of NiFe/TiO2-HT. The samples were denoted as Fe(x)/TiO2-HT where x indicated Fe loading (wt%). Fe(NO3)3·9H2O (0.22 g and 1.09 g for 3 and 15 wt% Fe, respectively) was used as the precursor of TiO2 (P25, 1.94 g and 1.40 g for 3 and 15 wt% Fe, respectively). NiFe catalysts supported on other supports including SiO2, Al2O3, ZrO2, and Nb2O5 were prepared using the methods identical to the HT method used for NiFe(3)/TiO2-HT. Instead of TiO2, SiO2, Al2O3, ZrO2, and Nb2O5 were used to prepare NiFe(3)/SiO2-HT, NiFe(3)/Al2O3-HT, NiFe(3)/ZrO2-HT, and NiFe(3)/Nb2O5-HT, respectively.The Ni and Fe metal contents were determined using the inductively coupled plasma-optical emission spectroscopy (ICP-OES) technique (Model: 730-ES; Varian; Palo Alto, California, USA).The thermal stability of catalysts was determined using the thermogravimetry (TG) technique. The catalyst (10–20 mg) was dried over 10 min under a flow of N2 (30 mL/min) in the temperature range of 30–200 °C prior to conducting the experiments. The sample was heated from 30 °C to 900 °C (heating rate: 10 °C/min) under a flow of air (80 mL/min). Thermal degradation was determined using a thermogravimetric analyzer (Model: Q600; TA Instruments; New Castle, Delaware, USA).The N2 physisorption behavior was studied using the MicrotracBEL (Osaka, Japan) BELSORP-mini II system. The Brunauer–Emmett–Teller (BET) surface area (SBET) was measured, and the pore volume (Vp) and pore size distribution were calculated following the Barret–Joyner–Halenda (BJH) method.The crystal structures of catalysts were determined using the Dmax2500/PC X-ray diffractometer (Rigaku, Japan) equipped with a scintillation counter and graphite monochromatic detector. The Cu Kα ave radiation (λ = 0.15418 nm) was generated at 40 kV and 200 mA. The X-ray diffraction (XRD) results were recorded in the 2θ range of 3–90° (scanning rate: 2° per minute, step width: 0.02°). For the TiO2 supports, the anatase-to-rutile ratios were calculated by analyzing the intensities of peaks at 2θ = 25.3° (anatase) and 27.4° (rutile). The Ni-based particle size (d Ni(111)) was determined using the Scherrer equation (equation 1): (1) d N i 111 = K λ B c o s θ where d Ni(111) denotes the size of crystal domain determined by analyzing the Ni(111) peak, K is the Scherrer constant (0.94 for spherical crystals exhibiting a cubic symmetry), λ denotes the wavelength (0.15418 nm for Cu Kα ave), B is the modified full width at half maximum (FWHM, rad) for the diffraction peak (calculated as ((FWHM)2 – (FWHM of bulk crystal, 0.2° in this study)2)1/2), and θ denotes the Bragg angle corresponding to Ni (111).The electronic structures of catalysts were investigated using the high-performance X-ray photoelectron spectroscopy (XPS) technique using the VG Scientific ESCALAB 250 spectrometer (ThermoFisher Scientific Inc., Waltham, Massachusetts, USA).The atomic structure was imaged and the elemental distribution was determined using the HR-(S)TEM technique. The Talos F200X microscope (Thermo Fisher Scientific, Waltham, Massachusetts, USA) was used to record the images. A small amount of catalyst powder was dispersed in anhydrous ethanol. A significantly high dilution ratio was maintained and the dispersion was ultrasonicated for 1 h. Following this, the dispersion was mounted on a Holey carbon-coated copper grid (Electron Microscopy Sciences, Hartfield, Pennsylvania, USA; 200 square mesh; average opening size: 50 μm). The as-prepared TEM grid was further dried in air at 105 °C for 16 h. The sample was plasma etched to completely remove the residual organic layer.Temperature programmed methods and pulse chemisorption were used to characterize the catalysts using the BELCAT–B system (MicrotracBEL, Osaka, Japan) equipped with a thermal conductivity detector (TCD) consisting of tungsten–rhenium (W–Re) filament. A quadrupole mass spectrometer (Qmass) equipped with a yttriated iridium (Y–Ir) filament detector (BELMass) was also used for analyzing the samples.The H2 temperature programmed desorption (H2 TPD) results using solid samples were recorded. The catalyst (∼50 mg) was heated at 300 °C for 1 h under a flow of He (flow rate: 30 mL/min). Following this, the sample was reduced at 450 °C for 2 h under a flow of H2/Ar (5 vol%; flow rate: 30 mL/min), prior to bringing the temperature down to 50 °C under a flow of He (30 mL/min). The catalyst was heated from 50 °C to 900 °C (ramping rate: 10 °C/min) under a flow of He (30 mL/min). The quantity of desorbed H2 was measured using a TCD and identified using a mass spectrometer (at m/z = 2).Catalyst reducibility was determined by analyzing the H2 temperature programmed reduction (H2 TPR) results recorded for the catalysts. The as-prepared catalyst (∼50 mg) was heated at 300 °C for 1 h under a flow of Ar (30 mL/min). Following this, the temperature of catalyst was brought down to 50 °C. The catalyst was heated from 50 °C to 900 °C (ramping rate: 10 °C/min) under a flow of H2/Ar (5% (v/v); flow rate: 30 mL/min). The produced water was removed using a molecular sieve trap. The amount of H2 uptake per gram of the catalyst (mmolH2·gcat –1) was determined using the TCD method. This indicated the quantity of reduced sites.The temperature programmed oxidation (TPO) method was used to examine the surface/lattice oxygen vacancies present in the catalyst. The catalyst (∼50 mg) was heated from 50 °C to 900 °C (ramping rate: 10 °C/min) under a flow of O2/He (5% (v/v); flow rate: 30 mL/min). The quantity of desorbed O2 was measured following the TCD method. The produced water was removed using a molecular sieve trap.The results obtained using the H2 TPD and H2 TPR methods were used for determining the Ni metal dispersion (DNi, %) (equation (2)). DNi is defined as the ratio of the quantity of Ni surface atoms available for hydrogen adsorption to the quantity of Ni atoms present in the catalyst. The quantity of desorbed H2 (measured following the H2 TPD method) represents the quantity of active sites present to achieve the dissociative adsorption of H2 on the Ni NPs [30,31]. DNi can be calculated as follows: (2) D N i % = 2 × V H 2 T P D × M N i × S F m × W S × V m × d r × 100 where VH2 TPD denotes the volume of chemisorbed H2 (at STP) determined by analyzing the H2 TPD results, MNi denotes the atomic weight of Ni (58.69 g/mol), SF denotes the adsorption stoichiometric factor for the H to Ni molar ratio (H/Ni = 1 atom/atom), m represents the weight of catalyst (g), WS is the actual weight fraction of Ni in the catalyst determined using the ICP-OES technique, Vm is the molar volume of H2 (22414 mL/mol) at STP, and dr is the degree of reduction of Ni determined following the H2 TPR method. The theoretical dispersion for the Ni NPs (Dt, %) was calculated using equation (5). The size of Ni particle was determined using the TEM. A spherical model was assumed, and the value of Dt was calculated as follows: (3) D t % = 3 × M N i × S F 2 × N A × R N i × ρ N i × σ × 100 where NA represents the Avogadro’s number (6.02214 × 1023 mol−1), RNi is the number-averaged radius of Ni NPs measured using the TEM, ρNi is the density of bulk Ni metal (8.908 g/mL), and σ is the atomic cross-sectional area of H atoms adsorbed on the Ni surface (0.0649 nm2). The results obtained using the H2 TPD (equation (2)) and HR-TEM (equation (3)) methods were analyzed to determine the surface coverage (θ Ni) on the Ni NPs. It was determined using equation (4) as follows: (4) θ N i % = 1 - 2 × σ × N A × n × m N i W S × π × d N i 2 × 100 where n is the quantity of desorbed H2 per mass of catalyst (mmol/g), mNi is the weight of a single Ni NP (g), and dNi represents the diameter of Ni NPs measured using the XRD [32–34]. Equation (5) was used to calculate mNi. (5) m N i = ρ N i × π × d N i 3 6 The CO diffuse reflectance infrared Fourier transform spectroscopy (CO DRIFTS) was used to analyze the samples and the experiments were performed at the UNIST Next-Generation Catalysis Center (Ulsan, Korea). A Nicolet iS10 Fourier transform infrared (FT-IR) spectrometer (ThermoFisher Scientific Inc., UK) equipped with a zinc selenide (ZnSe) DRIFT cell and mercury-cadmium-telluride detector was used to conduct the experiments. The samples were diluted with KBr and then pretreated at 100 °C (heating rate: 10 °C∙min−1) for 1 h under an atmosphere of He (flow rate: 60 mL∙min−1). CO adsorption was examined under a flow of 1% (v/v) CO/He (flow rate: 60 mL∙min−1) at room temperature. The data collection was allowed to proceed for 15 min.The HDO activities of prepared catalysts were studied using a custom-built SUS 316 batch reactor (Hanyang Precision, Gimpo, Korea). A mixture of guaiacol (2.50 g, 20.14 mmol) and the catalyst (0.25 g) was mixed in n-decane (50 mL) and the mixture was loaded into the reactor. The inner reaction system was purged with N2 (gas) three times prior to conducting the experiments. The system was pressurized (5 MPa; H2) at room temperature. The reaction was performed at the desired reaction conditions, and the reaction mixture was cooled down to 50 °C. A mixture consisting of the products and the catalyst was collected using a conical tube (50 mL). An aliquot (1 mL) of the fresh liquid product was filtered through a PTFE membrane syringe filter (pore size: 0.45 µm). Following this, the liquid was transferred to a volumetric flask (20 mL). An internal standard (n-dodecane; 1 μL) was added to the product mixture, and the mixture was further diluted using methanol for analysis using the gas chromatography (GC). The spent catalyst was recovered from the residual mixture of products following the vacuum filtration. The recovered catalyst was washed three times with acetone (25 mL each) to remove all traces of n-decane and organic compounds. The obtained solid was dried at 105 °C over 16 h, and the dried catalyst was denoted as the spent–recovered catalyst (SC). The liquid products were characterized using the gas chromatography-mass spectrometry (GC–MS). The GC–MS system (Agilent 78900A; 5975C inert MS XLD with triple axis-detector) was equipped with an autosampler injector (Agilent 7860 N) and an HP–5MS capillary column (60 m × 0.25 µm × 0.25 mm ID). The liquid product was quantitatively analyzed using a gas chromatography–flame ionization detector (GC–FID, Hewlett 5890 Packard Series II, USA) equipped with an autosampler injector (6890 series injector) and HP–5MS capillary column (60 m × 0.25 µm × 0.25 mm ID). The catalytic activity was determined by analyzing the conversion (Xf eed,%), product yield (Yp roduct,%), and selectivity (Sp roduct,%). The parameters were determined as follows: (6) X f e e d % = 1 - n feed n feed 0 × 100 (7) Y p r o d u c t % = n p r o d u c t n f e e d 0 × 100 % (8) S p r o d u c t % = n p r o d u c t ∑ n all p r o d u c ts × 100 where n f e e d 0 is the number of mol of the feed compound before reaction and nfeed, n p r o d u c t , and ∑ n all p r o d u c ts represent the number of mol of the residual feed compound, targeted product, and the total number of mol of all products after reaction, respectively.The cyclohexanol (CHNOL) production rate (rCHNOL, molCHNOL∙gNi −1∙h−1) was calculated as the ratio of the total number of mol of cyclohexanol produced per reaction time to the weight of Ni metal. (9) r C H N O L m o l C H N O L ∙ g Ni - 1 ∙ h - 1 = n C H N O L / r e a c t i o n t i m e m × W S The activity of prepared catalyst was also described in terms of the site time yield of cyclohexanol (STYCHNOL). The mol of cyclohexanol produced per unit time was divided by the mol of active Ni species (determined by the H2 TPD and TPR methods) to determine STYCHNOL. (10) S T Y C H N O L mol CHNOL · mol Ni - 1 · h - 1 = n C H N O L / r e a c t i o n t i m e n N i = n C H N O L / r e a c t i o n t i m e m × W S × D N i / M N i A two-step reaction using BO was conducted. The first step involved the catalytic depolymerization of pyrolytic lignin fraction in the presence of 5 wt% Pd/C to obtain phenolic monomers and the second step involved the selective HDO of phenolic monomers using NiFe(3)/TiO2-HT. The second step led to the production of cyclic alcohols. The first step of reaction was performed in a batch reactor. A mixture of BO (6 g, BTG) and 5 wt% Pd/C (0.5 g) was used for the experiments. The mixture was dissolved in n-decane (50 mL) at 200 °C and the reaction was allowed to proceed for 1 h under 5 MPa (pressure generated by H2; measured at room temperature). Note that all H2 pressures depicted in this manuscript were measured at room temperature, not at the high reaction temperature. The extents of product recovery and separation achieved are depicted in Fig. S1. Four types of products (light oil (LO), water-soluble oil (WSO), heavy oil (HO), and solid residue (SR)) were recovered at the end of the first step. The fresh product containing a mixture of the liquid and solid products was collected from the reactor. Following this, the constituents were separated following the vacuum filtration. PTFE membrane filters (pore size: 0.45 μm) were used for filtration to obtain the filtrate and residue. The filtrate consisted of organic (decane-soluble) and aqueous (water-soluble) phases. The liquid–liquid separation technique was used to separate the phases using a separatory funnel. The n-decane (organic) layer containing LO and the aqueous layer containing WSO formed two separate layers. These fractions were labeled BO-S1-LO and BO-S1-WSO, respectively. HO was present in the residue present in the upper part of the vacuum filtration setup. The layer could be dissolved in acetone, yielding the HO–acetone soluble phase. The insoluble SR was present on the filter paper containing the catalyst and coke. The liquid filtrate appeared dark brown when washed with acetone. This indicated the presence of HO containing high-molecular-weight compounds. Acetone was removed at 40 °C using a rotary evaporator to recover oil. The recovered phase was labeled as BO-S1-HO. The solid residue, denoted as BO-S1-SR, was then dried at 105 °C over 16 h. In the second step, selective HDO was conducted to produce cyclic alcohols. In this step, BO-S1-LO (LO obtained from the first step during the conversion of BO was used in this step) was used as the reactant in the presence of the NiFe/TiO2 catalyst (0.5 g). The conditions used were similar to the conditions under which the model reaction was conducted (temperature: 270 °C; time: 1 h; pressure: 5 MPa (H2)). The method followed for product recovery and separation was similar to the method followed in the first step. The fresh LP and SR obtained were denoted as BO–S2–LP and BO–S2–SR, respectively.Ni/TiO2-based catalysts were selected for the HDO of biomass-derived compounds because they could selectively cleave C-O bonds present at the perimeter of Ni-TiOx interfacial sites [27,28]. The oxygen defects at the interface between Ni and Ti3+ have been reported to be the active centers that participate in HDO [27]. Chemical treatment (such as acid-base and redox reactions) methods can be used to generate the defects on the Ni metal, TiO2 support, or Ni-TiOx interface [35]. HDO usually necessitates the acid–redox reactions [36–38]. Two synthetic strategies for the preparation of TiO2-supported Ni catalysts (namely IM and HT synthesis methods; Fig. 2 (A)) were investigated to understand their effects on the strength of metal–support interactions between Ni and TiO2. The strong electrostatic adsorption and HT treatment (SEA-HT) methods were combined to generate a highly dense interface between the metal and the support. The interface obtained under the conditions of the HT method was larger than that of the IM method. The adsorption of metal monolayer [39–41] and the ion-exchange method involving the metal cation complex precursors (on the negatively charged metal oxide surface containing defects) led to the formation of larger interace in the HT method [42]. Under these conditions, the surface chemical properties of supported metal catalysts could be tuned.To assess the accuracy of the presented hypothesis, the Ni crystal structures, their particle size, and H2 dissociation behavior were analyzed using XRD, HR-TEM, and H2 TPD. The XRD results revealed that the Ni crystal size of Ni/TiO2-HT was smaller (10.7 nm, calculated using the Scherrer equation) than that of Ni/TiO2-IM (117.2 nm) (entries 3–4; dNi, Table 1 ). These observations were further confirmed by the particle size distributions obtained using the HR-TEM (Fig. 3 (A1) and (B1)). The average particle diameter of Ni NPs (14.7 nm) present in Ni/TiO2-HT was approximately 10 times smaller than that of Ni NPs present in Ni/TiO2-IM (140.1 nm). The small Ni particles in Ni/TiO2-HT were correlated with the high extent of H2 desorption determined by the H2 TPD results (0.013 mmolH2·gcat −1) and the high apparent Ni dispersion (DNi = 24.4% determined by the H2 TPD). Compared to Ni/TiO2-HT, the Ni dispersion for Ni/TiO2-IM was negligible (entries 3–4; DNi, Table 1).Under conditions of SEA treatment (Fig. 2(B)), the Ni precursors in the form of Ni(II) hexahydrate complex cations (Ni(H2O)6 2+) were deposited on the surface of TiO2 support in the presence of NH4HCO3, leading to the formation of a negatively charged TiO2 surface (pH 9.0 > PZC 6.2) [39]. It has been previously reported that metal cation hexaaqua complexes can be strongly adsorbed on the surface of TiO2, forming mono- and di-substituted complexes containing a bridging surface oxygen (inner-sphere complex). The (hydr)oxobridge connection formed the metal–support interfacial region [40]. The Ti-O bonds present in the TiO2 support can be exploited to form the [Ti(OH)6]2− species under conditions of the HT method and high pH [43]. The complex can be further converted to the labile TiO6 2− species on the TiO2 surface to generate the defects on the uppermost layer of TiO2. The NPs can be grown from the dispersed nuclei which were formed from Ni(H2O)6 2+ (stabilized by the (hydr)oxobridges on the negatively charged TiO2 surface) [44]. The defects on the surface of TiO2 can be generated during the ion exchange between Ni(H2O)6 2+ and surface Ti. Notably, the IM method at pH ∼5 < PZC 6.2 (at room temperature and atmospheric pressure) formed the weak binding of metal precursor on the surface of TiO2 containing a smaller number of defects. The process forms metal particles that can easily aggregate during the reduction [45].The SEA-HT method was further used to synthesize the NiFe/TiO2 catalysts (exhibiting the ternary heterostructure). Under these conditions, the metal–support interfacial interactions with the defect-containing support were generated [46,47]. The addition of Fe(III) precursor led to the formation of [Fe(H2O)6]3+ and Ni(H2O)6 2+. Strong bonds were formed between the metal cation complex precursors and TiO2. It has been reported that aqua complexes containing Fe can be effectively deposited on the surface of TiO2 by forming tetradentate inner-sphere complexes [48]. In addition, the Ni2+ and Fe3+ cations can be deposited as NiFe-based hydroxide complexes. The mobile anions such as OH–, HCO3 –, and CO3 2– can be used to balance the charge under basic conditions. The deposition of bimetallic NiFe on the surface of charged and defect-containing surface of TiO2 follows a similar monolayer adsorption mechanism. The detailed mechanism of adsorption is explained in section 3.4.The pore structures and surface areas of monometallic Ni/TiO2 and bimetallic NiFe/TiO2 catalysts were examined by the N2 physisorption measurement. The adsorption–desorption isotherms observed for the catalysts exhibited a weak hysteresis at p/p0 = 0.9–1.0 (Fig. S2(A)), indicating the formation of inter-particular spaces between the TiO2 NPs. The formation of inter-particular spaces is reflected by the BJH pore size distributions (Fig. S2(B)) [49,50]. The measured BET surface areas revealed that HT treatment and the addition of metal components tuned the pore structures of TiO2. When TiO2 was subjected to the conditions of HT treatment in the absence of metals, a decrease in the BET surface area (56 to 44 m2⋅g−1) (SBET, Table 1) and an increase in the single point pore volume (0.19 to 0.44 cm3/g) were observed (Table S2). These results revealed that TiO2 (P25) was dissolved, and agglomerated TiO2 particles were formed (generating the inter-particular spaces or observed pores) under these conditions. The increase in the BET surface area (49–50 m2⋅g−1) and pore volume (0.48–0.56 cm3⋅g−1) in Ni/TiO2-HT (or NiFe/TiO2-HT) indicated that the presence of Ni and Fe precursors during the HT process influenced the nanoscopic pore structures of catalysts. The BET surface area of Ni/TiO2-IM was lower and the pore volume of Ni/TiO2-IM was higher than those measured for the HT-prepared catalysts.The crystal structures of catalysts, analyzed using the XRD, varied based on the synthetic methods followed and the Fe contents (Fig. 4 (A)). Anatase and rutile phases of TiO2 were present in all the catalysts. The major peak corresponding to the (101) anatase TiO2 (PDF#21–1272) appeared at 2 θ = 25.3°, and the major peak corresponding to the (110) rutile TiO2 (PDF#21–1276) appeared at 2 θ = 27.4°. The presence of fcc Ni (PDF#04–0850) was observed in all the Ni-containing catalysts. The formation of fcc Ni (PDF#04–0850) was confirmed by the presence of distinct diffraction peaks at 2 θ = 44.5° (111) and 51.8° (200) [51]. The results obtained using the Scherrer equation indicated that the Ni particles observed in the catalysts prepared following the HT method were smaller than those observed in the catalysts prepared following the IM method. When the monometallic Ni/TiO2-HT was analyzed, a broad Ni(111) peak was observed at 2θ = 44.43°. The particle size was calculated to be 10.7 nm (entry 4; dNi, Table 1). When Ni/TiO2-IM was analyzed, sharp diffraction peaks corresponding to Ni(111) and Ni(200) were observed, indicating the presence of bulky Ni metal (entry 3; dNi, Table 1). The crystal structures of TiO2 were adjusted by the metal deposition methods (SEA-HT or IM) followed during the synthesis. The weight fraction ratio (for the anatase and rutile phases) measured for Ni/TiO2-HT (5.33 w/w) was higher than the anatase-to-rutile weight fraction ratio measured for Ni/TiO2-IM (2.65 w/w) (entries 3–4; WA/WR, Table S3). The anatase structure was preferentially formed at temperatures below approximately 600 °C. The acceleration in the rate of formation of anatase can be attributed to the significant reconstruction of TiO2 (in the presence of Ni precursor) observed during the HT process. Under these conditions, dissolution, ion exchange, and recrystallization of TiO2 could occur.The formation of Ni was confirmed for the bimetallic NiFe/TiO2-HT catalysts. The permeation of Fe into the bulk Ni structures (leading to the possible formation of alloys) was also validated by a decrease in the 2θ value and an increase in the d-spacing when the Fe content was increased. The addition of Fe to form NiFe/TiO2-HT led to a shift in the peak (from 2θ = 44.43° (0 wt% Fe) to 44.29° (5 wt% Fe)) corresponding to Ni(111). The peak corresponding to Ni(200) shifted from 2θ = 51.82° (0 wt% Fe) to 51.62° (5 wt% Fe) with an increase in the Fe content from 0 wt% to 5 wt% (entries 4–7; 2θ, Table 1). The peaks corresponding to NiFe(111) and NiFe(110) appear at 2θ = 43.7° and 50.9°, respectively. The positions of diffraction peaks characterizing NiFe/TiO2-HT indicate the formation of a transient structure (between Ni and NiFe crystals). The atomic distribution of Fe in the Ni particles was also confirmed. The diffraction peaks corresponding to the Fe oxides in NiFe/TiO2-HT were of low intensity, indicating that the crystalline Fe oxides were present in low concentrations. It also indicated the low extent of segregation of the Fe species including Fe oxides. In Fe(15)/TiO2-HT, the formation of ilmenite FeTiO3 (PDF#29–0733) (alloyed Fe oxides with TiO2) was observed, and the formation could be attributed to high Fe loading. The formation of ilmenite FeTiO3 (PDF#29–0733) was not observed under conditions of low Fe loading (3 wt% Fe, Fig. S3(C)). The absence of FeTiO3 phase and inflated Ni structures in the NiFe/TiO2-HT catalysts indicate that the Fe species preferred to form alloys with Ni (over NiFe or Fe-Ti oxides) when subjected to the conditions of HT method [52].When the formation of supported NiFe was attempted using other oxide supports such as SiO2, γ-Al2O3, and Nb2O5, the formation of Ni metals or NiFe alloys was not observed. The diffraction peaks corresponding to Ni or NiFe were not observed under these conditions (Fig. 4(B)). The segregation of Ni and Fe species on the Al2O3 support was observed using energy dispersive X-ray spectroscopy (EDS) mapping and HR-(S)TEM (Fig. S4(A)). The XPS results confirmed the absence of metallic Ni on the Al2O3 support (Ni 2p in Fig. S5(B)). These observations indicate that the formation of NiFe alloy was not favored when Al2O3 was used as the support. The XRD results observed for NiFe/ZrO2 prepared following the HT method indicated the formation of monoclinic ZrO2 and well-developed large Ni metal crystals. These results were also validated by the TEM images (Fig. S4(B)) and the XPS results (Ni 2p in Fig. S5(B)). These observations indicate that the Fe-doped Ni particles can be formed on the surface of reducible metal oxides such as TiO2 and ZrO2 rather than that of non-reducible metal oxides.The TEM images of NiFe/TiO2-HT confirmed the formation of Fe-doped Ni cores and FeOx shells on the TiO2 support (Fig. 3). Overlayers formed of TiO2 were observed on the Ni NPs, indicating the presence of strong interactions between TiO2 and Ni. The formation of the overlayers and the generation of interactions were validated by the results obtained using the HR-TEM coupled with HAADF-STEM and EDS (Fig. 3). The structural morphology and elemental distribution were also studied. Large Ni NPs with an average Ni particle size of 140.2 nm were observed (in Ni/TiO2-IM). The result agreed well with the corresponding XRD results (Fig. 4(A)). The presence of TiO2 on the surface of Ni NPs was also observed (Fig. 3(A4)). A uniform distribution of small Ni NPs (average particle size: 14.7 nm) was observed when Ni/TiO2-HT was studied (Fig. 3(B)). The HR-TEM images of Ni/TiO2-HT revealed the presence of overlayer lattices. The d TEM-spacings (lattice spacings measured using the TEM images) were 2.03 and 3.51 Å for the Ni NPs and anatase TiO2(101), respectively (Fig. 3(B2)). The formation of amorphous TiO2 on the surface of Ni NPs (Fig. 3(B4)) can be attributed to the formation of Ni(OH)2-Ti(OH)4 precipitates (following the HT method) on the surface of TiO2. This led to the formation of TiOx clusters on the surface of Ni. The unique structural morphology of Fe species and core–shell structures consisting of NiFe cores and NiFe-FeOx shells were observed when NiFe/TiO2-HT was studied. The formation of core–shell structures was proved using the EDS mapping (Fig. 3(C5, C6(iv), and (v)). The core–shell structure was not observed when other supports (Nb2O5, SiO2, Al2O3, and ZrO2) were used, which can be attributed to the poor formation of Ni species (for NiFe/SiO2-HT and NiFe/Al2O3-HT as observed for the poor XRD peaks) and the permeation of Fe into Ni particles (for NiFe/Nb2O5-HT and NiFe/ZrO2-HT as observed for the shift of Ni peaks of XRD results). The average Ni particle size measured for NiFe/TiO2-HT was 19.3 nm (Fig. 3(C1)), which was larger than the average Ni particle size measured for monometallic Ni/TiO2-HT. This could be attributed to the Fe-containing shells present on the surface of Ni NPs. The formation of core–shell structure with Ni cores (m/z = 59 (Ni)) and NiFe shells (m/z = 56 (Fe), 48 (Ti)) was also confirmed from the HR-STEM images (Fig. 3(C6(i and ii)). The interplanar spacings were analyzed in detail (Fig. 3(C6(iii))). The lattice spacings on the NiFe nanoparticle surface were calculated using the inverse fast Fourier transform technique. The results revealed that the d TEM-spacings of 0.179 nm and 0.208 nm were measured for the core area (area a) of the tetrataenite NiFe phase (PDF#47–1417) or the Fe-doped fcc Ni particles. A large d TEM-spacing of 0.239 nm was measured for the shell (area b, 1.825 nm thick) of Fe2O3 phase (PDF#39–1088) and non-stoichiometric FeOx. The EDS mapping technique was used to analyze NiFe/TiO2-HT, and the results indicated the presence of highly dispersed Fe species on TiO2 (Fig. 3(C3)). The lattice distance (3.53 Å) measured for the TiO2 phase using the HR-TEM was slightly larger (Fig. 3(C2)) than the lattice distance measured for the monometallic Ni/TiO2-HT. This indicates that the incorporation of Fe species into TiO2 was promoted by the strong interaction present between Fe and TiO2. Ni-free Fe(3)/TiO2-HT was analyzed using the XRD (Fig. S3(C)), and the results indicated the formation of amorphous Fe2O3 at the perimeter of TiO2 support, indicating the presence of FeOx–TiO2 interface at the surface of NiFe(3)/TiO2-HT.Interestingly, the presence of amorphous TiO2 clusters was not observed on the NiFe NP surface. This indicated that Fe-containing shells that wrapped the Ni NPs suppressed the generation of strong Ni-TiO2 interactions [53]. The formation of these complex structures is attributed to the formation of mixed NiFe-OH precipitate. The Ti4+ species present on the TiO2 surface favored ion-exchange reactions with Fe3+ (over Ni2+) during the HT process [42]. Improved apparent dispersion and an increase in the fraction of surface-exposed Ni atoms (in NiFe NPs) at the FeOx-TiO2 anchoring sites were observed. In addition, it was observed that the surface coverage of Ni (θ Ni, Table 1) increased from 82.6% to 89.8% when the Fe loading was increased from 1 wt% to 5 wt%. This indicated that the Ni and Fe species strongly interacted with each other. The results confirmed the formation of Fe-doped fcc Ni NPs surrounded by the FeOx shells. The addition of Fe to the Ni cores led to a slight shift in the XRD peaks corresponding to fcc Ni. The peaks indicated the formation of NiFe alloys and transient crystals of Fe-doped Ni NPs. The formation of these complex core–shell structures following a simple deposition method under conditions of mild heat treatment (< 450 °C) can be achieved in the absence of structure-directing agents, organic surface modifiers [54–57] and carbonaceous precursors [58–60]. Although these have been discussed in the literature reports, papers reporting the results obtained following the process reported herein are rare. The literature reports do not discuss the results in detail, and hence the results reported herein could not be compared for effective validation.The chemical compositions and electronic states of the surface atoms present on the catalysts were observed using the XPS. Peaks corresponding to Ni 2p, Fe 2p, Ti 2p, and O 1s were observed (Fig. S5(A)). The Ni 2p3/2 peaks obtained for all the catalysts were deconvoluted into peaks corresponding to Ni0 (852.4 eV), Ni2+(856.0 eV), and satellite peaks (861.0 eV) [61,62]. The binding energy of Ni0 (852.5 eV) in Ni/TiO2-HT was comparable to that of Ni0 in Ni/TiO2-IM and the binding energy corresponding to Ni2+ in Ni/TiO2-HT (855.7 eV) was lower than the binding energy of Ni2+ in Ni/TiO2-IM (856.0 eV). This indicated that the Ni2+ ions present on the surface of Ni NPs (Ni/TiO2-HT) were partially reduced by the adjacent TiO2 species while the Ni0 ions present in the core of Ni NPs were not significantly adjusted by TiO2. The Ti 2p binding energy for Ni/TiO2-HT (458.5 eV) was lower than the Ti 2p binding energy for Ni/TiO2-IM (458.7 eV) (Fig. 5 (A) and Table 1). This indicated the formation of Ti3+ (and not Ti4+). These observations indicate that the presence of oxygen vacancies on the TiO2 surface improved the transfer of electrons from the surface Ti4+ to Ni2+. The oxygen vacancies on the TiO2 surface maintain charge balance. The charge imbalance can be attributed to the substitution of Ti4+ by Ni2+ at the octahedral sites of TiO2 structure [63]. The absence of octahedrally coordinated O2– anions can potentially lead to a reduction in the extent of charge transfer (usually from Ni2+ to O2–), leading to a decrease in the binding energy of Ni2+.O 1s peaks were adjusted by the electronic states of Ni and Ti. The O 1s spectra can be deconvoluted into three parts corresponding to three different oxygen species (lattice oxygen (OL, 529.1 eV), oxygen vacancy (OV, 530.8 eV), and chemisorbed oxygen (OC, 532.2 eV)) [64]. The spectral profile revealed that the OV/OL ratio measured for Ni/TiO2-HT (0.24) was higher than that measured for Ni/TiO2-IM (0.13). This indicated that the number of oxygen vacancies in Ni/TiO2-HT was higher than the number of oxygen vacancies in Ni/TiO2-IM. These observations further suggest that the ion-exchange reaction between Ni2+ and Ti4+ in the TiO2 crystal during the HT process led to the formation of a large number of oxygen vacancies, which in turn led to the formation of strong bonds between Ni and Ti. The interfacial charge transfer (from the electron-rich Tiδ+ (δ < 4)species to the Ni species (Niδ– on the Ni NP surface)) led to the formation of Niδ––OV–Tiδ+. We also studied the effects of Fe species on the selective HDO involving NiFe/TiO2-HT. The Fe 2p results for Fe(3)/TiO2-HT (a control catalyst) revealed the presence of peaks corresponding to Fe2+, Fe3+, and Fe satellite at 709.7, 712.6, and 716.8 eV, respectively, and the peak corresponding to Fe0 was not observed. The slight shift in the binding energy of Fe2+ toward the lower binding energy region (∼0.6 eV) indicated that electron transfer occurred from Ti4+ to Fe2+. The charge transfer could be attributed to the strong interaction between Fe and TiO2, hinting toward the formation of Fe–OV–Ti (FeOx-TiO2) in Fe/TiO2-HT.The Fe 2p results of the bimetallic NiFe/TiO2-HT confirmed the electron transfer between the Ni, Fe, and Ti species. The Fe 2p3/2 results of the NiFe/TiO2-HT catalysts revealed the presence of peaks corresponding to Fe0 in the range of 705.3–705.9 eV and Fe2+ in the range of 713.8–715.3 eV. The peaks corresponding to Fe2+ and Fe3+ present in NiFe(1)/TiO2-HT and NiFe(3)/TiO2-HT shifted to the lower binding energy regions, indicating the partial reduction of Fe oxides (to form Feδ+, δ < 3). The reduction was promoted by the electron transfer from the adjacent Ni species. This was confirmed by the shift in the binding energy (by 0.6 eV) of Ni0 (Ni/TiO2-HT) to the region characterizing Niδ+ (NiFe/TiO2-HT) (Ni 2p, Fig. 5(A)). The interaction between the Ni and Fe species can also be confirmed by the formation of NiFe alloy (at 450 °C) in the presence of NiFe(OH)x precursor under an environment of H2 (during the preparation of NiFe/TiO2-HT) [65]. The formation of FeOx shell on the surface of Ni core suppressed the generation of direct Ni-TiO2 interaction. The formation of the shell was confirmed using the HR-(S)TEM (Fig. 3(C6(iii))). It was observed that surface oxygen vacancies were present on the NiFe/TiO2-HT catalysts. The maximum OV/OL ratio of 0.32 was measured for NiFe(3)/TiO2-HT. The corresponding OV/OL ratio measured for NiFe(1)/TiO2-HT was 0.28 and that for NiFe(5)/TiO2-HT was 0.20. The fractions of surface oxygen vacancies measured for NiFe(3)/TiO2-HT catalysts were higher than that measured for monometallic Ni/TiO2-HT (OV/OL = 0.27). This could be attributed to the formation of surface oxygen vacancies at the NiFe-TiO2 interface (Fe–OV–Ti) and FeOx (Fe–OV–Fe) sites (NiFe/TiO2-HT, Fig. 5(B)). These observations further indicate that the presence of partially reduced Feδ+ (δ < 3) in the bimetallic NiFe/TiO2-HT catalysts led to the formation of iron oxide (FeOx) containing coordinatively unsaturated Fe sites that can promote the selective HDO reaction. The results were confirmed using the XRD and HR-(S)TEM results.The reduction behaviors of Ni, Fe, and TiO2 were investigated using the H2 TPR method (Fig. 6 (A)) and by analyzing H2 uptake (Table 1) to understand the metal oxide–metal interaction in NiFe/TiO2 catalysts. The monometallic Ni/TiO2-IM catalyst was studied, and the presence of two primary reduction peaks was observed. One of the peaks appeared in the temperature range of 230–250 °C, while the other appeared at 412 °C. A significantly low H2 uptake of 0.003 mmolH2·gcat –1 was measured. The peaks corresponded to the reduction of NiO to Ni0 weakly and strongly deposited on TiO2, respectively [32]. A strong and broad reduction peak spanning the range of 210 °C to 310 °C (representing the reduction of Ni2+ to Ni0) was observed for Ni/TiO2-HT. The peak temperature was observed to be 285 °C and the H2 uptake was 1.21 mmolH2·gcat –1. The H2 TPR results for the monometallic Fe(3)/TiO2-HT presented two-step reduction peaks at 292 °C (for the conversion of Fe2O3 to Fe3O4) and 735 °C (for the conversion of Fe3O4 to Fe) (Fig. S6) [66]. The lower reduction temperature observed for the first step (292 °C vs. ∼ 450 °C reported in the literature) for Fe(3)/TiO2-HT indicates that H2 was consumed by the tiny Fe2O3 particles present on the surface of TiO2 support. This was reflected by the presence of peaks of low intensities in the XRD results (Figs. S3(C)). These observations also confirmed that the HT method could be followed to promote the formation of small Ni and Fe NPs that interacted strongly with the reducible TiO2 (anatase) phase [67]. The incorporation of Fe into the Ni/TiO2 structure decreased the reduction temperature (for the reduction of Ni2+ to Ni0) to 250 °C. The H2 uptake measured under these conditions (1.21 mmolH2∙gcat -1) was higher than that measured for the monometallic Ni/TiO2-HT. This indicates the presence of Ni2+ species that can be easily reduced because of the presence of a large number of oxygen vacancies on the bimetallic NiFe/TiO2-HT catalyst [68,69] allowing the effective reduction of NiO present in the neighborhood of oxygen vacancies [69].The presence of the second reduction peak (for Fe2O3) at 300 °C (or the presence of the peaks in the high-temperature region) and the low H2 uptake measured for NiFe/SiO2 and NiFe/Al2O3 indicate that the oxygen vacancies improve the efficient reduction of Ni2+. With an increase in the Fe content, the peak corresponding to Ni2+ reduction shifted to the higher temperature region. The shift can be attributed to the fact that Fe oxide present in close proximity to Ni lowers the reducibility of Ni. Interestingly, the reduction peaks appearing in the temperature range of 450–700 °C were not observed for NiFe/TiO2-HT. This suggested that the reducibility of Fe oxides increased under the effect of hydrogen spillover observed in the Ni NPs located at the core–shell interfaces [70]. These observations further confirmed that the NiFeOx core–shells dispersed on the reducible TiO2 moieties tuned the hydrogen adsorption activity of metal.The hydrogen adsorption behavior was further studied using the H2 TPD-MS measurement (at m/z = 2) to study the activity and determine the quantity of hydrogen atoms adsorbed on the catalysts (Fig. 6(B)). Ni/TiO2-IM exhibited negligible H2 desorption, and this could be attributed to the presence of large Ni particles. Large peaks corresponding to H2 desorption appeared at 150 °C for Ni/TiO2-HT, indicating the presence of a large number of Ni active sites that could participate in the hydrogenation. These observations indicate that the HT method can be followed to increase the activity of Ni NPs during HDO. The H2 desorption peak observed for NiFe/TiO2-HT shifted to the higher temperature region (temperature ≤ 250 °C), indicating the presence of strong bonds between hydrogen and Fe-incorporated Ni.The TPO method was also used to confirm the presence of oxygen vacancies on the catalysts (Fig. 6(C)). An intense O2 consumption peak was not present for TiO2 (P25). A peak of low intensity, corresponding to O2 consumption (at ≥ 700 °C), was observed for TiO2-HT. The presence of the peak indicated the generation of oxygen vacancies in hydrothermally treated TiO2 (P25). The peak corresponding to O2 consumption appeared at 260 °C for Ni/TiO2-HT during the Ni oxidation. The peak position was lower than the peak position observed for Ni/TiO2-IM (450 °C). These observations indicate that the HT method can be followed to effectively activate the surface oxygen atoms at low temperatures. Under these conditions, the nearby Ni atoms were oxidized. A slight shift of the peaks corresponding to O2 consumption (for NiFe/TiO2-HT) toward the high temperatures regions indicates that the formation of NiFe alloys proceeds under conditions of thermal activation.The CO DRIFTS was used to further investigate the geometry of catalyst surface. Adsorption bands for monometallic Ni/TiO2 and bimetallic NiFe/TiO2 appeared in the regions spanning 2070–2020 cm−1 (linearly adsorbed CO) and 2000–1950 cm−1 (bridged CO) (Fig. 6(D)). The bands corresponding to the CO molecules adsorbed on the low-coordination Ni atoms appeared at 2063 cm−1 [32], and the bands corresponding to the CO molecules adsorbed on the Ni tetracarbonyl species (Ni(CO)4) appeared at 2057 cm−1. The peak at 2050 cm−1 was attributed to the CO molecules linearly adsorbed on the Ni atoms (CO-Ni0), and the peak at 2035 cm−1 was attributed to the CO molecules adsorbed on the negatively charged Ni (CO-Niδ−) [32]. Notably, the intensity of the band corresponding to CO-Niδ− appearing for Ni/TiO2-IM was lower than the intensities of the bands appearing for Ni/TiO2-HT and NiFe/TiO2-HT. This indicated that a large number of CO molecules could be adsorbed by the Niδ− present in the catalysts prepared by the HT method. Because the electron transfer from TiO2 to Ni at the Ni-TiO2 interface leads to the formation of electron-rich Niδ− [71], the extent of back donation of electrons from Niδ− to the adsorbed CO can be improved, leading to the formation of strong Ni-CO bonds at the Ni–OV–Ti interfacial sites [32]. The process also leads to the generation of high-intensity bands corresponding to CO-Niδ−.The bands appearing at ∼1995 cm−1 and ∼1983 cm−1 were attributed to the adsorption of bridged CO molecules on the highly coordinated Ni surface. The small intensity of the peak appearing at 1995 cm−1 revealed that negligible amounts of bridged-O species were adsorbed on the Ni surface present in Ni/TiO2-IM. The peak at 1995 cm−1, for Ni/TiO2-HT and NiFe/TiO2-HT, red-shifted toward 1993 cm−1. The intensity of the new peak was higher than that of the peak at 1995 cm−1. This indicated that a large number of CO molecules were adsorbed on the Niδ− surface present at the Ni–OV–Ti interfacial sites.Based on the results, it was inferred that isolated and terrace Ni sites were present in Ni/TiO2-HT. The result for NiFe/TiO2-HT contained adsorption peaks corresponding to linear and bridged CO molecules. As the Fe content in NiFe/TiO2-HT was increased, the intensity of the adsorption peaks at 1993 and 1983 cm−1, corresponding to bridged CO, decreased. The peak positions blue-shifted toward 1995 cm−1. These observations indicated that Fe encapsulation led to a reduction in the number of bridged CO adsorption sites. The process also improved the extent of electron transfer (from Ni0 to Fe). These results confirmed the presence of well-distributed FeOx species on the Ni surface. It was also hypothesized that electron transfer could potentially proceed at the Ni-Fe interface. The results agreed well with the XPS and HR-TEM results. The presence of multiple Ni geometries tuned the extent of CO adsorption achieved. This can further adjust the efficiency of HDO of the lignin-derived oxygenates.The interfacial sites present in the catalysts have been described in Fig. 5(B). The sites were analyzed based on the characterization results. It was observed that the ease of formation of the anatase form of TiO2 in Ni/TiO2-HT and NiFe/TiO2-HT was higher than the ease of formation of the anatase form of TiO2 in Ni/TiO2-IM. Compared to the rutile form, it was easier to reduce the anatase form of TiO2 [72–74]. Better HDO activity was observed with the anatase form [74]. It was observed that the amorphous Fe oxides were highly dispersed on the surface of the TiO2 present in Fe(3)/TiO2-HT, which was devoid of Ni. The tiny amorphous Fe oxide clusters were composed of reduced Fe2+ species. Negligible amounts of Fe2O3 and FeTiO3 were present and a strong metal–support interaction with the TiO2 support was generated (Fig. S3(B)). Two types of oxygen vacancies can be generated at the Fe-TiO2 interface: (i) Fe2+–OV–Ti3+ vacancies are formed during the cation exchange process (followed by the reduction step) between Ti4+ and Fe3+ during the HT method and (ii) Feδ+–OV–Feδ+ were formed during the reduction process [75] (Fe/TiO2-HT, Fig. 5(B)). For monometallic Ni/TiO2-HT, small Ni NPs on the surface of TiO2 form strong metal–support interactions. A small amount of metallic Ni was observed on the terrace of the Ni NPs (Fig. 6(D)). Ni atoms at the edges or isolated sites were predominant in the Ni/TiO2-HT and NiFe/TiO2-HT catalysts. The Niδ−–OV–Tiδ+ oxygen vacancies (here, Niδ− species are electron-rich) (Ni/TiO2-HT, Fig. 5(B)) were formed during the cation exchange process between Ti4+ and Ni2+ when the HT method was used for fabrication. The cation exchange process was followed by the reduction process. Although the presence of amorphous TiOx was observed on the surface of the Ni NPs (Fig. 3(B4)), the formation of Tiδ+–OV–Tiδ+ was hindered by the presence of the electron-rich Ni NPs. The well-developed core–shell NiFeOx NPs were dispersed on the surface of the reducible TiO2 support present in NiFe/TiO2-HT. Highly dispersed oxophilic centers were formed, which closely interacted with Ni NPs and the TiO2 support. The formation of NiFe and Fe-TiO2 interfacial sites was observed, but the formation of the Ni-TiO2 interfacial sites was not observed (NiFe/TiO2-HT, Fig. 5(B)). The generation of the surface oxygen vacancies at the NiFeOx interface can be potentially attributed to the phenomenon of hydrogen spillover. The generation of the vacancies at the Fe-TiO2 interface may help compensate for the charge imbalance attributable to the partial substitution of Ti4+ by Fe3+.To achieve the selective conversion of guaiacol to cyclohexanol, the active catalytic sites must be controlled during the demethoxylation of Ar-OCH3 and the hydrogenation of aromatic rings. NiFe(3)/TiO2-HT exhibited excellent catalytic activity with high STYCHNOL (182.7 molCHNOL·molNi −1·h−1). The rate of production was 0.49 molCHNOL·gcatalyst −1·h−1. This is the maximum rate reported to date at reaction temperature below 300 °C (Table S1). The selectivity achieved was 85% and almost all the guaiacol molecules could be converted to the desired product at 270 °C under 5 MPa of H2 (measured at room temperature) (Fig. 7 (A); Xfeed and Sp roduct, Table 1). HDO of guaiacol was allowed to proceed at 270 °C under 5 MPa (H2). The reaction was allowed to proceed for 1 h using the model catalysts to understand the roles of NiFe/TiO2-HT on the selectivity of the cyclohexanol formation (demethoxylation–hydrogenation) reaction.Monometallic Ni/TiO2-HT exhibited a high HDO activity (Fig. 7(A)). The STYCHNOL was 95.6 molCHNOL·molNi −1·h−1 and the selectivity was 60% (guaiacol conversion: 97.2%; entry 4, Table 1). When the reactions were performed using monometallic Ni/TiO2-IM, the cyclohexanol yield was 17.9% and the guaiacol conversion was 39.8% (entry 5, Table 2 ). In the absence of catalysts, negligible guaiacol conversion was observed, indicating that metals are required for effective conversion (entries 1–3, Table 2). The characterization results revealed that high Ni dispersion (DNi = 24.4%) and low surface coverage (θ Ni = 81.1%) could be achieved using Ni/TiO2-HT. The Ni dispersion was greater and the surface coverage achieved was lower than those achieved using the Ni/TiO2-IM catalysts. The good guaiacol conversion (to cyclohexanol) in the presence of Ni/TiO2-HT could be attributed to the presence of abundant HDO active centers in the system. The generation of a large number of active centers could be attributed to the close interaction existing between the Ni NPs (during hydrogenation) and the oxophilic centers (Niδ−–OV–Tiδ+) (during demethoxylation) present at the Ni-TiO2 interface. The low cyclohexanol yield achieved using Ni/TiO2-IM could be attributed to the presence of the inactive Ni surface covered by the TiO2 overlayers and the low density of the oxophilic centers participating in HDO.The effects of Ni loading on the efficiency of the Ni/TiO2-HT catalyst were investigated. It was observed that the cyclohexanol yield could be potentially improved by increasing Ni loading. An increase in the Ni loading helps promote the demethoxylation reaction by forming active centers for HDO at the Ni-TiO2 interface. When the Ni loading was increased (Fig. S7(A)), the cyclohexanol yield increased (yield range: 55.0–60.0%; at 15–20 wt% Ni). Under these conditions, the yield of 2-methoxycyclohexanol reached a constant (yield range: 16.6–20.0%; at 10–20 wt% Ni), but the quantity of hydrogenating Ni increased. Significant changes in the product yield and guaiacol conversion were observed when the Ni loading was increased (loading ≤ 15 wt%). When the Ni loading was increased to 20 wt%, significant changes in product selectivity were not observed. However, under these conditions, the yield of cyclohexane increased.The HDO of guaiacol was performed using Fe/TiO2-HT (in the absence of Ni) to understand the roles of Fe. The catalyst helped achieve high 1,2-dimethoxybenzene selectivity (80%) and 9.8% guaiacol conversion (Fig. 7(A); entry 4, Table 2). Other compounds such as 1,2-benzenediols (7), phenols (2), methylated phenols (8), and methanol were formed in less amounts (Fig. S8(A)). The selective formation of 1,2-dimethoxybenzene (6) indicates that transmethylation is favored in the presence of activated methyl (–CH3) radicals which are formed during the demethoxylation of guaiacol.During the demethoxylation, guaiacol is converted to phenol (or catechol). During this process, –CH3 radicals are adsorbed on the surface of the Fe/TiO2 catalyst (Fig. 8 C(i) and S8(B)). The CH3 radicals present on the catalyst surface can be used for transmethylating the (i) adsorbed guaiacol molecules to form 1,2,-dimethoxybenzene (7) and (ii) aromatic rings of phenols to produce methylated phenols (9) (Fig. 1(A)). The cleavage of Ar-OCH3 can potentially take place at the oxygen vacancies present at the perimeter of the reduced FeOx site. The cleavage follows the reverse Mars–van Krevelen mechanism [75]. The Fe species influence the efficiency of conversion of guaiacol. The formation of small Fe oxide particles containing reduced Fe2+ species was observed using the XRD (Fig. S3(C)) and XPS results (Fe 2p, Fig. 5). The results suggest the presence of highly dispersed oxophilic centers at the oxygen vacancies present in Fe oxides. The demethoxylation reactions conducted using the NiFe/TiO2 catalysts were promoted under these conditions.As highly dispersed oxophilic centers were present in the Fe species and the Ni/TiO2-HT catalyst (15 wt% Ni and 1–5 wt% Fe) exhibited good hydrogenation activity, the catalysts were used for the HDO of guaiacol. An increase in the cyclohexanol yield was observed as the Fe content was increased. The STYCHNOL (182.7 molCHNOL∙molNi −1∙h−1) achieved using 3 wt% Fe (1.5 wt% measured Fe content; entry 6, Table 1) was two-fold higher than that achieved using the monometallic Ni/TiO2-HT. As depicted in Fig. S7(B), the yield of 2-methoxycyclohexanol decreased (from 19.1% to 0.8%) as the Fe loading was increased (from 0 wt% to 5 wt%). The cyclohexanol yield increased (from 72.6% to 87.9%) with an increase in the Fe loading.A low HDO activity was observed for NiFe(5)/TiO2-HT (cyclohexanol yield: 72.1%; guaiacol conversion: 88.4%; entry 12, Table 2). The low activity hindered the formation of 2-methoyxcyclohexanol (following the direct hydrogenation) from guaiacol. The lower HDO activity of NiFe(5)/TiO2-HT (compared to the activity of Ni/TiO2-HT, Table 1) can be potentially attributed to the decreased Ni dispersion (from 19.73% to 17.46%), increased Ni surface coverage (from 81.1% to 89.8%), and decreased degree of reduction (from 76.1% to 29.9%). Although the HDO activity was suppressed, the STYCHNOL achieved with NiFe(5)/TiO2-HT was high (167.1 molCHNOL∙molNi −1∙h−1). The high STYCHNOL values confirmed that the active sites were related to the reduced Fe oxides present near the Ni NPs and TiO2. When Fe loading was further increased to 10 wt% and 15 wt%, the cyclohexanol yield and guaiacol conversion significantly decreased, indicating catalyst deactivation. The deactivation of the catalyst could be attributed to the accumulation of the inactive Fe oxides occupying the surface of Ni NPs.The results reveal that the ensembles of the NiFeOx and FeOx-TiO2 interfacial sites are selective HDO centers that promote the instantaneous demethoxylation–hydrogenation reactions. The highly electron-deficient sites (oxygen vacancies) present at the FeOx-TiO2 interface (Fe2+–OV–Tiδ+) help anchor the hydroxyl groups present in guaiacol. The oxygen vacancies present at the NiFeOx interface participate in the direct demethoxylation of guaiacol. The demethoxylation is followed by the hydrogenation of the phenyl ring (Fig. 8(C-iii)).The influence of the supports on the efficiency of HDO of guaiacol were studied using SiO2-, Al2O3-, Nb2O5-, ZrO2-, and TiO2-based supports (Fig. S7(C)). A high rate of cyclohexanol production (STYCHNOL = 160.6 molCHNOL·molNi −1·h−1) was achieved using NiFe/ZrO2-HT. The rate was slightly lower than the rate achieved using NiFe(3)/TiO2-HT (STYCHNOL = 182.7 molCHNOL∙molNi −1∙h−1). Other than TiO2 and ZrO2, NiFe catalysts on SiO2, Al2O3, and Nb2O5 could not be used for the effective HDO of guaiacol. This could be potentially attributed to the fact that the oxygen vacancies were not present on the surfaces of these non-reducible supports.The demethoxylation–hydrogenation of the phenyl rings was studied by analyzing the time-dependent HDO of guaiacol (Fig. 7(B)). The yield of cyclohexanol (produced with high selectively) increased (from 45.8% to 87.9%) with an increase in the reaction time during the first 40 min (includes the heating time of 20 min). This indicated that the formation of cyclohexanol was favored. The possible intermediates (2-methoxycyclohexanol (4) and cyclohexanone (3)) were formed in negligible amounts (Fig. 1(A)). The formation of phenol (2) was not observed under these conditions. These observations suggested that cyclohexanol (5) could be formed during the direct demethoxylation of guaiacol. The demethoxylation was followed by the rapid hydrogenation of the intermediate (2). Fig. 7(C) (representing the pressure-dependent HDO) suggests the formation of phenol following the direct demethoxylation of guaiacol. The low-pressure reaction performed under 2 MPa (H2; measured at room temperature) led to the production of phenol (selectivity: 20.6%) and cyclohexanone (selectivity: 31.6%). These compounds rapidly hydrogenated under the high pressure of 4 MPa (H2; measured at room temperature) even when the catalyst ratio was low and the temperature was 250 °C (entries 13 and 14, Table 2). The results confirmed that HDO could be achieved following the demethoxylation of guaiacol (leading to the production of phenol) at low H2 pressure.The reaction pathway was further investigated by tuning the reaction temperature (Fig. 7(D)). The yield of cyclohexanol increased from 63.5% to 87.9% as the reaction temperature increased from 230 °C to 270 °C. Approximately 98% of guaiacol could be converted to the desired product under these conditions. When the reaction temperature was increased to 290 °C, the cyclohexanol yield decreased to 17.9%. The formation of cyclohexanol (6) was accompanied by the formation of cyclohexane (Fig. 1(A); yield: 57.2%) and methylcyclohexanol (10; yield: 5.5%). These observations indicate that the demethoxylation–hydrogenation is favored when the temperature is ≤ 270 °C although complete hydrodeoxygenation occurs at 290 °C. The apparent activation energy (Ea) was determined (assuming first-order reaction kinetics for the conversion of guaiacol to cyclohexanol) to understand the efficiency of the selective HDO reaction. The Arrhenius plots (Fig. 8(A)) were analyzed, and it was observed that the Ea for the formation of cyclohexanol from guaiacol was the lowest (69.05 kJ/mol) when NiFe(3)/TiO2-HT was used as the catalyst and the Ea was high when Ni/TiO2-HT (78.66 kJ/mol) and Ni/TiO2-IM (91.74 kJ/mol) were used as the catalysts. These observations indicate that the active sites on NiFe/TiO2-HT can effectively cleave the Ar-OCH3 bond and hydrogenate the phenyl ring present in guaiacol. The different activation energy of Ni/TiO2-IM compared to Ni/TiO2-HT can be attributed to the structure difference by the catalyst preparation methods, which was confirmed by the characterization results of TEM, XRD, and CO DRIFTS. The different structure formed the different active sites, exhibiting the different activation energies.The feed-to-catalyst ratios were also tuned and high yields of cyclohexanol were obtained when large quantities of catalysts were used (Fig. S7(D)). Incomplete conversions of guaiacol and low yields of cyclohexanol were observed under conditions of high feed-to-catalyst ratios.The results revealed that the presence of the Fe species and Ni NPs in the core–shell structure of NiFe/TiO2-HT improved the generation of the oxygen vacancies at the perimeter of Feδ+, Tiδ+, and Ni NPs (Fig. 5(B)). The rate of formation of cyclohexanol increase with an increase in the Fe content (till 2.5 wt%). A typical volcano-shaped curve was generated, indicating that the demethoxylation–hydrogenation process was adjusted by the presence of Fe (Fig. 8(B)). When the Fe content was low (range: 0–1 wt%), the production of cyclohexanol was hindered as the C-O bond activation energy was high. When the Fe content was high (range: 2–2.5 wt%), the formation of cyclohexanol was hindered by the poisoning of the Ni surface by Fe. The maximum rate of cyclohexanol formation (STYCHNOL = 182.7 molCHNOL·molNi −1·h−1) was measured when 1.5 wt% Fe was used (actual content measured for NiFe(3)/TiO2-HT).Based on the results obtained using the XPS, H2 TPR, H2 TPD, and TPO measurements, it can be inferred that the incorporation of the Fe species led to the formation of a large number of oxygen vacancies at the interfacial sites. The amount of oxygen vacancies (OV/OL) was correlated with the Fe content, and it was observed that the OV/OL ratio increased with an increase in the rate of cyclohexanol formation. The trend in the change in the number of vacancies was similar to the trend observed when the Fe content was varied (Fig. 8(B)). These observations confirmed that the demethoxylation–hydrogenation activity was adjusted by the number of the surface oxygen vacancies.While the HDO of guaiacol can be achieved following the processes of demethylation, demethoxylation, hydrogenation, and hydrodeoxygenation [28,76], the formation of cyclohexanol can proceed via two major routes (Fig. 1(A)): (i) demethoxylation of guaiacol (1) to phenol (2) followed by the hydrogenation of the compound to form cyclohexanol (5) , and (ii) hydrogenation of guaiacol (1) to form 2-methoxycyclohexanol (4). The catalysts studied by us may not promote the demethoxylation of (4).The probable reaction mechanism followed (when the TiO2-supported catalysts were used) has been presented herein. When the Ni/TiO2-HT catalyst is used, the adsorption of guaiacol on the Ni surface can occur exploiting two different configurations (as inferred from the results obtained using various characterization methods): (i) The C=C bonds present in the phenyl ring in guaiacol can get adsorbed on the terrace surface, leading to ring hydrogenation in the presence of activated H atoms. This process leads to the production of 2-methoxycyclohexanol (Fig. 8(C-ii)). (ii) The methoxy or hydroxyl groups in guaiacol can get adsorbed on the defects of oxygen vacancies present at the electron-deficient Ni–OV–Ti interface to promote the demethoxylation reaction and not the dehydroxylation reaction (Fig. S8(C-ii)). As Ni/TiO2-HT contains Ni metal and the Ni–OV–Ti rich phase, routes (i) and (ii) could occur concomitantly on Ni/TiO2 producing a mixture of methoxycyclohexanol and cyclohexanol in the reaction product.For NiFe/TiO2-HT, the ternary structure of NiFe dispersed on TiO2 improved the adsorption of guaiacol. The methoxy (or hydroxyl) moieties in guaiacol improved the adsorption process. The highly dispersed FeOx species function as oxophilic centers and closely interacted with the methoxy group (Ar-OCH3) present in guaiacol. The NiFeOx (not pure Ni) present in the core–shell structure of NiFe/TiO2-HT suppressed the the hydrogenation of the phenyl ring. Improvement in the extent of demethoxylation at the NiFe interface was also observed under these conditions. An increase in the Fe content in the thick FeOx outer shell layer promoted the direct demethoxylation of guaiacol followed by the hydrogenation. A high rate of cyclohexanol formation can be attributed to the fact that the highly dispersed FeOx species can help in the generation of multiple interfacial sites at the NiFeOx and TiO2 interfaces, leading to the generation of multiple active centers that participate in demethoxylation (Fig. 8(C-iii) and Fig. S8(C-ii) without significantly poisoning the hydrogenation sites.HDO of pyrolysis oil proceeded via a two-step process (Fig. S1). Pyrolysis oil dissolved in n-decane was catalytically treated with 5 wt% Pd/C in the first step (BO-S1). The product, denoted as LO, containing alkyl methoxyphenols such as guaiacol, methyl guaiacol, ethyl guaiacol, propyl guaiacol, and propyl syringol as the major components, was obtained (Fig. S9(B)). In the absence of a catalyst, BO-S1-LO (Fig. S9(A)) consisted of alkyl methoxyphenols, eugenol, isougenol, and other small oxygenates (such as acetone, methyl acetate, 2,5-dimethylfuran, cyclopentanone, and methyl cyclopentanone). The product obtained at the end of the first step was analyzed using the GC. The results revealed that the amount of alkyl methoxyphenols (696.43 μg/gbio-oil) present in the LO fraction was higher than the amount of alkyl methoxyphenols present in the fraction obtained in the absence of catalysts (173.67 μg/gbio-oil). Hydrotreatment using 5 wt% Pd/C led to a low yield of solid coke and high yields of WSO (19 wt%) and HO (49 wt%) (Fig. 9 (A)). This indicated that recondensation reactions (C–C coupling of labile monomers) were suppressed under these conditions.HDO of the LO fractions was simulated using the phenolic model compounds, such as methyl guaiacol, propyl guaiacol, and allyl guaiacol. The reaction was performed using NiFe(3)/TiO2-HT as the catalyst, and the results indicated that the conversion of the alkyl guaiacols decreased as the length of the alkyl chains were increased. This can be attributed to the steric hindrance exerted by the alkyl chains (Fig. S7(E)).Based on the results obtained using the phenolic model compounds, the HDO of the LO fraction was conducted at the second stage of the reaction involving NiFe/TiO2-HT. The solvent was not removed during the process. The yield of the cyclic alcohols was 71.7%, and 93.0% of the alkyl guaiacol reactants could be converted (Fig. S9(C)). A low yield of the cyclic alcohols (20%) and a low guaiacol conversion (54%) could be achieved using Ni/TiO2-HT (Fig. 9(A)). This confirmed that NiFe/TiO2-HT was more effective (than Ni/TiO2-HT) and could be used to achieve the selective HDO of the LO fractions.The reusability of the NiFe/TiO2-HT catalyst (for the HDO of guaiacol) was studied under the optimized reaction conditions (270 °C; 5 MPa, H2; 1 h) (Fig. 9(B)). The guaiacol conversion and the yield of cyclohexanol slightly decreased after each cycle. At the end of the 5th cycle (6th run), the guaiacol conversion was 93.6%, and the yield of cyclohexanol was 83.5%. This indicated that an excellent catalyst reusability for the NiFe(3)/TiO2-HT catalyst could be achieved. Spent NiFe(3)/TiO2-HT was characterized using the TG and XRD measurements (Fig. 9(C) and (D)). The TG results revealed that the weight of fresh catalyst (under a flow of air) increased by 2.9 wt%, and this could be attributed to the oxidation of the Ni and Fe species. It was also observed that the weight of spent catalyst after the 6th run increased by only 0.8 wt% (vs. ∼20 wt% assuming the complete coking of guaiacol per run [77]) under conditions of coke deposition. This indicated negligible coking under the reaction conditions and this can be attributed to the absence of coke precursors such as phenol and catechol [77]. The 6.2 wt% increase in weight can be explained by the complete oxidation of metallic Ni and Fe species in NiFe(3)/TiO2-HT.The spent catalyst obtained following the HDO of BO-S1-LO (using Pd/C) contained 1.1 wt% coke (formed during the condensation of the labile monomers). The spent NiFe(3)/TiO2-HT samples were characterized using the XRD, which revealed negligible changes in the crystal structures. A close inspection of the peaks corresponding to Ni(111) indicates the presence of large particles (18.4 nm) and low 2θ (or higher d-spacing) values. These observations indicated that the Ni-based particles were slightly sintered and were further alloyed with Fe in the presence of Fe species. The segregation (followed by agglomeration) observed for the Ni-based particles was not observed for Fe2O3. Peaks corresponding to Fe2O3 were not observed in the XRD results. Significant changes in the crystal structure of TiO2 were not observed, either. These observations indicate the highly stable nature of NiFe/TiO2-HT. The XRD results for the spent catalyst (following HDO of BO-S1-LO (using Pd/C)) revealed that the fraction of the rutile TiO2 phase was larger than the fraction obtained using the fresh catalyst. The Ni particles present in the spent catalyst were larger than the Ni particles present in the fresh catalyst. Peaks corresponding to Fe2O3 were not observed for fresh and spent NiFe/TiO2-HT catalysts. These observations indicate that NiFe/TiO2-HT was stable, and the stability was slightly adjusted by the acids present in BO-S1-LO.The NiFeOx core–shell (ternary heterostructure) structure on the TiO2 support, prepared following the HT method, exhibited excellent chemoselective catalytic activity and stability during the direct demethoxylation–hydrogenation process followed to produce cyclic alcohols from BO-derived alkyl methoxyphenol complex mixtures. The Ni metal active sites, tuned by the presence of highly dispersed Fe species in a core–shell environment on reducible anatase TiO2, formed complexes with Fe and TiO2, leading to the generation of abundant oxygen vacancies at the NiFeOx-TiO2 interfacial sites. This led to the highly selective direct demethoxylation–hydrogenation reactions that could be conducted to obtain cyclic alcohols. The results reported herein reveal the synergism between metal–metal oxide support interfacial sites in NiFeOx/TiO2. The effects can be tuned to selectively form cyclic alcohols from lignin-derived phenolic oxygenates and BO.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 Technology Development Program to Solve Climate Changes (2020M1A2A2079798) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT of Republic of Korea.Supplementary data to this article can be found online at https://doi.org/10.1016/j.cej.2022.136578.The following are the Supplementary data to this article: Supplementary data 1

TiO2-supported Ni catalysts are promising candidates that can be used to achieve biomass valorization following the selective hydrodeoxygenation (HDO). Their catalytic activity can be tuned and they are characterized by strong metal–support interaction (SMSI). The SMSI observed at the interfaces of Ni nanoparticles (NPs) and the TiO2 support was tuned by adding Fe and subjecting the synthesis system to hydrothermal treatment conditions. The prepared catalysts promoted the selective conversion of alkyl methoxyphenols, including lignin-derived guaiacol, to saturated cyclic alcohols. The low Fe content (∼1.5 wt%) in Ni/TiO2 significantly promoted the formation of cyclohexanol from guaiacol (rate: 183 molcyclohexanol·molNi −1·h−1; selectivity: 85%). Alkyl methoxyphenols present in the biomass pyrolysis oil could be effectively converted into cyclic alcohols (yield: 71%). It was observed that the interfacial sites of highly dispersed NiFe-FeOx core–shell structures on the TiO2 support promote the demethoxylation of reactant to selectively produce cyclic alcohols.

The combinatorial materials chip (CMC) method [1], combined with high-throughput characterization techniques, has been proven a powerful tool for rapid material studies such as in phase diagrams [2], ferromagnetic shape memory alloys [3], and amorphous alloy [4], etc. With this approach, the magnetism of FeCo-based alloys was rapidly screened [5–8] as candidates for rare-earth-free permanent magnet materials. In these researches, W, Mo, Nb, and V were added to the FeCo-based alloys as dopped elements. Because of that the added refractory elements are found to have strong spin-orbit couplings, which may result in high magnetocrystalline anisotropy in FeCo-based alloys by hybridization [9–12].FeNi-based alloy is a class of important materials in engineering. As permalloy, it exhibits superparamagnetic properties and thus can be used for various applications related to drug delivery and magnetic sensors [13]. However, the magnetic properties of Fe-X-Ni (X stands for the third element) ternary systems were studied in a systematical way only in rare occasions [14], because it required a great deal of labor and cost under the premise of the traditional trial-and-error approach. In our previous work [15], the magnetic properties of the CMC of the Fe-Co-Ni ternary system were characterized systematically with an HT-MOKE system. Combined with the composition data by high-throughput μ-XRF and structural data by high-throughput XRD, the correlation between composition, crystal structure, and the magnetic property was established and readily visualized. With such a panoramic view, the compositional regions exhibiting distinctive magnetic properties such as large coercive force, high saturation magnetization, etc. were revealed.In this work, the magnetic properties of CMCs of the Fe-X-Ni (X = Cr, W and V) ternary systems covering the full composition range were screened systematically by measuring the in-plane and out-of-plane hysteresis loops using an HT-MOKE system. Maps of composition-phase-magnetic properties were thus constructed very rapidly. Through these relationships, the effect of sample forms, MOKE modes, alloying elements, and heat treatment conditions on the structure and magnetic performance were systematically studied.Following a procedure described by Xing et al. [16], the Fe-X-Ni thin-film CMCs covering the full composition range were deposited on quartz glass substrates (25 mm × 25 mm × 2 mm) at room temperature with a base vacuum pressure of 1 × 10−7 Torr (1 Torr = 133.322 368 4 Pa). The chip is composed of 10-cycle of Fe, X (= Cr, W, V), and Ni layers with a total thickness of about 100 nm. As shown in Fig. 1 a, the Fe-Cr-Ni ternary region is an equilateral triangle of 20 mm side length in the center, surrounded by three unary regions (Fe, Cr, and Ni) and three binary regions (Fe-Cr, Fe-Ni, and Cr-Ni) separated by laser marking lines drawn after deposition. The composition C i (Mole fraction/%) at each point on the chip is determined by the thickness of each component [16] computed by the following equation: (1) C i = t i ρ i ∕ z i Σ i t i ρ i ∕ z i ∑ i t i = const  where t i is the thickness, ρ i the density, and z i the atomic mass of the ith component. The chips were annealed isothermally under dynamic pumping to maintain the vacuum at 1 × 10−7 Torr followed by rapid cooling after deposition.Time-of-flight secondary ion mass spectrometry (TOF-SIMS) depth profiling was conducted to evaluate the degree of interdiffusion between layers. The TOF-SIMS (TOF-SIMS 5–100, IONTOF GmbH, Germany) was operated in in-depth profile mode under the base pressure of the analysis chamber below 1.1 × 10−9 mbar (1 bar = 0.1 MPa). The layer-by-layer pealing was performed with an ion beam raster on 300 μm × 300 μm areas using 1 keV O (69.27–82.74 nA current). As shown in Fig. 1b, TOF-SIMS elemental depth profiles of Fe, Cr, and Ni are quite flat in the majority of the depths except for the very top cycle, indicating that interdiffusion is essentially complete in the thickness direction of the Fe-Cr-Ni system. Microbeam X-ray fluorescence (μ-XRF) spectroscopy (M4 TORNADO, Bruker Co., Germany) was employed to calibrate the composition distribution of the chip with a step size of 50 μm. A linear distribution of each element over the chip surface was confirmed in Supplementary Materials.For structural characterization, part of the samples was mapped by a combined XRD + XRF device on BL09B at the Shanghai Synchrotron Radiation Facility (SSRF), China. Micro-beam X-ray with an energy of 12 keV (λ = 0.103 3 nm) was focused by a pair of KB mirrors to 24 × 35 μm2. The incident angle of X-ray to sample surface was set to 15° and diffraction pattern was collected by a Dectris Pilatus 3S 2M detector covering a 2θ range from 16° to 52°. The XRF signal was collected by an energy-dispersive silicon drift detector simultaneously. The rest of the samples were characterized by a laboratory micro-beam X-ray diffractometer (Bruker D8 advance) with a rotating anode (Cu K α radiation) and area detector, whose beam diameter is 300 μm.The characterization of the magnetic hysteresis loops of the CMCs was conducted on a home-built HT-MOKE system, where a He-Ne laser with a wavelength of 632.8 nm is used as the light source. The beam was focused to 300 μm diameter and linearly polarized by a polarizer placed at an angle of 0° (parallel to the sample surface). After being reflected by the sample, the elliptically polarized light passes through a photoelastic modulator (PEM) (II/FS50LR, Hinds Co.) with a modulation frequency of 50 kHz, and a phase delay parameter A 0 = 2.405 rad. After modulation, the light passes through the analyzer set at 45° and then is received by the detector. A lock-in amplifier is employed to improve the signal-to-noise ratio. The mathematical derivation process of the Kerr signal can be seen in the Supplementary Materials.The magnetic field is imposed by a 4-pole electromagnet composed of two pairs of electromagnets arranged at 90° from each other. By controlling the intensity ratio of the two electromagnet pairs, the resultant magnetic field vector relative to the sample is adjustable so that the measurement can be switched between the transverse (in-plane) and polar (out-of-plane) mode. Typically, polar geometry has a much larger Kerr effect [17,18] and is commonly used in MOKE experiments. The magnetic field is scanned between −0.5 T and +0.5 T during hysteresis loop measurement.MOKE magnetic hysteresis loops are mapped automatically using a custom-designed sample stage system. The combinatorial chip is mounted on a sample holder with a 20 cm long aluminum shaft and positioned in the center of the 4-pole electromagnet for measurement. The shaft is attached to a three-dimensional stage system (Zolix TSA50-B stage) located on the side of the electromagnet, with a horizontal movement in the X-Y plane and vertical movement in the Z-axis. The sample can rotate around the x-axis or y-axis to change sample orientations. The whole process was controlled by a LabVIEW program with each data point taking about 60 s to measure. Python programs were written to automate the visualization of hysteresis maps and data analysis.A coordinate transform system (CTS) capable of correlating multiple instruments is required to ensure measurements are performed in the same position. To establish the correlation between two coordination, the three vertices of the ternary region triangle were first marked using a laser marking machine as the reference point. Considering the coordinate value of any point (x, y, z) in the rectangular Cartesians coordinate system O 1 -x 1 y 1 z 1 (sample coordinate system) and O 2 -x 2 y 2 z 2 (equipment coordinate system) are noted as coordinate sets U i and X i , respectively. The conversion between the two systems is expressed as (2) X i = R ∗ U i + T in which R is the rotation matrix and T the translation matrix. Using the coordination data for the three reference points, R and T can be obtained mathematically based on the Singular Value Decomposition (SVD) method [19]. This relationship is extendable to any two coordinates.For comparison purposes, bulk samples of 5 selected compositions were prepared by arc melting and drop-casting under a pure Ar atmosphere. Each bulk sample weighed about 20 g, and the purity of each of the raw materials was at least 99.9%. The as-cast ingots were homogenized at 1 200 °C for 2 h followed by cold rolling at room temperature with 75 % thickness reduction and subsequent recrystallization annealing at 1 000 °C for 0.5 h to refine the grain size. Optical microscopy (OM) was used for microstructure observation on specimens metallurgically polished and then etched with aqua regia (HCl: HNO3: C3H8O3 = 3:1:1) for 10–30 s. Magnetic properties were measured on a magnetic performance measurement system (Quantum Design, MPMS3) in fields up to 2 T along the long axis of samples having a dimension of 2 mm× 2 mm× 4 mm. Phase constitution of the bulk samples was identified by XRD (Rigaku Ultima IV) using Cu K α radiation with a scanning speed of 5 (°)/min.According to the direction of the magnetization and the plane of incidence, MOKE measurement can be conducted in longitudinal, transverse, and polar modes. As shown in Fig. 2 a, in the transverse, and longitudinal geometries, the in-plane signal is obtained, in which the magnetization is perpendicular (transverse) or parallel (longitudinal) to the plane of incidence. In the polar geometry, the out-of-plane signal is detected. The magnetization is parallel to the light propagation direction and perpendicular to the surface of the sample. The polar mode usually shows the largest Kerr effect followed by the longitudinal mode and the transverse mode. In this study, due to the restriction of the geometric setting, the longitudinal mode cannot achieve high-throughput measurement. Therefore, we use the transverse (T-MOKE) and polar (P-MOKE) modes to obtain the in-plane and out-of-plane Kerr signals.As shown in Fig. 3 a, 210 hysteresis loops in transverse mode (in-plane), color-coded with the crystal structure, are displayed for the Fe-Cr-Ni CMC sample annealed at 600 °C for 2 h. 5050 points of crystal structure data were characterized by synchrotron radiation at SSRF. With the help of CTS, phase structure data were integrated with the hysteresis loops in the map (Fig. 3a). The Fe-Cr-Ni ternary system is divided into three-phase regions including fcc and bcc two single-phase regions and fcc + bcc one two-phase region. The fcc single-phase region (color-coded red) resides on the Ni-rich portion and the bcc single-phase region (color-coded black) mainly distributes along the Fe-Cr edge with the fcc + bcc two-phase region (color-coded blue) in between (Fig. 3a).Based on the shapes of hysteresis loops, ferromagnetism is mainly found in the fcc single-phase region and the other regions exhibit weak ferromagnetic or paramagnetic characteristics. Fig. 3b shows that the largest saturation magnetization appears on the Ni-side along the Fe-Ni edge where the Fe concentration ranges between 10% and 40%. The saturation magnetization decreases with increasing Cr content because Cr is an antiferromagnetic element. The coercivity distribution displays a certain trade-off with the saturation magnetization even though the largest coercivity is ∼36 mT, which appears at the boundary between the bcc + fcc two-phase region and fcc single-phase region (Fig. 3c).To verify the results of thin-films by MOKE, the magnetic performance of some selected bulk alloys was compared with the CMC. Five compositions were selected from the fcc single-phase region in Fig. 3a, labeled as #1-#5 with an order of increasing Cr content, namely Fe38.7Cr5.8Ni55.5, Fe56.6Cr6.7Ni36.7, Fe41.7Cr11.7Ni46.6, Fe26.7Cr11.7Ni61.6, and Fe41.6Cr21.7Ni36.7. After recrystallization annealing, all the five alloys are fcc single phase and the grain size is ∼50 μm (Fig. 4a and b). The hysteresis loops by VSM show that four of the five alloys are ferromagnetic while one alloy with the highest Cr content is non-magnetic (Fig. 4c). The saturation magnetizations of the four alloys decrease with increasing Cr content which agrees with the CMC results. The saturation magnetizations of the #1 (Fe38.7Cr5·8Ni55.5) and #2 (Fe56·6Cr6·7Ni36.7) alloys are similar, ∼0.9 T despite their slightly different Cr content. The saturation magnetizations of the #3 and #4 alloys are ∼0.7 T and ∼0.6 T respectively due to the increased Cr content. Compared with the hysteresis loops of the corresponding compositions measured by T-MOKE (Fig. 4d), there is an agreement with the magnetization in general. However, the signals by MOKE measurement are much noisier, which makes a detailed comparison of magnetization as a function of composition impossible. In addition, the measured coercivities of the bulk alloys are almost neglectable while the MOKE measurements showed a coercivity of ∼20 mT for all the compositions. This can be attributed to the grain size effect because the grain size in the CMC of ∼100 nm thick is at the nano-meter scale and is at the micro-meter scale in the bulk form.The hysteresis loops measured in the polar mode (out-of-plane) on the same Fe-Cr-Ni CMC display certain differences from those in the transverse mode (in-plane) (Fig. 5 a). On the one hand, the signal-to-noise ratio of the polar mode is higher than the transverse mode. The saturation magnetization decreases with increasing Cr content (Fig. 5c) and the highest saturation magnetization resides in a similar region as measured by the transverse mode. On the other hand, relatively large coercivity whose maximum is ∼400 mT is found in the region with Ni concentration between 25% and 75% and Cr concentration of ∼20% (Fig. 5d). Fig. 5b shows the out-of-plane hysteresis loops of the same composition points #1–#5 as in Fig. 4d. The signal-to-noise ratio is ∼10 times higher than in Fig. 4d, which makes the plots more recognizable. Among them, #1 and #4 compositions have the maximum saturation magnetization followed by #2 and #3, and #5 with the highest Cr content shows the lowest value. This is roughly in agreement with Fig. 4d whereas the sequence of the saturation magnetizations of #1–#4 compositions is different from the bulk alloys (Fig. 4c). Unlike the bulk alloys and the hysteresis loops measured in the transverse mode, the five compositions show a coercivity from 70 mT to 310 mT, which is in reverse relationship with their saturation magnetization. The relatively large coercivity observed in the polar mode is probably due to the size effect because the film thickness is ∼100 nm and the beam size of MOKE measurement is ∼300 μm. Fackler [20] also showed that the out-of-plane coercivity is greater than the in-plane one either in the MOKE test or VSM test for thin-film. In the thin-film, the shape anisotropy effect is particularly important, where the shape of the magnet is essentially an infinite plane.The out-of-plane hysteresis loop maps of the Fe-W-Ni and Fe-V-Ni systems show a decrease in saturation magnetization with increasing W or V content like the Fe-Cr-Ni system as a general trend (Fig. 6 ). The compositions with obvious ferromagnetism were mainly distributed in the fcc single-phase region. The largest saturation magnetization values are distributed along the Fe-Ni edge where the distributed range of the region is slightly different between the two systems (Fig. 6b and e). A relatively large coercivity of ∼300 mT was found in an area where W is less than 20 % and Fe is between 20% and 70% in the Fe-W-Ni system. The overall coercivity of the Fe-V-Ni system was extremely weak and distributed evenly except for several abnormal points at the Fe-rich and Ni-rich ends. As can be seen from the contour map Fig. 6c, the shape of the area with a large coercivity (more than 150 mT) was similar to that of the Fe-Cr-Ni system (Fig. 5d) with a maximum value of about 300 mT.It is well-understood that while FeNi bulk alloys are soft magnetic and Cr, W and V are non-magnetic elements, when a small amount of them is added to FeNi-based alloys, the magnetic moment of ferromagnetic materials would be diluted, leading to a decreased saturation magnetization. This agrees with both in-plane and out-of-plane MOKE measurements of the CMCs. However, relatively large coercivity was found in the Fe-Cr-Ni and Fe-W-Ni systems in the out-of-plane measurements. Similarly, a scenario was observed in the Fe-Co-Ni CMC in our previous work [15]. The characteristic of magnetism is a strong function of both composition and crystal structure. The saturation magnetization is mainly composition-dependent while the coercivity can be sensitive to microstructure. This on the one hand can be explained by the film thickness effect (shape anisotropy) in the out-of-plane measurement. On the other hand, there is a microstructure effect in the corresponding CMCs. Fig. 7 compares the 2D diffraction patterns in the fcc single-phase region in the three systems heat-treated under the same conditions. The #1 composition corresponds to Fe38.7 X 5.8Ni55.5, #3 Fe41.7 X 11.7Ni46.6, and #5 Fe41.6 X 21.7Ni36.72 where X = Cr, W, and Ni. In the Fe-Cr-Ni system (Fig. 7a-c), all three compositions showed strong (111) texture. In the Fe-W-Ni system (Fig. 7d-f), the intensity (111) diffraction distributed evenly in #1 indicating weak texture while texture became strong in #3 and #5 with increasing W content. In comparison, the textures of the three compositions were all weak in the Fe-V-Ni system (Fig. 7g-i), and the diffraction intensity decreased with the increase of V content. Correlating the composition and structural characteristics with magnetic properties, it appears that the coercivity is closely related to crystal orientation and crystallinity. Because the (111) plane in fcc structure is a closely packed plane with low surface energy, the thin film tends to form the (111) texture and such a tendency may vary depending on the type of and the content of the alloying element.The Fe-Co-X (X = W [5], Mo [6], Nb [7], and V [8]) systems have been surveyed by a similar method trying to design a magnetic crystal with sufficient magnetization and a large magnetic anisotropy without the use of rare-earth elements. Similar to the Fe-X-Ni systems in this study, Fe-Co-X ferromagnetic materials are distributed in the region with low X element concentrations. Among these systems, the coercivity of 230 mT in the Fe-Co-W system and 260 mT in the Fe-Co-V system were obtained in the out-of-plane direction, which is in the same order of magnitude as what was found in this work. It should be noted that the annealing time of the Fe-Co-X systems was not always clear. For instance, The Fe-Co-Nb system was annealed at 700 °C for 1 h, the Fe-Co-Mo system, at 700 °C for 45 min and the Fe-Co-W and Fe-Co-V systems, at 600 °C or/and 700 °C [5–8]. In fact, heat treatment condition is closely related to magnetic properties, which is to be discussed in the next section. In the Fe-Co-W system, the vertically standing platelet-like grain structure was ascribed to the enhanced coercivity. The increase of coercivity in the Fe-Co-V system was explained by the shape anisotropy from column-shaped grains. A new noncubic ferromagnetic phase with a hexagonal crystal structure (C36) embedded in a FeCo-based matrix was identified in the Fe-Co-Nb system. These suggest that the coercivity is sensitive to the local microstructure of the corresponding system, including the crystal orientation and crystallinity.To study the effect of heat treatment conditions on magnetic properties, the #3 composition in the Fe-Cr-Ni system was chosen to compare systematically under different heat treatment conditions (600 °C, 700 °C and 800 °C for 1 h and 700 °C for 4 h, respectively). The hysteresis loops were collected in polar mode due to their high signal-to-noise ratio. As shown in Fig. 8 a, when the heat treatment time is fixed for 1 h, the magnetic properties of all samples are roughly similar despite the annealing temperature increasing from 600 °C to 800 °C. The coercivity is increased from 230 mT for annealing at 600 °C to 350 mT at 700 °C but remains almost unchanged at 800 °C, or for an extended time. For the remanence, an average of 6 was obtained for all samples annealing at 600–800 °C for 1 h. However, the value of remanence rose to 19 with annealing at 700 °C for 4 h. In addition, under 600–800 °C for 1 h heat treatment, the hysteresis loops of samples are barely smooth, and the shape is less regular at 600 °C, for 1 h treatment. Both the shape of the hysteresis loop and the squareness are better when sample annealing at 700 °C for 4 h. Meanwhile, it can be seen from the #3 hysteresis loop in Fig. 5b that the coercivity at the same composition point after annealing at 600 °C for 2 h is 240 mT and the remanence is 14 The results indicate that coercivity has a stronger dependency on temperature. However, the remanence or saturation magnetization is more closely related to the duration of heat treatment but is not sensitive to the heating temperature.To find out the structural difference between the materials under different heat treatment conditions, the XRD spots (Fig. 8b-d) are compared. Two peaks corresponding to (111) and (200) of the fcc phase, respectively, exist for the samples annealed at 600–800 °C for 1 h. Obviously, the fcc (200) spot for the sample annealed at 600 °C for 1 h showed a wider 2θ range than that at 700 °C and 800 °C for 1 h, indicating more residual stress for annealing at 600 °C for 1 h, which caused that the hysteresis loop at 600 °C looked more irregular (Fig. 7b, c & d). However, the (200) peak disappeared for the sample annealed at 700 °C for 4 h (Fig. 8e). Meanwhile, the remaining ring of (111) peak is more aggregated than that of the samples annealed at 600–800 °C for 1 h. This indicates that the grains grow preferentially under this extended heat treatment. In general, grains grow along with a given orientation due to interface and surface energy. Hence, the whole film will move towards its thermodynamically most stable state by orienting all grains with the plane of lowest surface tension parallel to the substrate. As shown in Fig. 2b and c, for the fcc crystal structure, the close-packed (111) surface energy is the lowest energy surface among all indices. Thus, the films with fcc structure show a strong [111] preferred orientation perpendicular to the surface of the sample. However, the easy axis [111] of the fcc phase is parallel to the magnetization direction of the P-MOKE. Meanwhile, the degree of crystallization of the material is positively correlated with the heat treatment time. The higher the crystallinity, the greater the coercivity [21]. Therefore, materials annealed at 700 °C for 4 h show better magnetic properties than that of 600–800 °C for 1 h may be attributed to the texture and crystallinity of materials. This also explains why in Section 3.1, the saturation magnetization of the film sample is slightly higher than that of the bulk, especially #5, due to the texture of the film.A series of Fe-X-Ni (X = Cr, W and V) CMC samples were fabricated, covering the full range of component elements. After heat treatment with different conditions, the phase structure was characterized by XRD, and the magnetic properties were characterized by a self-built HT-MOKE system. The composition-phase-magnetic properties relationship maps were established for the Fe-X-Ni (X = Cr, W and V) systems. Based on these relationships, the effects of sample forms, MOKE modes, alloying elements, and heat treatment conditions on magnetic properties were systematically studied.The results showed that the saturation magnetization of all systems has a strong dependence on alloying elements, and generally decreases with increasing Cr, W, and V content. This is due to the magnetic moment of ferromagnetic materials being diluted by these added non-magnetic elements. The largest saturation magnetization appeared in the fcc single-phase region and is on the Ni-rich side near the Fe-Ni edge in all three systems. A large difference in coercivity between the in-plane (transverse mode) and out-of-plane (polar mode) hysteresis loops was noticed due to the shape anisotropy. At Fe37Cr21Ni42, the maximum coercivity was found. The value is ∼36 mT by the in-plane measurements and it increases to ∼400 mT by the out-of-plane measurements. The trend of magnetic properties in CMC samples was not totally repeated on 5 bulk samples with selected compositions surrounding Fe37Cr21Ni42. Although the trend of saturation magnetization in bulk is in good agreement with that from CMC, all bulk samples show almost no coercivity, attributable to the much smaller grain size, grain shape anisotropy, and stronger texture in CMC samples. Comparing the Fe-X-Ni systems under a similar condition, Cr alloying obtained the largest coercivity followed by W alloying and then V alloying. The maximum out-of-plane coercivities in the Fe-W-Ni and Fe-V-Ni systems were ∼300 mT and ∼200 mT, respectively. We suggest that alloying with different elements leads to the diverse orientation and crystallinity of the fcc phase resulting in different magnetic properties.Meanwhile, the effect of heat treatment on magnetic properties was evaluated in the Fe-Cr-Ni system. The materials annealed at 700 °C for 4 h show better saturation magnetization than that of 600 °C, 700 °C and 800 °C for 1 h, indicating that coercivity has a stronger dependency on temperature. However, the remanence or saturation magnetization is more closely related to the duration of heat treatment but is not sensitive to the heating temperature. This can be explained by the highly crystallized fcc structure and the strong (111) texture formed.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 from the National Key Research and Development Program of China (Grant Nos. 2021YFB3702102 and 2017YFB0701900) and the Major Science and Technology Project of Yunnan Province “Genome Engineering of Rare and Precious Metal Materials in Yunnan Province (Phase One 2020)" (Grant No. 202002AB080001-1). Boyue Instruments (Shanghai) Co., Ltd for support of μ-XRF is also acknowledged.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.jmat.2022.07.006.

Fe-X-Ni (X = Cr, W and V) combinatorial thin-film (∼100 nm thick) materials chips covering the full composition range of ternary systems were fabricated. The crystal structure distribution was mapped by micro-beam X-ray diffractometers (XRD) and the magnetic hysteresis loops over the chip were characterized by a high-throughput magneto-optical Kerr effect (HT-MOKE) system to establish the composition-phase-magnetic properties relationships. The results showed that saturation magnetization for all systems has a strong dependency on alloying composition, and decreases with increasing dopped elements content as a general trend. Although the trend of saturation magnetization in bulk is in good agreement with that from thin films, all bulk samples show almost no coercivity, attributable to the much smaller grain size, and stronger texture in thin-film samples. Comparing the Fe-X-Ni systems under a similar condition, in the out-of-plane, Cr alloying obtained the largest coercivity (∼400 mT) followed by W alloying (∼300 mT) and then V alloying (∼200 mT). We suggest that alloying with different elements leads to the diverse orientation and crystallinity of the fcc phase resulting in different magnetic properties. Meanwhile, the effect of heat treatment on magnetic properties indicates that saturation magnetization is more closely related to the duration of heat treatment.

Chemical reactions are always around our lives, and more importantly, they affect and even change our lives. 1 Almost 80% of synthetic chemical reactions are dependent on the efficiency of catalysts. 2 Compared with traditional trial-and-error methods, the rational design of catalysts based on efficient descriptors for catalytic activity could improve the reaction efficiency, 3 decrease the cost of catalysts, and result in significant economic benefit. 4 In the long history of the development of catalysts, single-atom catalysts (SACs) have shown the most potential as the altar of catalysts because of their high active atom utilization, high catalytic activity, and high catalytic selectivity. 5–7 Except for the difficult synthesis of SACs with high loading and high stability, the optimal SAC for a specific reaction is fairly random and unpredictable, 8 which hinders the development of SACs in the field of catalysis. 9 , 10 Some effective strategies have been proposed to address this issue; however, these designed SACs are still mainly based on traditional trial-and-error methods. 11 , 12 Even though some insights on active-metal atoms have been developed, 13 , 14 the activation mechanism of reactants for SACs remains obscure. The catalytic activity of SACs is not only dependent on the active atoms but also heavily influenced by the substrates and metal-substrate interactions. 15 , 16 A number of theories and descriptors have been proposed for predicting the adsorption energy and for further determining the catalytic activity of corresponding reactions. 13 , 17–20 For example, the d band center model has been widely applied to metal-based catalysts. 21 Unfortunately, the descriptor cannot be effectively applied to SACs because the d band structures of active-metal atoms are drastically influenced by the substrate. 19 , 22 Moreover, Gao et al. proposed a model with a new electronic descriptor of ψ and the generalized coordination number (CN) of active sites for quantitatively predicting the adsorption energies of small molecules on metallic materials and oxides. 17 For SACs, Xu et al. presented a universal design principle for a rational design of graphene-based SACs using the coordination number, the electronegativity of active center, and the electronegativity of the nearest neighbor atoms. 13 These efforts have made a great contribution for rational catalyst design from the two aspects of atomic and electronic structures. However, identifying the real atomic structures of coordination environments is too complex, especially during the catalytic reaction process.Chemical reactions include the breakage of chemical bonds in the reactants and the formation of new chemical bonds in the products, which are essentially a process of electronic redistribution. The role of the catalyst is to guide the redistribution of electrons purposefully. Thus, the catalytic activity of SACs is expected to be predicted only through the intrinsic electronic characteristics of single metal atoms and the substrates without consideration of the complex coordination environment, where these electronic characteristic parameters are easily available through publicly available databases.In this study, we constructed 126 SACs with 9 two-dimensional (2D) substrates and 14 metal atoms for developing a machine learning (ML) model based on density functional theory (DFT) calculation. The nitrogen reduction reaction (NRR) is taken as the model reaction because the product of NH3 is vital for many important chemicals, particularly fertilizers and potential hydrogen storage materials. 23 , 24 The developed ML model takes intrinsic electronic characteristics of single metal atoms and the substrates as input and shows a high accuracy for predicting the catalytic activity of SACs for NRR, which is verified by available experimental data and independent DFT computations. More importantly, a new bidirectional activation mechanism is proposed for thoroughly analyzing the activation of N2 by considering the number of isolated electrons in the d band (N ie-d) of active-metal atoms. Our work not only gives insights on the relationship between the electronic structure of SACs and their catalytic performance but also provides a guided direction for the rational design of SACs.We chose nine 2D materials (C3N4, graphdiyne, C2N, InSe, black phosphorus, BN, MoS2, WSe2, and Mo2C) as the substrates for SACs. As shown in Figure 1 , these nine 2D materials include single atomic layers, multiple atomic layers, single elements, double elements, flat surfaces, furrowed surfaces, and porous structures, which represent almost all types of 2D materials. Moreover, 14 single metal atoms with d electrons (3d: Mn, Fe, Co, Ni, Cu; 4d: Mo, Ru, Rh, Pd, Ag; 5d: W, Ir, Pt, and Au) were selected as active atoms of SACs. The single metal atoms are adsorbed on the energetically favorable sites of different 2D materials, which are determined by their electronic structures and high-symmetry sites. The most stable adsorption configurations are shown in Figure S1 and are consistent with the previous DFT studies. 25–27 The binding energy of single metal atoms could be used for evaluating the thermodynamic stability of SACs, where the strong binding energy indicates high thermodynamic stability. From the viewpoint of the stability of SACs, C2N, C3N4, Mo2C, and InSe have more potential than others as substrates for SACs because of their stronger binding energy values (Figure S2). For example, experimentally, C3N4 has been proved to have high stability for supporting Ru SAC even at high temperatures. 28 Note that although the binding energies of single metal atoms on GDY are medium among all 2D materials, GDY endows a huge potential as the substrate for SACs because of the confinement effect of the uniform distribution of pore configurations, which has been demonstrated by recent theoretical and experimental results. 29–31 On the basis of these 126 SACs, we further explore the activation mechanism of N2 and predict the catalytic activity of NRR through ML.The reaction mechanisms of NRR are very complex and include distal, alternating, enzymatic, and dissociative mechanisms. Except for the dissociative mechanism, the first protonation of N2∗ exists in all mechanisms, which is usually the potential limiting step (PLS) of NRR, especially for SACs. 32 , 33 The dissociative mechanism is almost impossible on SACs given that the single metal atom does not have enough activity to break the strong N≡N triple bond with a bond energy of 9.75 eV. Therefore, the most common distal mechanism on SACs is the only one considered for all systems in this work, as shown in Figure 2 A, which includes six protonations: N2∗ + H+ + e− → NNH∗ (R1) NNH∗ + H+ + e− → NNH2∗ (R2) NNH2∗ + H+ + e− → N∗ + NH3 (R3) N∗ + H+ + e− → NH∗ (R4) NH∗ + H+ + e− → NH2∗ (R5) NH2∗ + H+ + e− → NH3∗ (R6) Among all catalyst systems (126 SACs), only the PLS of nine SACs (W/MoS2, W/WSe2, Au/WSe2, W/BN, Mo/Mo2C, W/Mo2C, Ru/Mo2C, Cu/Mo2C, and Ag/Mo2C) are not the first protonations of N2∗, as shown in Table S1. Note that about half of these nine SACs are W/SACs, and the reaction free-energy values of R1 (ΔG R1) on all W/SACs are very low, as shown in Figure S3 and Table S2, demonstrating that the W single atom has enough activity to activate the N2 molecule, which is mainly due to its outmost 5d orbitals with four isolated electrons, as discussed in detail below. Moreover, more than half of these nine SACs are based on the substrate of Mo2C. As shown in Figure S2, the strong binding energy of the single metal atom (ΔE b-M) on Mo2C indicates the strong interaction between single metal atoms and Mo2C, which could lead to the different reaction mechanisms, where the PLS changes from R1 to other steps, as shown in Table S1. Figure 2B demonstrates the catalytic activity of all SACs with the ΔG PLS for NRR. About 34% of all designed SACs have better catalytic activity than the state-of-the-art Re(111) surface. 34 In addition, considering the 100% utilization rate of active atoms, SACs have been proved to have great potential for NRR. These 14 metals are divided into three regions according to their periods in the periodic table, as shown in Figure 2B. Apparently, in the same period, the ΔG PLS increases from left to right on the periodic table, where the VIB group (Mo and W) SACs show excellent catalytic activity for NRR, whereas the IB group (Cu, Ag, Au) and its nearest VIII group (Ni, Pd, Pt) SACs are almost inert for activating N2 molecule. This interesting phenomenon can be illustrated by the following activation mechanism.The activation process of N2 includes two continuous steps, as shown in Figure 3 A: the lone-pair electrons of N2 in 2σ bonding orbitals first transfer to the d orbitals of active-metal atoms, and then the d electrons of active-metal atoms feed back to the 1π∗ antibonding orbital of N2. However, researchers tend to focus on electron transfer from the d band of metal atoms to the antibonding orbital of N2 while ignoring electron transfer from the lone-pair electrons of N2 to the d orbitals of metal atoms. 35 In fact, the initial electron transfer is more important because this process is also an activation process with reduced bond order due to the electron transfer from the bonding orbitals of N2 to active-metal atoms. Subsequently, the d electrons of active-metal atoms feed back to the antibonding orbital of N2, which further reduces the bond order of N2. As shown in the molecule orbitals of N2 (Figure 3A), before activation, the bond order of N2 is as large as 3, indicating a stable N2 molecule. After activation, the bond order of N2 reduces to 3 − (a + b)/2, where a and b represent the number of electrons transferred in both directions. We define this bidirectional electron transfer for activating molecules as the “bidirectional activation mechanism.”The nature of the bidirectional activation mechanism indicates that the optimal catalyst should accept electrons from the bonding orbitals and provide electrons to antibonding orbitals of pre-activated molecules. Therefore, a new descriptor of isolated electron number in d orbitals (N ie-d) for metal catalysts is proposed for evaluating bidirectional activation mechanism. Figure 3B describes the relationship between the catalytic activity and N ie-d in SACs on C2N; this substrate has strong binding energies for SACs and uniform distribution of pore configurations. In the same period, the larger N ie-d shows a smaller ΔG PLS, indicating a better catalytic activity for NRR. This trend can be understood by the electron configuration rule where the d orbitals with isolated electrons could gain electrons and lose electrons, while the d orbitals with an electron pair could not gain electrons and the empty d orbitals could not lose electrons. The electron configurations of Mo and W are shown in Figure 3C. Mo and W SACs with high catalytic activity have high N ie-d values of 5 and 4, respectively, indicating a strong ability for both gaining and losing electrons. Because of the high N ie-d of Fe (4) and Mo (5) based on the bidirectional activation mechanism, we are able to better understand the high catalytic activity of Fe- and Mo-based catalysts in experiments, which endow high faradic efficiencies and high yield rates for NRR. 36–38 Note that the bidirectional activation mechanism is directly related to the catalytic activity rather than the adsorption energy. More importantly, this proposed mechanism can be applied not only to the activation of nitrogen but also to the activation of other molecules with large bond order, such as O2, CO, and CO2. When these molecules (O2, CO, and CO2) are absorbed on active atoms of catalysts, the first step is always electron transfer from the bonding orbitals of molecules to the active atoms and then electron transfer from the active atoms to the antibonding orbitals of molecules. This means that the bidirectional activation mechanism in this work can be extended to other catalytic reactions, such as the oxygen reduction reaction, CO2 reduction reaction, and CO reduction reaction, which is worthy of further study in the field of catalysis.To further elucidate the bidirectional activation mechanism, we investigated the local and partial density of state (LDOS and PDOS, respectively), as demonstrated in Figure 4 . The sixth-period elements with the best catalytic activity of the W element are used as examples for analyzing their electronic structure characteristics on the substrate of C2N. As shown in Figure 4A, the d band centers of metal atoms are distant from the Fermi level from W to Au in the same period, indicating that the d electrons of W are more active and thus transfer to the antibonding orbitals of N2 more readily than others. Moreover, there are more vacant d orbitals in the W/C2N system. These half-filled d orbitals could give more room for the lone-pair electrons of N2, which is consistent with our previous analysis.The PDOS of W/C2N, N2 adsorbed W/C2N (N2-W/C2N), and N2 molecule are depicted in Figure 4B. A strong hybridization between W-dxy, W-dz2 orbitals, and the bonding orbitals of N2-σ is found, where the electrons transfer to W-d orbitals from the N2-σ orbitals through W-dxy and W-dz2 orbitals. This means that the number of electrons in the bonding orbitals of N2 decreases, indicating the reduced bond order of N2. Moreover, all the W-d orbitals have hybridization with antibonding orbitals of N2-π∗, leading the N2-π∗ orbitals to be broadened and half filled by the electrons from the W-d orbitals. In particular, there is a strong hybridization between W-dyz, W-dxz, and N2-π∗ orbitals, denoting that the transfer of electrons from W-d orbitals to N2-π∗ orbitals is mainly through W-dyz and W-dxz orbitals. The electrons occupied in N2-π∗ orbitals further reduce the bond order of N2. Therefore, both the loss of electrons in bonding orbitals and the acquisition of electrons in antibonding orbitals could achieve activation of the N2 molecule, indicating the bidirectional activation mechanism. This insight into the catalytic nature of N2 on SACs plays a decisive role in the rational design of SACs for NRR.To quantify the effect of the bidirectional activation mechanism on SAC’s catalytic properties, here we leverage the predicting power of ML models trained on DFT-calculated data. In addition, ML models can address the challenge presented by the vast chemical space of 2D material-supported SAC given that it is time consuming to exhaustively characterize all possible materials either theoretically or experimentally.ML models are developed to simultaneously predict three crucial physical quantities to NRR performance: ΔG PLS for NRR, the adsorption free energy of hydrogen (ΔG H∗) for the hydrogen evolution reaction (HER), and ΔE b-M. ΔG H∗ is an important factor in determining NRR selectivity because HER is always a competing side reaction in NRR. 39 Moreover, ΔE b-M is directly related to the stability of the SAC, a high value of which is desired to resist sintering. 40 This is the first time that the outputs include the activity, selectivity, and stability of SACs for NRR rather than one aspect of NRR performance.The ML prediction is made on the basis of a total of seven input features, three intrinsic features of which characterize the 2D materials (Fermi energy, electrostatic potential, and the work function of pure 2D materials) and four intrinsic features of which characterize the single atoms (electronegativity, electronic affinity, ionization energy, and N ie-d of the single atom). Note that all features are only related to the electronic structures rather than complex coordination environments. Moreover, these seven input features can be obtained directly from some databases and even the periodic table. The electronic properties of each SAC are not required in our ML models, and their acquisition is quite time consuming and laborious. Therefore, the simple input features are a huge advantage of our ML models.The developed ML models are based on the boosted-regression-tree ensemble method, which utilizes an ensemble of regression trees such that each regression tree makes its own prediction and the final prediction is the collective prediction over all individual regression trees. One ML model is built for each output variable (ΔG PLS, ΔG H∗, and ΔE b- M), and three models are built as a result. All models take the same input features and are subject to the same hyperparameters and evaluation criteria.The developed ML models have high accuracy as demonstrated by their parity plots shown in Figures 5 A and S5. The predicted values match those obtained from DFT calculations both in the training set and in the testing set for all three quantities predicted. The developed ML models show a testing-set mean absolute error (MAE) of 0.21, 0.24, and 0.43 eV and a root-mean-square error (RMSE) of 0.28, 0.31, and 0.52 eV for ΔG PLS, ΔG H∗, and ΔE b- M, respectively, with a testing-set percentage of 20%. These ML performance metrics are reasonably high given the moderate dataset size, drastic value ranges of the outputs, and the complexity involved in the electronic structure of a metal-substrate combination. The models are only moderately overfit, as indicated by the slightly lower MAE values on the training sets corresponding to each of the three output variables. To compare more directly with the experimental data, our ML model also predicted an overpotential (η) for NRR, which is a direct indicator of catalytic reactivity in experiments, and a smaller η value represents a faster NRR. 41 The ML model still has high accuracy with an MAE of 0.21 V and an RMSE of 0.28 V, as shown in Figure S4. More importantly, the ML model also possesses high generalizability given that the ML-predicted η values match those obtained from the previous theoretical and experimental works with reasonable accuracy (Figure 5B). 19 , 27 , 42–44 This implies that the ML models are generalizable to a broad range of literature. For instance, according to the ML models, the predicted η value of Ru/N4-C is 0.55 V versus a reversible hydrogen electrode (RHE), indicating an excellent catalytic performance for NRR, which is consistent with the experimental work of Geng et al., where Ru/N4-C achieved a high yield rate of 120.9 μgNH3 mg−1 cat. h−1 for NRR. 44 To understand how important each input feature is to the output prediction, we analyzed feature importance on the prediction of the catalytic performance of NRR and HER. The relative importance of input features is shown in Figure 5C (for HER) and in Figure 5D (for NRR). Interestingly, the overall feature importance is different for HER and NRR such that HER is mostly influenced by the Fermi energy of the 2D material and the electronegativity and electronic affinity of the single metal atom and NRR is mostly influenced by N ie-d. The high feature importance in N ie-d for NRR, as opposed to HER, is in excellent agreement with the important role that directional activation mechanism plays in activating nitrogen molecules, as highlighted in the earlier discussion. Since the adsorption of H does not involve bidirectional electron transfer, the new descriptor of N ie-d shows low feature importance in HER. Meanwhile, the different feature importance is conducive to regulating the catalytic activity of NRR and HER from different features. This means that it is easier to design SACs with high catalytic activity of NRR but low catalytic activity of HER, which is crucial for the selectivity during the NRR process.The accuracy and generalizability of the developed ML models could potentially act as a tool for mass-scale screening of combinations of 2D materials and single metal atoms. These ML models highlight the holistic approach toward rational SAC design by their ability to simultaneously predict NRR activity, selectivity, and SAC stability. It is important to note, on the other hand, that because of the nature of tree-based ML models, it might be difficult to accurately extrapolate our ML model for new types of 2D materials or new types of single metal atoms. However, for a new type of 2D material, the developed model requires the calculation of only a few single-atom metal types to predict for all other single-atom metal types, which was demonstrated in the randomly shuffled training-testing experiment above. This can still provide much time savings in comparison with calculating all combinations of 2D materials and single atoms via DFT.In conclusion, we investigated the catalytic performance of a series of SACs for electrocatalytic NRR on the basis of DFT calculations and developed a ML model to represent the structure-activity relations. Based on the bidirectional activation mechanism proposed, a new descriptor of N ie-d is closely related to the catalytic activity for NRR. The large N ie-d indicates high catalytic activity, which is further revealed by electronic structure analysis, where the electrons in σ orbitals of N2 transfer to the d orbitals of active atoms through dxy and dxz orbitals and the electrons go back to π∗ of N2 mainly by dyz and dz2 orbitals, resulting in the activation of N2. The insights achieved in this work on the bidirectional activation mechanism will further promote the rapid development of metal catalysts. More importantly, we can predict the catalytic activity of SACs for NRR through ML, which is further validated by both DFT calculations and experimental works. We hope that our strategy and the ML model can be extended to other catalytic systems.Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Chandra Veer Singh ([email protected]).All unique reagents generated in this study will be made available upon request.There is no dataset or code associated with this paper.All computations were performed with the spin-polarized DFT calculations using the Vienna Ab initio Simulation Package. 45 The Projector-augmented wave pseudopotential and Perdew-Burke-Ernzerhof functional of the generalized gradient approximation were utilized to describe the interactions between valence electrons and ionic cores, as well as the exchange-correlation effects. 46 , 47 The kinetic energy cutoff for the wave-function calculations was set to 550 eV. A smearing width of 0.1 eV was applied for the Fermi smearing function. The supercells of all systems were calculated with a 4 × 4 ×1 Monkhorst-Pack grid of k-points, and a vacuum gap of about 15 Å was used to avoid interactions between the system and its mirror images. The Tkatchenko-Scheffler method was applied to describe the van der Waals interactions, 48 which has been successfully used in 2D materials such as boron nitride, 49 MoS2, 50 and boron monolayer. 51 The geometric relaxation was stopped when the incremental changes in total energy and forces were smaller than 1 × 10−4 eV and 0.02 eV/Å, respectively.To evaluate the electrochemical catalytic reactions with the transfer of proton-electron pairs, we used the computational hydrogen electrode model to obtain the free energies, 34 , 52 where we set the reference potential to be RHE and determined the chemical potential of the proton-electron pair by one-half of the chemical potential of H2. Reaction free energy (ΔG) was determined by (Equation 1) ΔG = ΔE + ΔZPE – TΔS, where ΔE, ΔZPE, T, and ΔS denoted the reaction energy, zero-point energy change, temperature, and entropy change, respectively. The overpotential (η) was a good indicator of catalytic activity, where a smaller η value indicated a better catalytic activity for NRR. The η value was determined by, (Equation 2) η = U equilibrium − U limiting where U equilibrium and U limiting denoted the equilibrium potential (U equilibrium = −0.16 V versus RHE for N2 + 6 H+ + 6e − → 2NH3) and the applied potential required for PLS (U limiting = −ΔG/e). We determined the stability of SACs by computing the binding energy of metal atoms on 2D materials (ΔE b) as shown below, (Equation 3) ΔE b = E SACs – E 2D – E M where E SACs, E 2D, and E M are the total energies of the SACs, the substrate of 2D materials, and metal atoms, respectively.The ML models were developed on the basis of the boosted-regression-tree ensemble method, which utilized a collection of individual regression trees. Each regression tree made its own prediction, and the final prediction was the collective prediction over all decision trees. The tree-based method was chosen primarily because each 2D material possessed distinct characteristics, making the tree-based method suitable. The algorithm was implemented in MATLAB with its built-in function, “fitrensemble.” The boosting algorithm was chosen to be least-squares boosting (“LSBoost”). All data were first shuffled randomly, and training and testing data were then split into a 80%/20% ratio. Based on Bayesian hyperparameter optimization from preliminary investigations, a learning rate of 0.05, number of learning cycles of 180, minimum leaf size of 3, and number of sampled variables of 10 were used as the hyperparameters throughout training. To analyze the feature importance, we used MATLAB’s built-in function “predictorImportance,” which determined the importance of each feature by summing feature importance values from all decision trees in the ensemble.We acknowledge financial support from the Natural Sciences and Engineering Research Council of Canada, the Hart Professorship, and the University of Toronto. We also acknowledge Compute Canada for providing computing resources at the SciNet, CalculQuebec and the Westgrid consortia.Z.W.C., Z.L., and L.X.C. contributed equally to this work. C.V.S. and Z.W.C. conceived and designed the study. Z.W.C., Z.L., and L.X.C. carried out the DFT simulations and machine-learning model development. Z.W.C., Z.L., L.X.C., and C.V.S. wrote the paper. All authors discussed and revised the manuscript.The authors declare no competing financial interests.Supplemental information can be found online at https://doi.org/10.1016/j.checat.2021.03.003. Document S1. Figures S1–S5 and Tables S1 and S2 Document S2. Article plus supplemental information

Single-atom catalysts (SACs) have provided new impetus to the field of catalysis because of their high activity, high selectivity, and theoretically full utilization of active atoms. However, the ambiguous activation mechanism prevents a clear understanding of the structure-activity relationship and results in a great challenge of rational design of SACs. Herein, by combining density functional theory (DFT) calculations with machine learning (ML), we explore 126 SACs to analyze and develop the structure-activity relationship for the electrocatalytic nitrogen reduction reaction (NRR). We first propose a bidirectional activation mechanism with a new descriptor for catalytic activity, which provides new insights for the rational design of SACs. More importantly, we establish a ML model for predicting the catalytic performance of NRR, validated by both DFT calculations and experimental works. The successful ML prediction in this work helps with the accelerated design and discovery of new catalysts by computational screening with high practical significance.

Substituting the traditional fossil fuels by high-energy-density hydrogen with zero-environmental impact promises a carbon-free era in fuel utilization [1–3]. Water electrolysis driven by renewable energy is an effective way for hydrogen production with little environmental impact [4,5], and therefore has drawn increasing attention in a world-wide scale [6–8]. Exploring non-noble electrocatalysts with high activity, long-term stability and easy deployment is of importance for practical applications.Many non-precious electrocatalysts in a powdery state have been designed [8–10]. The indispensable inert organic binders used to paste the powdery catalysts onto the conductive substrates tend to block active sites [11]. In addition, powdery catalyst film is prone to peel off from the substrate during long-term electrolysis, especially at large current densities in practical applications [12]. Moreover, fleeing-away of hydrogen bubbles from the organic binder-containing surfaces is impeded due to the deterioration of wettability [11,12], resulting in increased overpotential and energy consumption.Some non-precious metal catalysts such as Ni, Cu, Mo, and their alloys show appreciable catalytic activity for HER [3,5,8,13–15]. Importantly, these metals with high ductility can be easily processed and manufactured into various shapes according to the practical requirement, serving as both substrates for conducting current and catalysts for boosting HER [12,15]. Therefore, the above drawbacks of powdery catalysts can be avoided, enabling long-term stability and large-current-density electrolysis. For example, Ni is the most common HER catalysts for industrial water electrolysis in alkaline solutions [15,16]. However, the activity of these metal catalysts needs to be essentially improved.Constructing interfaces between metal catalysts and oxide promoters is an effective strategy to improve the activity of metal catalysts [13,15–18]. For example, addition of alkaline oxides into Fe-based catalysts can promote the Fischer-Tropsch synthesis reactions [18]. Enhancements of other catalytic reactions including CO oxidation, CO2 hydrogenation, and methane reforming are also reported by delicately designed metal/oxide interfaces [15,16,19,20]. Recently, HER is also greatly boosted by a metal/oxide interface-induced synergistic effect according to a “chimney effect” [14]. Improving the activity of metal catalysts for HER by constructing metal/oxide interfaces is an appealing strategy.Herein, Ni-V2O3 interfaces are one-pot constructed by electrochemical reduction of NaVO3 on porous nickel cathode in NaCl molten salt. The high-temperature molten salt enables the tightly adherence of electrodeposited V2O3 on porous nickel. The high solubility of NaVO3 in molten salt contributes to an ultra-fast homogeneous deposition of V2O3, contributing to binder-free construction of V2O3 coatings on nickel substrates within only several minutes. Specially, the nanostructured strip-like V2O3 is perpendicularly anchored on the surfaces of nickel substrate, promising abundant Ni-V2O3 interfaces and fully surface exposure of both nickel and V2O3. The V2O3-modified nickel results in interface-induced synergy between Ni and V2O3, contributing to enhanced activation of H2O and improved reduction of H* due to efficient electron transfer between substrate and reactive intermediates. Resultantly, V2O3-modified nickel shows much improved activity toward HER in alkaline solution when compared with the bare nickel counterpart.All chemical reagents were of analytical purity and used as received without further purification. NaCl (anhydrous, 99% purity) was provided by Shanghai Titan Scientific Co., Ltd. NaVO3 (anhydrous, 99.9% purity) was purchased from the Aladdin (Shanghai, China). Porous nickel (99.9% purity) was provided by Dongguan Zehui New Material Technology Co., LTD. Before electrolysis, NaCl was well mixed with 5 wt% NaVO3 and contained in an alumina crucible (inner diameter, 70 mm; height, 150 mm). The alumina containing the salts was then transferred into an alumina tube reactor, which was continuously flushed by an argon flow (150 mL min−1) to keep an inert atmosphere. The physically adsorbed water by the salts was removed by keeping the temperature at 400 °C for 12 h (h). Then the temperature is increased to 850 °C and maintained for half an hour to enable the complete melting of the salts. After that, a graphite anode (15 mm in diameter and 99.9% in purity) and porous nickel cathode were immersed into the molten salt and the electrolysis was then initiated. All the electrolysis temperature was kept at 850 °C, with the cell voltage being 2.5 or 2.8 V. The electrolysis time was varied from 1 to 60 min. After electrolysis, the cooled nickel cathode was immersed in de-ionized water for more than 12 h to remove the entrained salts and then dried at 60 °C in air overnight.X-ray diffraction (XRD) spectra were acquired on Rigaku Miniflex600 at a scan rate of 4° min−1 with Ni filtered Cu K α radiation (λ = 1.5406 Å). The field-emission scanning electron microscope (FESEM, Zeiss SIGMA) and transmission electron microscopy (TEM, Titan G 2 60–300) were used to probe the morphology of prepared samples. The composition and elemental mapping images of the samples were analyzed by energy dispersive X-ray spectroscopy (EDS, GENESIS 7000 and OXFORD IET 200) attached to TEM apparatus. X-ray photoelectron spectra (XPS) were collected on X-ray photoelectron spectrometer (ESCALAB250Xi, Thermo Fisher Scientific). All the XPS results were calibrated by C 1 s at 284.8 eV.All the electrochemical texts were conducted in a 1 M KOH aqueous solution, using a standard three-electrode configurations programmed by a CHI 760 electrochemical workstation. Before test, the newly prepared solution was continuously bubbled for 1 h by a N2 flow at 100 mL min−1. A carbon rod (10 mm in diameter and 99.999% in purity) and a home-made reversible hydrogen electrode (RHE) were used as the counter electrode and reference electrode, respectively [11]. The porous nickel deposited with V2O3 (Ni-V2O3) was cut into a square piece (1 cm × 1 cm) and directly used as the working electrode, without addition of any organic binders or conducting agents. For comparison, performance of bare porous nickel (Ni) with the same dimension was also tested. In another case, a part of V2O3 powders were scraped off from the porous nickel substrate by sonication in de-ionized water at 60 °C for more than 12 h. After centrifugation and being dried at 60 °C overnight, solid V2O3 powder was obtained, which was then made into ink by addition of 2 mg as-prepared V2O3 powder into 400 μL ethanol and sonication for 1 h. The V2O3 working electrode (denoted as V2O3) was prepared by dropping 17.5 μL ink onto the surface of a glassy carbon rotating disk electrode (RDE, Pine Research Instrumentation, 5 mm in diameter), with further addition of ~2 μL Dupont Nafion 117 solution (5 wt%). The Pt/C working electrode was prepared by the same way, replacing the catalysts with the Pt/C powder (20 wt% Pt, Sigma-Aldrich) instead. The mass loading for Pt/C and V2O3 sample are both 0.5 mg cm−2, which are the same with that in Ni-V2O3 obtained by electrolysis at 2.8 V for 5 min. For all the test of the V2O3 and Pt/C working electrodes, the rotating speed was fixed at 1600 rounds per minute. All the linear scanning voltammogram (LSV) curves were recorded at 5 mV s− 1. All polarization curves were iR-corrected unless noted. The electrochemical impedance spectra (EIS) were acquired in the frequency range of 0.01–105 Hz, with the amplitude and overpotential being fixed at 0.01 V and 230 mV (Solartron 1470E).The density-functional-theory (DFT) calculations were performed by using Vienna Abinitio Simulation Package (VASP) with Projector Augmented Wave (PAW) method [21–24]. Perdew-Burke-Ernzerhof (PBE) functional for the exchange-correlation term was used with the projector augmented wave method [25,26], with the kinetic energy cutoff of electron wave functions being set as 400 eV. The convergence of energy and forces were set to be 1 × 10−4 eV and 0.05 eV Å−1, respectively [27–30]. The crystal surfaces with the highest XRD peak intensities of Ni and V2O3, namely Ni(111) and V2O3(104), were used to construct the geometry structures [31].The solubility of NaVO3 in NaCl molten salt is high up to 8 wt% [32,33], which means that the NaVO3 and NaCl can form a homogenous system in a molten state at the electrolysis temperature [34–36]. When the cell voltage is applied, the soluble VO3 − on the surface of porous nickel cathode can be easily electroreduced to V2O3 (2VO3 2− + 4e− =V2O3 + 3O2−) [7,32,33], which can be in situ adhered onto the porous nickel to form V2O3-coated nickel surfaces (denoted as Ni-V2O3), as illustrated in Fig. 1 .Deposition of V2O3 on the porous nickel cathode is strongly validated by the XRD and XPS results after electrolysis (Ni-after, Fig. 2 a–c). In addition to Ni (JCPDS No. 01–1258), extra diffraction peaks relating to V2O3 (JCPDS No. 84–0318) are observed in the XRD pattern of the sample Ni-after (Fig. 2a). The appearance of V 2p signals in the XPS survey spectrum of the sample Ni-after (Fig. 2b) also reveals the existence of vanadium element on the surface of porous nickel, which is further manifested by the comparison of high-resolution V 2p spectra between the samples of Ni-before and Ni-after.The morphology characterizations further uncover the adherence of nanostructured V2O3 on the surface of porous nickel cathode. After deposition of V2O3, the smooth surface of bare porous nickel (Fig. 2d) becomes furry (Fig. 2e), with deposition of closely anchored strip-like V2O3 nanoarrays (Fig. 2f). The length and width of single V2O3 strips are ~1 μm and 100–200 nm, respectively (Fig. 2g), with the thickness being several nanometers (Fig. 2f). Specially, the V2O3 strips are perpendicularly attached to the surface of nickel (Fig. 2f), which is further verified by the TEM images in Fig. 2(g). Such a configuration contains Ni/V2O3 interfaces and maximize the surface exposure of both Ni and V2O3. The Ni/V2O3 interface is further manifested by the high-resolution TEM image in Fig. 2(h), as evidenced by the clear boundary between (200) plane of Ni and (104) plane of V2O3. Fig. 2(i) shows the high-resolution TEM image of region 3 in Fig. 2(g), in which the (104) plane of V2O3 is also observed. The results also reveal that the strips are V2O3, as further manifested by the homogenous distribution of V and O in one detached strip (Fig. 2j–m).Ni-V2O3 shows much more superior HER performance in alkaline solution than the Ni and V2O3 counterparts. The lowest onset potential (20 mV) occurs in Ni-V2O3 among the samples of Ni-V2O3, Ni, and V2O3 (Fig. 3 a), which is further validated by the overpotential comparison at 10 mA cm−2 (η@10 mA cm−2) for various samples (Fig. 3b). The optimal electrolysis time for electrodeposition of V2O3 is 5 min, as evidenced by both the η@10 mA cm−2 (Fig. 3b) and LSV curves (Fig. S1). Too short electrolysis time can hardly provide enough V2O3 strips to construct abundant Ni/V2O3 interfaces (Fig. S2) while too long electrolysis time leads to overgrowth of V2O3 for blocking the surfaces of nickel (Fig. S3). Electrolysis of 5 min. can realize the balance between the as-mentioned two aspects. A η@10 mA cm−2 value as low as 136 mV (Fig. 3b) hence appears in the optimal sample, much smaller than that of Ni (256 mV, Fig. 3b) and V2O3 (632 mV, Fig. 3b). Of note, this strategy with such a short time (5 min) and extremely easy post-processing of samples (just leaching in de-ionized water) for constructing the well-defined metal/oxides interfaces is much more convenient when compared with previously reported methods [15–17]. Compared with electrolysis time, the influence of electrolysis voltage on the performance of Ni-V2O3 samples is minor (Fig. S4).The activity enhancement of porous Ni by deposited V2O3 is further manifested by the comparisons of Tafel slopes (Fig. 3c), electrochemical surface areas (Fig. 3d), and EIS spectra (Fig. 3e). Compared with Ni and V2O3, Ni-V2O3 presents the lowest Tafel slop (125 mV dec−1), meaning much improved reaction kinetics for HER after deposition of V2O3 on nickel. By extrapolation of the Tafel curves, the exchange current density (j 0) can be obtained, which is an indicator for the intrinsic activity of electrocatalysts. As shown in Fig. S5, deposition of V2O3 on porous nickel increases the exchange current density from 0.34 mA cm−2 (Ni) to 0.53 mA cm−2 (Ni-V2O3), which are also higher than that of V2O3 (0.04 mA cm−2). According to the CV curves at different scan rates recorded in the non-faradic regions (Fig. S6), the electrochemical surface area (ECSA) is measured based on the double-layer capacitance (C dl). The C dl of Ni-V2O3 (0.95 mF cm−2) is much higher than that of Ni (0.21 mF cm−2) and V2O3 (0.38 mF cm−2), even close to that of Pt/C (1.09 mF cm−2). The results also reveal that deposition of V2O3 on nickel can increase the reactivity sites for HER. The electrochemical impedance spectra (EIS) reveal the lowest charge transfer resistance of Ni-V2O3 among the three tested samples (Fig. 3e), indicating the constructed Ni/V2O3 interfaces can facilitate the electrons transfer between substrate and reactive intermediates. Such a synergy also endows the Ni-V2O3 electrode a good long-term durability, as evidenced by very small overpotential increases after electrolysis for 20 h at various current densities (Fig. 3f). The performance of Ni-V2O3 is comparable/superior to reported Ni-based electrocatalysts or other HER catalysts (Table S1). Surface modification of Ni substrate by forming a phosphide or carbide layer is a potential route for further enhancing the performance [3,5,8,13–15].The synergy between Ni and V2O3 induces enhanced electron transfer between the Ni/V2O3 interfaces. Fig. 4 (a) shows the XPS spectra of Ni 2p before and after deposition of V2O3. Shift of Ni0 peak toward a higher binding energy is observed in the XPS spectra after deposition of V2O3 on Ni surfaces, revealing a tendency for electron transfer from Ni to V2O3 [13]. Such a phenomenon is also observed by other researchers [15,16]. Ni2+ in Fig. 4(a) should be attributed to the surface oxidation of Ni substrate [15,16]. Compared with bare Ni, the increase in the peak intensity of Ni2+ in Ni-V2O3 indicates the tendency for the electron transfer from Ni surface to V2O3. Such results are in concert with the decreased peak intensity of Ni0 after decorating V2O3 on Ni, further validating the electron transfer from Ni to V2O3. Of note, over-oxidation of Ni surface can decrease the surface conductivity of Ni substrate, which is one reason for limiting the further enhancement of HER performance. The accumulation of electrons around V2O3 in the Ni/V2O3 interfaces is further uncovered by the local charge density difference (Δρ, Fig. 4b). From a closer observation of O and Ni atoms at positions close to the interfaces, positive Δρ near O (denoted as ONearby in Fig. 4b) is found, indicating the increase of electron density. The increase of electron density in ONearby is in concert with the decrease of electron density in NiNearby, as evidenced by the negative Δρ value of Ni atoms at the interfaces [14,38]. Such results mean that electrons are transferred from Ni atoms of nickel sheet to O atoms of V2O3 in the Ni/V2O3 interfaces, being consistent with the XPS results [38].The induced electron transfer from Ni to surface V2O3 promotes the activation of H2O and combination of H* to generate H2. Two steps are involved for hydrogen evolution reaction (HER) in alkaline solution, namely prior water dissociation to form H* intermediates (Volmer step) and subsequent combination of H* to form H2 (Tafel step or Heyrovsky step) [17,39–41]. The activated water adsorption energy (ΔG H2O) is applied as the activity descriptor for the former step while the binding free energy of H* (ΔG H*) is used as the activity descriptor for the latter step [17,42,43]. Details for calculating the ΔG H2O and ΔG H ⁎ are provided in the supporting information (see Text 1), with the corresponding results being presented in Fig. 4(c) and optimized structures being shown in Figs. S7 and S8. The construction of Ni/V2O3 interfaces decreases the value of ΔG H ⁎ from −0.549 eV for Ni to −0.092 eV for Ni-V2O3 (Fig. 4c), approaching to the optimal value (ΔG H*=0) [17]. Therefore, the Ni/V2O3 interfaces promote the H2 generation. More specifically, accumulation of electrons around the interface-positioned O atoms stimulates the combination of H* toward generation of H2. Although ΔG H ⁎ (−0.143 eV, Fig. 4c) for single V2O3 is also close to the optimal value (0 eV), the high ΔG H2O (0.269 eV) for single V2O3 retards the dissociation of water to generate H*, leading to sluggish HER kinetics. The same case is observed for single Ni, as evidenced a large value of ΔG H2O for Ni (0.525 eV, Fig. 4c). Interestingly, the lowest ΔG H2O value (0.073 eV) is observed for Ni-V2O3, implying that the V2O3 on the surfaces of nickel accelerates the cleavage of HOH [17]. Hence, V2O3 strips attached onto the Ni surface can act as a water dissociation promoter to generate hydrogen intermediates, which are then combined to generate H2 on the O sites of V2O3.The deposited V2O3 strips with cuspidal ends are perpendicularly anchored on the surface of nickel, creating abundant Ni/V2O3 interfaces (as illustrated in Fig. 5 a). The V2O3 strips attached onto the surface of Ni promote the dissociation of H2O to generate hydrogen intermediates (denoted as step ① in Fig. 5b). The Ni substrate with high conductivity swiftly guides the electrons from external circuit to the surfaces (denoted as step ② in Fig. 5b). Importantly, the synergy between Ni and V2O3 pumps the electrons on surfaces of Ni to surroundings of O atoms in V2O3, promoting the combination of H* for generating H2 (denoted as ③ in Fig. 5b). Therefore, the high-efficiency construction of Ni/V2O3 interfaces by molten salt electrodeposition of V2O3 from highly soluble NaVO3 provides a facile and swift way for improving the activity of Ni toward HER. Other metallic substrates can be also used for fabricating free-standing metal-oxide electrodes by this method. When a carbon cloth is used as the substrate, formation of free-standing carbides is highly possible during molten salt electrolysis [33,36,37]. Of note, NaVO3 is the primary intermediate product during processing of vanadium slag, which is a typical industrial waste from metallurgy of vanadium-containing ores [33,36,37]. Therefore, the strategy provided herein can hopefully be also applied for the deep processing of vanadium slags.V2O3-modified Ni with superior HER activity in alkaline solution is fabricated by molten salt electrodeposition of V2O3 on porous nickel. The high solubility of NaVO3 in molten salts and high-temperature medium enables the swift deposition and tightly adhesion of V2O3 strips on Ni, creating abundant Ni-V2O3 interfaces with a special perpendicular-anchoring configuration. Such Ni-V2O3 interfaces induce a synergy between Ni and V2O3, promoting the activation of H2O and combination of H* to form H2. Resultantly, the V2O3-modified Ni shows a η@10 mA cm−2 value as low as 136 mV and a Tafel slope of only 125 mV dec−1 as well as a long-term stability for HER in alkaline solution. The strategy herein can provide for not only the activity improvement of Ni towards HER but also value-added processing of vanadium slag wastes.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 funding support from the National Natural Science Foundation of China (51722404, 51674177, 51804221 and 91845113), the National Key R&D Program of China (2018YFE0201703), the China Postdoctoral Science Foundation (2018M642906 and 2019T120684), the Fundamental Research Funds for the Central Universities (2042017kf0200), and the Hubei Provincial Natural Science Foundation of China (2019CFA065).Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jechem.2020.03.048. Image, application 1

Implementation of non-precious electrocatalysts is key-enabling for water electrolysis to relieve challenges in energy and environmental sustainability. Self-supporting Ni-V2O3 electrodes consisting of nanostrip-like V2O3 perpendicularly anchored on Ni meshes are herein constructed via the electrochemical reduction of soluble NaVO3 in molten salts for enhanced electrocatalytic hydrogen evolution. Such a special configuration in morphology and composition creates a well confined interface between Ni and V2O3. Experimental and Density-Functional-Theory results confirm that the synergy between Ni and V2O3 accelerates the dissociation of H2O for forming hydrogen intermediates and enhances the combination of H* for generating H2.

Intense research on the water-oxidation catalyst (WOC) center in photosystem II (PSII) over the last decades has revealed deep insights on the mechanisms by which nature liberates electrons and protons from H2O, two critical ingredients for downstream reactions such as CO2 reduction and N2 fixation. 1 , 2 This knowledge has propelled research on using molecular catalysts to oxidize water, and impressive progress has been made in terms of catalyst performance as measured by turn-over frequencies (TOFs) and turn-over numbers. 3 , 4 From a technological development perspective, there is a strong incentive to perform the reaction on heterogeneous catalysts, especially on those of low-cost and outstanding stability. Indeed, recent years have witnessed a surge of such research activities. 5–11 Despite the apparent variety of these catalysts, they share important commonalities in the chemical mechanisms. For instance, it is generally believed that the reaction proceeds through a series of proton-coupled electron transfer steps that lead to the formation of M=O (where M represents an active metal center) intermediates. 12 , 13 It is also agreed upon that the subsequent O–O bond formation is of critical importance to the overall reaction. 14 However, the details of the O–O formation and the subsequent steps have been the subject of diverging views. At least two possible pathways have been proposed and supported. 15–18 One involves direct nucleophilic attack of water, followed by O2 release and regeneration of the catalyst. In the literature, this mechanism is referred to as water nucleophilic attack (WNA) (Figure 1 , right pathway). 4 , 15 The other involves the coupling of two metal-oxo intermediates followed by O2 release, which is referred to as intramolecular oxygen coupling (IMOC) (Figure 1, left pathway). 15 For Ir- and Ru-based molecular catalysts, density-functional theory (DFT) calculations predicted that the IMOC pathway dominates at low overpotentials, whereas the WNA pathway becomes accessible at higher overpotentials. 17 , 19 The two pathways were also predicted to be competitive on a heterogenized dinuclear Ir oxide cluster. 17 With optical pump-probe spectroscopy, Cuk et al. 18 monitored the microsecond decay of oxyl (Ti–O⋅) and bridge (Ti–O⋅–Ti) intermediates on SrTiO3 photoelectrodes. They found that the two species decay with distinct reaction rates on a microsecond timescale. It was suggested that Ti–O⋅’s convert to Ti–O–O–Ti by dimerization (IMOC pathway) and Ti–O⋅–Ti converts to Ti–OOH by nucleophilic attack of water (WNA pathway). Furthermore, it was found that the relative predominance of the two pathways was controlled by the ionic strength of the electrolyte, with the WNA pathway dominating at low ionic strength. However, how the relative predominance of these mechanisms depends on the applied electrode potential has not been investigated in experiments. Herein, we address this central question.Inspiration on how to further this understanding could be drawn from progress made in molecular WOC-based studies. To discern different pathways for the water-oxidation reaction by molecular catalysts, researchers have resorted to a strategy of correlating the reaction rate with the catalyst concentrations. 4 With the help of additional experiments such as isotope labeling, significant knowledge has been gained. 20–22 However, similar approaches are challenging to implement for heterogeneous catalysts, because the active sites, including their structures and densities, are often poorly defined on a heterogeneous catalyst. The challenge could be circumvented using clever experimental designs. For instance, Durrant et al. 23 have identified a change of reaction orders relative to the hole concentration from the first to the third order on Fe2O3 using photoinduced absorption spectroscopy. Frei et al. 13 have succeeded in observing both the metal-oxo and superoxo species, using an infrared spectroscopy (IR) technique. In both studies, different reaction mechanisms were proposed for different light intensities. Nevertheless, owing to the lack of detailed information on the active centers, particularly their density under different conditions, it remains difficult to directly corroborate these early observations for an unambiguous understanding of water oxidation on heterogeneous catalysts. Although it is possible to address this challenge by synthesizing heterogeneous catalysts with well-defined active centers, as has been demonstrated recently by others and us, 24 , 25 the catalyst library remains limited, and significant work is needed before the values of such catalysts can be materialized. An alternative approach is to study how the reaction kinetics changes as a function of water activity, which is the main strategy for this present work.To appreciate the significance of this strategy, it helps to examine the proposed WNA and IMOC pathways on a heterogeneous Co phosphate (Co–Pi) catalyst (Figure 1). Previous studies have suggested that the initial electron/proton transfer steps (vertical arrow in the center) are fast in comparison with the O–O formation. Therefore, these steps are quasi-equilibrated, whereas O–O formation limits the rate of the reaction. From the oxidized state of the catalyst shown on the bottom of the scheme, the water-oxidation process can proceed through two distinct pathways: the WNA pathway involves a water molecule within the electric double layer in the rate-determining O–O forming step (right arrow). By contrast, the IMOC pathway only involves surface species in the rate-determining step (RDS) (left arrow). On the basis of this simplified mechanistic picture, the water-oxidation reaction is expected to be (pseudo) first order in the water activity when proceeding through the WNA pathway, whereas it is (pseudo) zeroth order when proceeding through the IMOC pathway. This simplified view assumes that the change in the water activity does not significantly affect the positions of the quasi-equilibria before the presumed RDS of O–O bond formation, as discussed later. Therefore, it is possible to discern the reaction mechanisms even without detailed knowledge of the active centers by altering the water activity, which has not been investigated previously.The problem is now reduced to how to alter water activity in a water-oxidation reaction. Indeed, most previous studies on this subject have treated water as a substrate of invariant activity, such that it was excluded in most kinetic considerations. 26 , 27 Only recently did we see advances where the water activity could be suppressed significantly in aqueous solutions. 28–30 The so-called “water-in-salt” electrolyte containing high concentrations of salts (e.g., up to 21 m [mole per kg of H2O]) represents one such system. The strong solvation effect of the high-concentration ions renders its H2O behaviors drastically different from those in bulk H2O. It becomes possible to perform water-oxidation reactions in an aqueous system where the water activity is no longer unity. Therefore, we are offered an opportunity to test the hypothesis proposed in the previous paragraph. That is, we expect a different sensitivity of the kinetics on the water activity for different mechanisms.To prove this concept, we have chosen Co-oxide-based catalysts as a study platform because they represent a class of most studied heterogeneous WOCs, with Co–Pi receiving arguably the most attention. A broad knowledge base has already been generated. 15 , 27 , 31–36 For example, the coordination environment of Co has been identified by a suite of spectroscopic techniques. 34 That the O–O formation is the RDS has been supported by numerous studies. 15 , 27 , 31 , 32 , 35 , 36 Both WNA and IMOC mechanisms have been proposed and supported by either experimental or computational studies for this catalyst. 15 , 32 , 35–39 Herein, we report the new observation of a switch from the IMOC pathway at low applied potentials to the WNA mechanism at high applied potentials.Previous studies have shown that various implementations of infrared and surface-enhanced Raman spectroscopies are powerful probes of water-oxidation intermediates. 12 , 13 , 18 , 40–48 To examine the mechanistic proposal (Figure 1), we employed surface-enhanced infrared absorption spectroscopy (SEIRAS) in the attenuated total reflection (ATR) geometry. In SEIRAS-ATR, the surface plasmon resonance of rough metal films locally enhances the evanescent field, rendering the technique sensitive to sub-monolayers of species adsorbed on the electrode. 49 With this work, we establish SEIRAS-ATR in the Kretschmann configuration as a tool for probing water-oxidation intermediates on metal oxide catalysts. For this purpose, we first electrochemically deposited a thin layer of CoOx(OH)y 31 onto a chemically deposited Au thin film (CoOx(OH)y-Au) 50 on a micro-machined Si-ATR crystal, 51 which affords high infrared transparency below 1,200 cm−1. A scheme of the setup is shown in Figure S1 in the supplemental information. For SEIRAS-ATR, CoOx(OH)y instead of Co–Pi was employed as the prototypical catalyst because the latter would greatly complicate the interpretation of the IR spectra in the region around approximately 1,000 cm−1 owing to the phosphate anion and its response to the applied potentials. As will be discussed in detail later in this work, the electrochemical behaviors of CoOx(OH)y are comparable with Co–Pi. It also features structurally similar active sites and the same cobalt oxidation states under water-oxidation conditions as Co–Pi. 31 , 52 The CoOx(OH)y-Au film exhibits a large activity for water oxidation in comparison with the Au substrate (Figure S2). Figure 2 shows the steady-state spectra of the CoOx(OH)y-Au electrode in 0.1 M potassium phosphate (KPi) in D2O, H2O, and H2 18O. The absorbance was calculated according to absorbance = −log(S/R), where S and R refer to the sample and reference spectra, respectively, taken at 2.21 and 1.61 V. Unless otherwise noted, all potentials in this work are relative to the reversible hydrogen electrode. The spectrum in the D2O-based electrolyte exhibited a band centered at 1,014 cm−1 (at 2.21 V) (Figure 2A). The intensity of this band increased with increasing applied potential (Figure S3), suggesting that it is caused by a water-oxidation intermediate. To assign the band to a water-oxidation intermediate, we performed the following control experiments: First, to exclude the possibility that the band (1,014 cm−1) arises from a phosphate species in solution, we confirmed that the band also appears when the electrolyte is 0.1 M KCl in D2O and in H2O (Figure S4). Second, the band is absent on an Au electrode without the CoOx(OH)y film (Figure S4).These observations strongly suggest that the band centered at 1,014 cm−1 is a water-oxidation intermediate on CoOx(OH)y-Au. According to the proposed mechanism, this spectral feature can be associated with either Co–O–O⋅–Co from the IMOC pathway or Co–O–O⋅ or Co–O–OH from the WNA pathway (Figure 1). To further assign this band, we conducted isotopic labeling experiments with H2O and H2 18O. The lack of an isotopic shift when the solvent was switched from D2O to H2O implies that the vibrational mode of the species does not involve a hydrogen atom (Figure 2B). Upon switching to the H2 18O electrolyte, this band shifts to 966 cm−1 (Figure 2C). The 48 cm−1 difference (from 1,014 to 966 cm−1) indicates that the intermediate involves an O-containing motif. These experimental observations support the conclusion that the 1,014 cm−1 band is associated with the superoxide intermediates (Co–O–O•–Co or Co–O–O•). 13 , 43 , 53 The other possible water-oxidation intermediate, hydroperoxide (Co–O–OH), would feature characteristic bands in the 740–920 cm−1 region. 42 , 44 , 54 , 55 Owing to the absorption by the H2O librational mode, the signals were too weak to be discernable in that spectral range. The other bands in the spectra in Figures 2A–2C are due to the enrichment and depletion of electrolyte phosphate species at the interface with changes in applied potential. The magnitude of those spectral changes depends on the characteristics of a specific electrode, such as film thickness and homogeneity, and the electrolyte system. The negative band at ∼1,050 cm−1 in Figures 2A and 2B is likely due to a surface-adsorbed phosphate species. 56 , 57 The spectrum of a bulk KPi solution is shown in Figure 2D. Duplicate experiments confirm the reproducibility of the spectroscopic results (Figure S5). Taken together, this set of experiments demonstrates the utility of the SEIRAS-ATR technique for the detection of water-oxidation intermediates under operating conditions. Importantly, the result confirms the presence of a superoxo species, consistent with the mechanistic proposal (Figure 1). Future research should be directed to further distinguish between Co–O–O⋅–Co and Co–O–O⋅.To further probe the mechanisms as shown in Figure 1, we monitored the electrochemical water oxidation current as a function of electrode potential in water-in-salt electrolytes of varying water activities. As noted earlier, different reaction orders with respect to water activity are expected from the two competing mechanisms: a (pseudo) first-order dependence on H2O activity (a w ) is expected for the WNA route, whereas a (pseudo) zeroth-order dependence on a w is expected for the IMOC pathway. In a practical electrochemical system, the dependence of the kinetics on a w is likely more complicated because of a number of other factors, including the participation of H2O as a solvent; these potential complications notwithstanding, the value of quantitatively analyzing the reaction rates as a function of water activity becomes obvious. Figure 3 A compares the steady-state electrochemical current densities due to the oxidation of water on Co–Pi in contact with 0.1 M KPi containing 0, 2, 4, and 7 m NaNO3. The corresponding water activities are shown in the legend and were calculated on the basis of empirical equations. 58 These values describe the activity of bulk water in these water-in-salt electrolytes. We caution that the activity of water at the electrocatalytic interface may be different from those values. Nevertheless, the activities of interfacial water are expected to qualitatively follow the same trend with increasing water-in-salt concentration. All electrolytes were at neutral pH and were stirred during measurements, which were performed on electrodeposited Co–Pi on fluorine-doped tin oxide (FTO) substrates in a single-compartment electrochemical cell. The potential window was carefully chosen so as to avoid mass transport limitations (i.e., >1.71 V) or large experimental errors due to low current densities (i.e., <1.62 V). Details of the data collection protocol are given in the supplemental information, and a representative dataset is shown in Figure S6. As shown, with increasing molality of NaNO3 and, hence, decreasing a w , the current density of water oxidation is increasingly suppressed. A similar trend was observed for CoOx(OH)y (Figure S7), suggesting that the observed trend is a more general feature of cobalt oxide-based catalysts. This finding further corroborates our assertion made earlier that CoOx(OH)y is an appropriate alternative model system for Co–Pi.The observed suppression of the water-oxidation reaction could arise from a number of different physical phenomena. First, to test whether the catalyst undergoes irreversible structural changes in the different electrolytes, we recorded the cyclic voltammograms (CVs) of the same Co–Pi electrode in 0.1 M KPi before and after collection of 3 cycles of CVs in the four electrolytes (of molalities 0, 2, 4, and 7 m). As shown in Figure S8, the CVs in 0.1 M KPi before and after catalysis in the water-in-salt electrolytes overlap. These data suggest that no irreversible changes in catalytic activity occur during water oxidation in the water-in-salt electrolytes.Second, to test whether the mass transport of water to the electrode limits the reaction rate at high salt concentrations, we collected the steady-state electrochemical current densities of a Co–Pi-coated Pt rotating disk electrode (RDE) at rotation rates of 2,000 rpm (Figure S9) and 0 rpm (Figure S10). Comparison of the two figures reveals that the recorded current densities on the RDE exhibit the same trend with increasing salt concentration, irrespective of the rotation rate. Moreover, as demonstrated in Table S1, the increase in the thickness of the stagnant layer with electrolyte concentration is expected to be small. Collectively, these results suggest that the suppression of the water-oxidation reaction is not caused by limited mass transport of water to the electrode.Third, at high concentrations of NaNO3, nitrate anions are expected to limit the enrichment of phosphate anions in the electric double layer with increasing potential. As a result, the pH buffer capacity at the electrocatalytic interface might decrease with increasing NaNO3 concentration. Changes in the pH in the vicinity of the electrode (local pH) could impact the reaction rate and mechanism. 27 , 59 To exclude local pH effects as a possible reason for the reactivity trends with increasing NaNO3 concentration, we performed three different control experiments: (1) we monitored the electrochemical current density as a function of solution pH at a fixed (absolute) electrode potential. As shown in Figures S11–S13, the pH dependence of the current density was independent of the rotation rate of the RDE. (2) We performed galvanostatic titration experiments. The potential shows an approximately Nernstian shift of 60 mV/pH for all electrolytes (Figures S11–S13). (3) We varied the concentration of KPi in the electrolytes containing 4 and 7 m NaNO3. As shown in Figure S14, the potential dependence of the reaction rate is insensitive to the concentration of KPi. Taken together, these control experiments suggest that the local pH does not significantly depend on the concentration of NaNO3.Fourth, to test whether nitrate anions block catalytic sites, we recorded the electrochemical current density as a function of potential in 7 m NaClO4. Perchlorate typically does not chemisorb on electrodes. 60 As shown in Figure S15, the impact of 7 m NaClO4 on the current density is similar to that of 7 m NaNO3. This result indicates that nitrate anions do not block catalytic sites of Co–Pi.Fifth, alkali metal cations are known to influence the rate of the water-oxidation reaction on various electrocatalysts. 61–65 In the case of Ni oxyhydroxides, intercalated electrolyte cations have been proposed to stabilize reaction intermediates. 62 , 64 To test whether the catalytic activity is affected by the identity of the cation, we conducted additional control experiments in 2 m KNO3. As shown in Figure S16, the current modulation ratio virtually overlaps with the one obtained in 2 m NaNO3 (higher concentrations of KNO3 could not be tested because of the lower solubility of that salt relative to NaNO3). This result is consistent with earlier work 66 showing that the substitution of K+ in Co–Pi by Na+ has no significant impact on the catalytic activity of this catalyst. On the basis of this finding and our observation that the catalytic activity of Co–Pi is retained after a sequence of CVs in three water-in-salt electrolytes (Figure S8), we can exclude the incorporation of Na+ into the Co–Pi film as the origin of the change in catalytic activity with increasing electrolyte concentration. Cations can also influence an electrocatalytic process by altering the properties of the electric double layer in a number of distinct ways, 67 which are not fully understood to date. One of the principal ways in which cations can impact the catalytic activity is by altering the structure and dynamics of water at the interface. 65 , 67 This possibility is included in our interpretation of these results in terms of the decreasing activity of water with increasing concentration of the water-in-salt electrolytes.Sixth, to exclude the possibility that impurities, for example, Fe, incorporate into the catalyst 68 with increasing salt concentration, we performed CV tests in electrolytes with reagent grade and trace metal grade salts. As shown in Figure S17, the same water-oxidation activity was observed in both electrolytes.Finally, to test whether the electrochemical currents arise from the oxidation of water to molecular oxygen, we conducted gas chromatography measurements. Figure S18 shows that O2 is produced with near-unity faradaic efficiency. This measurement demonstrates that: (1) other possible oxidation products (such as H2O2) are not produced in appreciable amounts and (2) parasitic chemical reactions (such as the oxidation of nitrate) do not occur.Taken as a whole, this set of results indicates that the observed suppression of the water-oxidation reaction is most likely caused by the decrease of water activity (a w ) from 1 to 0.83 as the concentration of NaNO3 increases from 0 to 7 m.To further analyze the data shown in Figure 3A, we plotted the ratio of the current density at a w  = 1 over that at a w  = 0.83 at different potentials (Figure 3B). This ratio quantifies the extent to which the reaction rate is modulated by the water activity. It is clear that the impact of the water activity strongly depends on the electrode potential: at 1.71 V, the rate is suppressed by a factor of ≈4.3. By contrast, at a potential of 1.615 V, the modulation factor is only ≈1.2, indicating that the rate of the reaction is less sensitive to the change in water activity at that potential. Identical trends were obvious for the other a w ′s (i.e., 0.94 and 0.89), albeit with different magnitudes.That the reaction rate is suppressed by up to a factor of 4.3 by an a w change from 1 to 0.83 at 1.71 V strongly suggests that H2O is involved in the RDS at that potential. Conversely, for the same a w , the modulation is close to unity at 1.615 V, indicating that H2O involvement in the RDS is less likely. Taken as a whole, the data suggest that a mechanistic switch occurs between 1.615 and 1.71 V. A possible mechanistic switch that is consistent with our observations is the transition from the IMOC pathway ([pseudo] zeroth order in a w ) to the WNA route ([pseudo] first order in a w ) as the electrode potential is increased from 1.615 to 1.71 V.To corroborate further this assertion, we measured the steady-state current density on the FTO-supported Co–Pi electrode in 0.1 M KPi in heavy water (D2O) as a function of electrode potential. The ratio of the current density of the corresponding measurement in light water over that in heavy water is the apparent kinetic isotope effect (KIE). The apparent KIE is close to 2 at 1.625 V and increases to ≈4.2 as the potential is tuned to 1.71 V. Because the IMOC pathway does not involve water in the RDS, we expect the rate of the reaction to be insensitive to H/D substitution. By contrast, the WNA involves a water molecule in the RDS. Therefore, a dependence of the rate on the isotope of hydrogen is expected. Collectively, the KIE measurements further corroborate our notion that the mechanism switches from the IMOC route to the WNA pathway with increasing potential.We note that the interpretation of KIE effects can be highly complex. For example, a similar KIE dependence on potential might be explained by a switch of the oxidized substrates from OH− to H2O, as has been reported by Zhao et al. on Fe2O3. 69 However, that mechanism is not applicable to the Co–Pi catalyst because OH− is unlikely to be the oxidized substrate at pH 7. Furthermore, Hammes-Schiffer et al. demonstrated that the relative contributions that specific reactant/product vibronic states make to the current density are dependent on the isotope. 70 They showed that this effect could give rise to a potential dependence of the KIE. Although we cannot fully rule out that such effects contribute to the potential dependence of the KIE in the present case, the corroboration between the KIE data and the potential-dependent impact of the water activity on the reaction rate supports the conclusion of a potential-induced switch from the IMOC mechanism to the WNA pathway with increasing potential. A KIE on the WNA pathway was also reported by Cuk et al. during the photocatalytic oxidation of water on SrTiO3. 18 As far as the KIE effect is concerned, it is noted that Dau and co-workers also found a suppression of the water-oxidation reaction in D2O relative to that in H2O. 32 Their electrokinetic results were similar to those reported herein. However, they interpreted these data differently. In particular, the authors found that the redox potential of the pre-equilibrium [CoIII–OH] ⇋ [CoIV–O] + H+ + e− shifts by approximately 60 mV in the anodic direction upon switching the solvent from H2O to D2O. Because galvanostatic measurements for water oxidation in H2O and D2O show a similar shift, they suggested that the suppression of the water-oxidation reaction is due to the shift in this pre-equilibrium (rather than a KIE on the RDS of the water-oxidation reaction). This pre-equilibrium is a critical factor determining the activity of Co-oxide-based catalysts. 27 , 31 , 71 This alternative interpretation could also account for the observed suppression of the water oxidation in D2O. However, we note that on the basis of the CVs of Co–Pi in H2O and D2O (Figure S19), we estimated a shift of ≈28 mV in the Co(II)/Co(III) redox half-wave potential. The relatively small shift in the pre-equilibrium suggests that it may not be the sole reason for the observed dependence of the rate of the water oxidation on the H/D isotope. Most importantly, this interpretation cannot account for the suppression of the current with increasing salt concentration (Figure 3). As discussed earlier, our control experiments in which we varied the rotation rate of the RDE (Figures S9 and S10), the pH of the electrolyte (Figures S11–S13), and the concentration of KPi (Figure S14) confirm that the buffer capacity is sufficient to maintain the [CoIII–OH] ⇋ [CoIV–O] + H+ + e− equilibrium in the water-in-salt electrolytes. To further corroborate this notion, we analyzed the Co(II)/Co(III) redox equilibrium of Co–Pi in contact with the water-in-salt electrolytes with cyclic voltammetry. As shown in Figure S20, the Co(II)/Co(III) redox half-wave potential is shifted by only 10–20 mV in the cathodic direction with increasing salt concentration. This small shift indicates the pre-equilibrium is not significantly affected by the presence of water-in-salt electrolytes. Therefore, when the isotope effect results are viewed in the context of the electrokinetic results for the water-in-salt electrolytes, our interpretation provides a cohesive, self-consistent picture, whereas the hypothesis of the shift in the pre-equilibrium can only partly explain the collective results. Although the shift may be a contributing factor, we conclude that it is not the dominating effect.In the following section, we discuss two possible molecular origins for our proposed potential-induced mechanistic switch. First, we show that the interfacial electric field at the electrocatalyst/electrolyte contact may affect the relative activation barriers of the two pathways and, thus, the relative weight of each route as the potential is altered. Second, we performed a DFT study of the two routes. These calculations show that only at high potentials does the WNA mechanism become thermodynamically accessible. In a practical system, the two effects may synergistically combine to favor the WNA pathway at high electrode potentials. Next, we discuss the impact of the interfacial electric field on the activation barriers; then we describe the insights derived from the DFT modeling.The key distinction between the IMOC and WNA pathways is the involvement of water in the RDS of the latter one (Figure 1). On the basis of this observation, we expect the energetics of the two pathways to exhibit distinct sensitivity to the interfacial electric field. The magnitude of the interfacial electric field of the electric double layer increases as the potential of the electrode is increased. It is well established that electric fields can profoundly impact the rates and selectivity of chemical reactions. 72–76 Reaction intermediates with sufficiently large dipole moments and polarizabilities can interact with the electric fields. As a result of this interaction, the free energy profile of the reaction processes can be altered. 72 , 73 Nørskov et al. have shown that the impact of electric fields on surface-bound water-oxidation intermediates (M–OOH, M–OH, M=O) is typically very small because these species have small dipole moments and polarizabilities. 75 On the basis of these findings, it is likely that the interfacial electric field has a negligible impact on the IMOC pathway. Because the rate-determining O–O bond-formation step is a chemical step, we expect the principal activation barrier of the IMOC pathway to be independent of the electrode potential. By contrast, because water has a relatively large dipole moment and polarizability, the orientation and dynamics of water molecules at electrified interfaces may strongly depend on the electrode potential. 76–78 It has been suggested that the water dynamics and structure at interfaces affect the rates of various electrocatalytic processes, such as water oxidation and reduction. 65 , 76 Therefore, even though O–O coupling in the WNA as hypothesized in Figure 1 is a chemical step, we expect the activation barrier of this process to depend on the electrode potential: Δ G ¯ W N A ≠ = Δ G W N A ≠ − Δ μ → ⋅ E → , where Δ G W N A ≠ is the standard chemical free energy of activation in the absence of an electric field; Δ μ → represents the change in the surface dipole when going from the reactant to the activated complex state; and E → is the interfacial electric field, which depends on the electrode potential. These qualitative considerations show that because of the participation of water in the rate-determining chemical step of O–O bond formation for the WNA mechanism, the activation barrier of this step is a function of electrode potential. Nevertheless, without knowledge of the molecular-level structure of the electrocatalyst/electrolyte interface at the present time, our considerations must remain qualitative at the present stage. Irrespective, this model describes one possible origin of the observed mechanistic switch from the IMOC route to the WNA pathway with increasing potential.To explore further other possible causes of the potential-induced switch, we studied the energetics of the two pathways with DFT. All calculations were performed with the B3LYP functional and def2-SV(P) and def2-TZVP basis set implemented in the Gaussian 16 software package. Further computational details are provided in the supplemental information. We constructed atomic models on the basis of previous EXAFS 34 and X-ray pair distribution function analysis. 52 The Co7O24H27 cluster has a Co ion surrounded by 6 Co ions at the edge that are connected to the center Co ion by μ3-O bridges (Figure S21). The energetics of the water-oxidation reaction is sensitive to the protonation state of the cluster. 35 , 36 We considered different protonation states and found that the lowest energy protonation state is a highly symmetric cluster with one side of the μ3-O being protonated and each pair of edge Co ions having strong hydrogen bonds between nearby hydroxide and water ligands (Figure S22). The protonation of the hydroxide ligand of the edge Co atoms is energetically unfavorable because it destroys the strong hydrogen bond interaction between OH− and nearby H2O. However, the edge OH− group can be protonated by reducing the corresponding edge Co(III) to Co(II) (Figure S23).On the basis of this structural model, we investigated the water-oxidation mechanism (Figure 4 ) starting from the H2O–Co(II)–(μ-Ο)2–Co(III)–OH2 intermediates (I). We note that our computational method overestimates the potential for oxidation potential of Co(III) to Co(IV) by ∼0.3 V (Figure S24). All potentials quoted herein are not corrected for this systematic error. The oxidation of Co(II) to Co(III) requires 0.95 V, which is much lower than the applied potential during catalysis. The second oxidation requires 1.98 V to generate intermediate III with one Co oxidized to Co(IV). This oxidation is a metal-center oxidation, consistent with X-ray absorption near edge structure results of the Co–Pi catalyst under catalytic conditions, which suggest a valence of Co greater than 3. 34 When the overestimation of the redox potential is accounted for, this intermediate is predicted to be prevalent under water-oxidation conditions. Consistent with the prediction, the resting state of the catalyst has been assigned to intermediate III in previous reports. 15 , 27 , 31 , 32 , 59 , 79 The hydroxide coordinated to the Co(IV) center in intermediate III has a partial radical character as indicated by a Mulliken spin population of 0.21 (Figure S25). Therefore, the two hydroxides can couple to form hydroperoxide through the IMOC mechanism. Thermodynamically, this pathway is favored over the WNA pathway under low applied potentials. The following two oxidations require low potentials. Therefore, it is fairly easy to form intermediate VI. The release of O2 and binding of two water molecules complete the catalytic cycle.Under high applied potential, intermediate III can be further oxidized to form intermediate IV′ with two nearby Co being oxidized to Co(IV). The terminal O atom coordinated to Co(IV) is best described as an oxyl radical because the Mulliken spin population on the O atom is 0.89 (Figure S25), close to 1 for a perfect radical. The intermediate IV′ can react with a water molecule from the solution to form intermediate V′ through the WNA mechanism. The incoming H2O forms hydroperoxide with the oxyl radical and releases a proton to the nearby OH− group. Intermediate V′ can be further oxidized to intermediate VI′, which releases O2 and adsorbs a water molecule to complete the catalytic cycle.We note that both IMOC and WNA mechanisms feature a superoxo intermediate (VI and VI′, respectively). This prediction is consistent with our spectroscopic results, which indicate the presence of a superoxo species. On the basis of the simulated O–O vibrational frequencies (Figure S26) alone, we cannot identify which of the two species gives rise to the vibrational band at 1,014 cm−1 (Figure 2). We reserve a more detailed assignment for future investigations.Although alternative reaction pathways may be available, 36 , 80 the DFT computations show that: (1) the IMOC and WNA pathways are feasible from a thermodynamic perspective and (2) their energetics are consistent with the proposed mechanistic framework (Figure 1) and the interpretation of our electrokinetic results (Figure 3); at low overpotential, the IMOC pathway predominates, whereas the WNA pathway becomes accessible at high overpotential. Finally, it is noted that, in line with previous precedence, we only considered the thermodynamics of the pathways. 19 , 80 The calculation of the activation barriers is complicated by spin-state changes during the conversion of intermediate III to IV. Furthermore, the activation barriers are sensitive to the protonation state of the catalyst, which is a complex function of applied electrode potential and reaction conditions. Fully accounting for these complications will require additional research that is beyond of the scope of the current work.Taken as a whole, the thermodynamic description of the two pathways and the qualitative considerations of the impact of the interfacial field on the relative magnitude of activation barriers of the O–O bond-forming steps provide strong support for the conclusion of a potential-dependent mechanistic switch. The DFT modeling predicts that a certain threshold potential for the WNA pathway needs to be surpassed before this pathway becomes thermodynamically feasible. In addition, the involvement of water in the RDS may further lower the activation barrier for the O–O bond-formation step for the WNA route, leading to a further acceleration of the reaction rate. Our conclusions are graphically summarized in Figure 5 .Previous research on homogeneous water-oxidation mechanisms has revealed that the 4-proton, 4-electron process of water oxidation can take place on a mononuclear or a dinuclear catalyst. Whether WNA or oxygen coupling is the preferred mechanism has been at the center of intense studies for the natural PSII and for molecular catalysts. In testing the various hypotheses for the reaction mechanisms, researchers mainly relied on kinetic models that depend on the information of key species, such as the concentration of the catalyst and the TOFs. In principle, the same methodology could be applied for the establishment of a similar knowledge base for heterogeneous water-oxidation reactions. However, the lack of knowledge on the detailed information of the catalytically active centers creates a critical challenge for such an approach. Our strategy of probing the kinetics of heterogeneous water oxidation by varying water activities is new. It generates information that permits the verification of hypotheses that are difficult or impossible to obtain by other methods. How the water-oxidation reaction proceeds is sensitive to a number of factors, including the local catalytic environment (e.g., the availability of mononuclear, dinuclear, or multinuclear active centers), substrate concentration, and the driving forces (see, e.g., a recent report on how varying applied potentials change water-oxidation products on a copper porphyrin in acidic solutions. 81 ) Although our studies suggest that the WNA mechanism is favored at high driving force, we are inspired to understand that in a practical water-oxidation system (such as the oxygen evolution catalyst in PSII or in an electrolyzer), both mechanisms may co-exist. The fact that this switch is observed on Co–Pi and CoOx(OH)y (Figures S7 and S27) suggests that the potential-induced changes in pathway may be a more general phenomenon of Co-oxide-based electrocatalysts. The dynamic switch of the mechanisms could help to explain how nature ensures the most efficient route for the utilization of solar energy to liberate electrons and protons; it also implies that future designs and optimization of heterogeneous catalysts for large-scale engineering implementations of water oxidation should consider the facile switch of reaction mechanisms. It is noted that the WNA mechanism could proceed through a mononuclear site or a dinuclear site depending on the catalytic conditions. 13 , 82 , 83 However, it likely makes only a minor contribution to our study because of the narrow and low overpotential regime investigated and the equivalent involvement of a water molecule in the RDS on both sites. Finally, although we envision that studying water oxidation by varying water activities indeed adds a new dimension to the tool kit, it faces limitations for systems at highly alkaline conditions where OH− but not H2O is being oxidized.In conclusion, this work introduced two key innovations. Using SEIRAS-ATR, we detected a key intermediate corresponding to O–O bond formation in Co-based water oxidation. This information lends support to the proposed mechanisms. By varying the water activity, we established a kinetic model that allowed us to verify the two competing mechanisms of water oxidation. We found that the dinuclear route (i.e., IMOC) is favored at relatively low driving forces, whereas the mononuclear route (i.e., WNA) is preferred at relatively high driving forces. The results contribute significantly to the understanding of water oxidation by heterogeneous catalysts.Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Dunwei Wang ([email protected]).This study did not generate new unique reagents.This study did not generate any datasets.The work at Boston College is supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division (DE-SC0020261). Work at Yale University was supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division (DE-FG02-07ER15909). V.S.B. acknowledges the computer time from the National Energy Research Scientific Computing Center (NERSC) and Yale Center for Research Computing (YCRC).C.L. and Y.W. performed the electrochemical experiments; J.L. performed the SEIRAS-ATR experiments; K.R.Y. conducted computational studies; J.E.T., Q.D., D.H., and Y.Z. contributed experimentally; all authors participated in discussions and the writing of the manuscript; V.S.B., M.M.W., and D.W. co-directed the project.The authors declare no competing interests.Supplemental information can be found online at https://doi.org/10.1016/j.chempr.2021.03.015. Document S1. Figures S1–S27, Table S1, supplemental experimental procedures, and supplemental references Document S2. Article plus supplemental information

O–O bond formation is a key elementary step of the water-oxidation reaction. However, it is still unclear how the mechanism of O–O coupling depends on the applied electrode potential. Herein, using water-in-salt electrolytes, we systematically altered the water activity, which enabled us to probe the O–O bond-forming mechanism on heterogeneous Co-based catalysts as a function of applied potential. We discovered that the water-oxidation mechanism is sensitive to the applied potential: At relatively low driving force, the reaction proceeds through an intramolecular oxygen coupling mechanism, whereas the water nucleophilic attack mechanism prevails at high driving force. The observed mechanistic switch has major implications for the understanding and control of the water-oxidation reaction on heterogeneous catalysts.

Magnetic nanoparticles are highly exploited in various fields like ferrofluids, magnetic separations, magnetic drug delivery, magnetic data storage systems, magnetic resonance imaging ( Cao et al., 2014 ; Song et al., 2015 ; Xuan et al., 2011 ) recording, absorbents ( Andjelkovic et al. 2018 ; Jacek, 1984 ; Mazen and Abu-Elsaad, 2015 ; Saha et al., 2018 ), telecommunications, transformers, etc. ( Amor et al., 2019 ). Their properties differ in their methods of preparation and hence the oxide content ( Karimi et al., 2015 ; Mohamed and Abu-Dief, 2020). They are highly sensitive to the extrinsic elements used as dopants ( El Foulani et al. 2019 ). The doping can be done in a controlled way that tunes their properties and performance easily ( Supriya et al., 2017 ). The quality of the ferrites depends on their electromagnetic properties ( Ahmed et al., 2021 ). The most common fabrication methods for such ferrites are Auto combustion, hydrothermal, citrate gel, microemulsion, ceramic, co-precipitation, etc. ( Chandramouli et al., 2021 ; Himakar et al., 2021 ; Hussain et al., 2021 ; Jesus Mercy et al., 2020; Mulushoa et al., 2018 ; Parajuli et al., 2021b; Parajuli and Samatha, 2021b ; Subrahmanya Sarma et al., 2022). The external elements can either be added or substituted for the betterment of their electrical and magnetic properties. The most commercially used mixed ferrites are cobalt-cupper, manganese-zinc, nickel-zinc, magnesium-manganese ferrites, etc. ( Jasrotia et al., 2020 ). The content of dopant is another parameter for the resulting property of such ferrites ( Farea et al., 2008 ). Recently, F. Matloubi Moghaddam et al. studied the CoCuFe2O4 ferrites and found a good catalyst for the cyanation of amines ( Ramakrishna et al., 2018 ). Hadi et al. studied Cu-doped Co-Zn ferrites and their applicability in multilayer inductor chip applications in Moghaddam et al. (2021 ). Sabih et al. investigated Ce and Zn substituted Co-Cu ferrites and found them to be of soft magnetic behavior with their applicability in water treatment, recording, and memory devices (Qamar et al., 2020). We have studied vigorously different ferrites materials based on which this work is initiated ( Hadi et al., 2021 ; Parajuli and Samatha 2021a , Parajuli and Samata 2021b ; Parajuli et al., 2022a ; Parajuli et al., 2021a).In addition, Ajeesha et al. (2022 ) studied the antibacterial activity of Ca1-xCuxFe2O4; x = 0, 0.2, 0.4, 0.6, 0.8, 1) incorporating BET analysis and found 96% degradation in 180 min using methylene blue dye. Almessiere et al. (2019 ), have studied CoNdxCexFe2–2xO (x = 0.00, 0.03, 0.05, 0.09, 0.10, 0.15, and 0.2) and found their excellent anticancer activity. Ameerah and her group investigated hydrothermally prepared Co0.5Ni0.5BixFe2-xO4 (x = 0.00–0.10), found a single domain structure through its squareness ratio of less than 0.5 and are sensible and efficient enough for spin reversal (Ameerah et al., 2022 ). Recently, the rare earth elements, like , Thulium substituted CoTmxFe2-xO4 (x = 0.00, 0.04 and 0.08)(Ünal et al., 2020), Dysprosium substituted Mn0.5Zn0.5Dy0.03Fe2–0.03O4 (x = 0.005, 0.01, 0.015, 0.02, 0.025, and 0.03) for sensor applications. In the last couple of years, K. M. Jadhav et al. have intensively studied the interesting properties of different types of ferrites with the substitution of various elements. The researches were focused on surface modification (Somvanshi et al., 2020a), RE metal Dy substituted yttrium iron garnet (Y3-xDyxFe5O12) nanoparticles for advanced high frequency devices ( Bhosalea et al. 2020 ), RE metal Gd substituted Zn-Mg ferrite for biomedical applications (Somvanshi et al., 2020b), thermal conductivity ( Kharat et al., 2020 ), Viral RNA-extraction protocol for COVID detection (Somvanshi et al., 2021), green method to prepare CoFe2- x Al x O4 to enhance the dielectric parameters ( Chavan et al., 2021 ), reviewed ferrites materials for biomedical applications ( Kharat et al., 2020 ), Ni-Zn ferrites incorporated with rhodamine B for photocatalytic purposes ( Jadhav et al., 2020 ), Zn Fe2O4 for chalcones preparation which gives pigment to flowers in nature ( Borade et al., 2020 ), surface functionalized CoFe2O4 in hyperthermia for cancer and other treatments ( Eivazzadeh-Keihan et al., 2021 ; Eivazzadeh-Keihan et al., 2020 ; Kharat et al., 2020 ), NiFe2O4 for toxic dye removal ( Jadhav et al., 2021 ), thermoacoustics ( Kharat et al., 2019 ), Mg substituted ZnFe2O4 for absorbents ( Somvanshi et al., 2020 ), hyperfine interaction study for identifying ferrimagnetic nature ( Humbe et al., 2020 ), green vs ceramic synthesis comparative study of multifunctional Mg-Zn ferrites nanoparticles for electronic and biomedical treatments ( Khirade et al., 2020 ). Similarly, many bionanocomposites of ferrites nanoparticles are used to prepare Pyranopyrazoles for analgesic agents in medical treatments ( Kamalzare et al., 2021 ) and various others for the recyclable nanocatalysts, biosensors and so on ( Bahrami et al., 2020 ; Kamalzare et al., 2020 ; Maleki et al., 2020 ;Maleki et al., 2017).We have used the sol-gel auto combustion method in this work to prepare Ni, Zn, and Mg substituted Co-Cu ferrite nanoparticles and studied their structural, dc electric resistivity, and magnetic properties. The reason for choosing this method was due to its simple process, easily controlled stoichiometry, cheap starting materials, and producing ultrafine particles in a short time at very low temperatures ( Parajuli et al., 2022 ).The highly analytical grade (99.9% purity) nitrates of cobalt, copper, magnesium, nickel, zinc, and iron were mixed with citric acid in a 1:1 molar ratio. Citric acid acts as a fuel for auto-combustion and makes the precipitation homogeneous at relatively low temperatures. The dropwise addition of ammonium hydroxide makes the solution neutral. The electronegative ions due to oxygen attract the metal ions with electropositive nature and get dissolved. The solution is stirred at 150 °C until a gel is formed and dried. It is sintered at 1000 °C for 3 h. The product is then grounded finely using agate mortar and pestle. Finally, they are pressed with 5 tons hydraulic system to get pallets. The two flat surfaces of the pallets can be coated with silver to form two electrodes which can be used for the determination of their electric and other properties. The chemical name and molar mass of the chemical used are listed in Table 1 .In this study, a Rigaku Miniflex II X-ray diffractometer incorporated with CuKα radiation (wavelength = 1.5406 Å) system was used for their structural properties. TESCAN, MIRA II LMH microscope with attached EDX, Inca Oxford was used for morphological and compositional properties respectively. Their functional groups were confirmed with the FTIR study. The magnetic properties and DC resistivities were studied with the help of the EZ VSM model and a two-probe DC resistivity system.The XRD pattern of Mg, Ni, and Zn substituted Co-Cu nano spinel ferrite is shown in Fig. 1 . The lattice parameter is affected by several factors such as the atomic size, interatomic forces, particle or grain size ( Ramakrishna et al., 2018 ), etc. X-ray diffraction measurements validated the mono-phase formation and nanocrystalline character of the produced samples. The peaks (111), (311), (222), (400), (422), (511), and (522); obtain after the indexing confirms the cubic structure of the samples. The lattice parameters obtained were 8.474 Å, 8.374 Å, and 8.393 Å for synthesized ferrites respectively. The Zn substituted sample has a higher lattice constant than the other two. The ionic radii of Cu2+, Zn2+ and Mg2+ ions were 0.73 Å, 0.74 Å, and 0.72 Å respectively higher than that of Ni2+ (0.69 Å) (Anand et al., 2017).The crystallite sizes of the samples were obtained with the help of the following relation ( Amar et al., 2019 ) considering the highest peak of the (311) plane; (1) a = d h 2 + k 2 + l 2 where d is interplanar spacing which can be calculated from Bragg's law. The accuracy of the lattice constant is done by the Nelson-Riley extrapolation function given by ( Batoo et al. 2016 ), (2) F ( θ ) = 1 2 [ ( c o s 2 θ sin θ ) + ( c o s 2 θ θ ) ] ± 0.002 Å The data obtained were listed in Table 2 which agrees well with the previous ( Yadav et al., 2018 ).The XRD patterns show the single-phase spinel ferrite matching with a JCPDS-card no. 22–1086 corresponding to (311) peak ( Jnaneshwara et al., 2014 ). The nano range of the average crystallite sizes as indicated by the sharpness of the peaks is connected by Debye Sherer's equation. The crystallite sizes were found at 47.19 nm, 40.15 nm, and 29.01 nm, respectively while sintering at 1000 °C gives. The crystallite sizes (below 100 nm) confirmed the nanocrystalline nature of the sample prepared ( Bhosale et al., 2020 ). The average crystallite size is obtained by the use of Scherer's formula for (311) peak ( Akhtar et al., 2019 ), (3) D ( 311 ) = 0.9 λ β c o s θ Where λ is equal to 1.5406 Å, β and θ are full width half maximum (FWHM) of (311) peak and angle of diffraction respectively.The X-ray density is given by ( Rouhani et al., 2018 ), (4) d x = 8 M ′ N a 3 M' is the prepared samples' molecular weight, a is the lattice constant, and N is Avagadro's number.Bulk density is given by ( Chaudhari et al., 2015 ), (5) d b = W 1 W 1 − W 2 W1 is the weight of the sample in air, and W2 is the sample's weight in water.Porosity gives the empty spaces or voids in a material with the help of the relation (Yoon and Raju, 2016) (6) p = 1 − d b d x % where db and dx are the bulk and X-ray densities, respectively. The plot between lattice constant and crystallite size is displayed in Fig. 2 . Fig. 3 shows the X-ray density and porosity variation of the synthesized ferrite nanoparticles. Here, the X-ray density and porosity values have been increasing in the order of Mg, Ni and Zn substituted Co-Cu nano ferrites samples.The volume of the unit cell is given by, (7) V = a 3 Å 3 showing the same trend as that of the lattice parameter since it is directly related with it.The ionic radii can be calculated with the help of two relations (Somvanshi et al., 2020a), (8) r A = ( u − 1 4 ) a 3 − r ( O 2 − ) Å (9) r B = ( 5 8 − u ) a − r ( O 2 − ) Å FESEM images of the Mg, Ni, and Zn substituted Co-Cu nano spinel ferrite are as shown in Fig. 4 (a) – (c), offering their spherical grains with particle sizes 40 nm, 50 nm – 74 nm, in agreement with XRD graphs. The figure shows clear and inhomogeneous crystal grains. Nevertheless, the density is reduced due to Cu with higher atomic weights (Nikmanesh and Eshraghi, 2019).The EDS analysis was used to determine the identity of metals in the synthesized Mg, Ni, and Zn substituted Co-Cu nano spinel ferrites are shown in Fig. 5 (a) – (c). From the EDS images, it is clear that Co, Cu, Mg, Ni, Zn, Fe, and O ions are visible. The histograms of compositions are shown in Fig. 6 , and their elemental composition has weight and atomic values, which are listed in Table 3 . The results are in good agreement with the standard stoichiometric ratio.The FTIR spectrum of M substituted (M = Mg, Ni, and Zn) Co-Cu nano spinel ferrite is shown in Fig. 7 . Two absorption peaks below 600 cm−1 of all ferrites indicate their spinel structures. The spectra of all the considerable series of powders show the presence of two primary absorption bands in which one is intense in the wavenumber around 300 cm−1 while the other band is around 600 cm−1. The higher band (ʋ1) is generally observed in the range 579–592 cm−1 Mtetra↔O. is caused by the stretching vibrations of the tetrahedral metal-oxygen band. The extension of FT-IR bands might be a result of the cation integration in the crystal layers. The lowest band (ʋ2) commonly seen in the range 357–397 cm−1 is caused by the metal-oxygen vibrations in the octahedral site Mocta ↔O. Such an IR wave number absorption difference can be expected because of the difference in bond lengths (Me-O) at the two sites.The two prominent peaks observed in 550 cm−1 and 453 cm−1 are due to the presence of the M–M bond and M–O stretching frequencies, respectively ( Shahbaz et al. 2012 ). The two peaks at 1603 cm−1 and 1115 cm−1 are associated with C=O and CO2 stretching vibrations respectively ( Batoo et al. 2016 ). The Mtetra↔O stretching bond has with higher frequency (ʋ1), and Mocta↔O has a lower frequency (ʋ2) with respective force constants K t and K o at tetrahedral and octahedral sites (Kumar and Shirage, 2017) as listed in Table 4 .The hysteresis loops of Mg, Ni, and Zn substituted Co-Cu nano spinel ferrite at room temperature are shown in Fig. 8 . They show the showing ferromagnetic nature. The saturation magnetization (Ms), remanent magnetization (Mr), coercivity (Hc), experimental and theoretical Bohr magnetron, and anisotropy values of the samples were obtained from these hysteresis loops ( Draack et al., 2019 ). Likewise, the variations of saturation magnetization (Ms) and magnetic moment (nB) of the samples are shown in Fig. 9 . From the figure, the Mg, Ni, and Zn substituted Co-Cu nano spinel ferrites have the saturation magnetizations (Ms) 42.95, 49.5, and 41.3 emu/g; remnant magnetizations (Mr) 12.4, 14.8, and 11.8 emu/g; magnetic moments (nB) 1.79, 2.09 and 1.7 µB; and the coercivities (Hc) 264.14, 566 and 409.5 Oe respectively. The FESEM micrograph shows the agglomeration which decreases the surface defect of the particle ( Zubair et al., 2017 ). We have observed that the values of the Ms, Mr, nB, and Hc are higher for the Ni2+ substituted Co-Cu than that of Mg2+ and Zn2+ substituted ones. The addition of nonmagnetic Cu2+ ions decreases the magnetic saturation as the divalent Cu2+ ion occupies tetrahedral sites, thereby replacing magnetic Fe3+ with octahedral ( Li et al., 2016 ). It gives the inverse spinel structure ( Zubair et al., 2017 ) according to Neel's sublattice model ( Ati et al., 2021 ). The ferrites have A-A, B-B, and A-B exchange interactions, out of which A-B exchange interactions are stronger. The Co2+, Cu2+, Mg2+, Ni2+, and Zn2+ ions occupy the octahedral site (B-site) thereby reducing magnetization as tetrahedral (A-site) has constant sublattice. Among these three interactions, the A–B interaction is (antiferromagnetic and) the strongest, and it predominates over AA and BB (ferromagnetic) interactions in determining the magnetic moment of the spinel ferrite compounds. The net magnetic moment of the lattice is therefore the difference between the magnetic moments of B and A sub-lattices, i. e., µB = µB(B) - µB(A) and its magnitude is given by µB = |µB(B) - µB(A)|. The occupancy of ions on tetrahedral and octahedral can be explained [Co0.5Cu0.2Fe]A [Co0.5M0.3Fe]BO4 where M = Mg, Ni, and Zn. M replaces Cu, the Co2+ cations do not change their occupancy. The reason seems to be clear that M2+ and Cu2+ occupy their preferred sites, i.e., tetrahedral and octahedral sites, respectively.The value of the remnant ratio (R = M R /M S) below 0.5 is isotropic with a single domain ferrimagnetic nature ( Mandal et al., 2017 ). Magnetic recording media need high M s ( Nayeem et al., 2017 ). The magnetic parameters are listed in Table 5 . The coercivity (HC) variation vs. magneto-crystalline anisotropy constant (K1) plots for Mg, Ni, and Zn doping Co-Cu nano spinel ferrite samples are shown in Fig. 10 .The coercivity (HC) is obtained from ( Humbe et al., 2018 ), (10) H C = 2 K 1 μ o M s The anisotropy constant (K1) and aspect ratio (MR/MS) can be obtained using ( Mandal et al., 2017 ), (11) Aspe ct Ratio = M R M S (12) K 1 = H C × M S 0.96 where MS, MR , and μo are the saturation magnetization , the remnant magnetization, and the free space permeability respectively.The magnetic moment (nB) of synthesized nano ferrites can also be calculated by the relation as specified below ( Chuang et al., 2015 ), (13) μ B = M s × MW 5585 where MW denotes the molecular mass of each synthesized specimen.There might be three reasons for the decrease in saturation magnetization in Mg and Ni substituted compounds: (1) single domain to multi-domain phase transition with increasing size (2) combined surface effect and its surface anisotropy (3) inert surface layer. The variation in magnetic saturation of impurity less and single-phase homogeneous crystal without defects is due to particle size variation of the samples. This behavior was already reported for different ferrites ( Ren and Xu 2014 ).The observed resistivity variation for the Mg, Ni, and Zn substituted Co-Cu nano spinel ferrites is according to the hopping mechanism ( Saini et al., 2016 ) which was reported by Verwey and his co-workers. During the sintering process, there is a possibility of forming Fe2+ and Fe3+ ions in both divalent and trivalent iron ions in octahedral sites between which their hopping takes place. For higher sintering temperatures, the formation of Fe2+ ions can also be from the evaporation of some elements. Co2+ ⇔ Co3+ and Cu+ ⇔ Cu2+ hopping also take place in the present system, mainly in octahedral sites ( Stergiou and Litsardakis 2014 ).In Mg, Ni, and Zn substituted Co-Cu ferrite samples; the resistivity is expected to get modified with the increase in Zn compound in place of Co ions through the following mechanisms: 1 Fe2+ ⇔ Fe3+ ions; electron hopping between iron ions in different valence states is supposed to be the dominant conduction mechanism thereby reducing resistivity considerably. 2 Cu+ ⇔ Cu2+ ions; the probability of forming Cu+ ions may be small. Due to this, conduction through this mechanism is also expected to be relatively low thereby resulting in low resistivity. 3 Co2+⇔ Co3+ ions; there is some probability of this kind of conduction mechanism in which there is the possibility of the formation of Co3+ ions in B-sites thereby lowering the resistivity. 4 Fe2+ + Co3+ ⇔ Fe3+ + Co2+ ions; the simultaneous presence of iron in cobalt ions in different valence states may often get locked in this mechanism thereby increasing the resistivity. 5 Fe2+ + Cu2+ ⇔ Fe3+ + Cu+ ions; the simultaneous presence of iron in copper ions in different valence states may also often get locked in this mechanism. But the increase in resistivity is small. 6 The presence of Zn ions between the iron ions increases the bond length thereby reducing electron hopping or increasing the resistivity. 7 Fine-grained microstructures with relatively more grain boundaries; since grains contribute to conduction and their boundaries in insulation, the fine-grained microstructures often increase resistivity. Fe2+ ⇔ Fe3+ ions; electron hopping between iron ions in different valence states is supposed to be the dominant conduction mechanism thereby reducing resistivity considerably.Cu+ ⇔ Cu2+ ions; the probability of forming Cu+ ions may be small. Due to this, conduction through this mechanism is also expected to be relatively low thereby resulting in low resistivity.Co2+⇔ Co3+ ions; there is some probability of this kind of conduction mechanism in which there is the possibility of the formation of Co3+ ions in B-sites thereby lowering the resistivity.Fe2+ + Co3+ ⇔ Fe3+ + Co2+ ions; the simultaneous presence of iron in cobalt ions in different valence states may often get locked in this mechanism thereby increasing the resistivity.Fe2+ + Cu2+ ⇔ Fe3+ + Cu+ ions; the simultaneous presence of iron in copper ions in different valence states may also often get locked in this mechanism. But the increase in resistivity is small.The presence of Zn ions between the iron ions increases the bond length thereby reducing electron hopping or increasing the resistivity.Fine-grained microstructures with relatively more grain boundaries; since grains contribute to conduction and their boundaries in insulation, the fine-grained microstructures often increase resistivity.Mechanisms 1 and 2 are dominant in our present study of Zn-like element substitution where the resistivity is decreased significantly first and further decrease is controlled by 3, 4, 5, 6, and 7 mechanisms. Fig. 11 shows the plot of temperature vs. dc electrical resistivity (log ρ vs. 1000/T) of Mg, Ni, and Zn substituted Co-Cu nano spinel ferrite. From the figure, it is clear that the resistivity is decreased with the increase in temperature of all the synthesized nano spinel ferrite showing their semiconducting nature ( Hossain et al., 2021 ). The impurities increase conductivity at a lower temperature. The energy needed for hoping of the electrons or activation energies are obtained from the fitted curves of the Arrhenius equation (Roy and Bera, 2006): (14) ρ = ρ o e − ( Δ E K T ) where ρ is the room temperature dc electrical resistivity at temperature T, ρo is the pre-exponential factor , Δ E is the activation energy, K is the Boltzmann constant, and T is the absolute temperature. The activation energy based on hopping is found in 0.58, 0.51, and 0.583 eV. Above this range, the hopping is called polaron hopping. The hopping can be of either electron or hole. In the case of polaron hopping, there is p-type or hole hopping in which there is hardly a quantum mechanical effect due to the larger grain size. The graph of the activation energies for Mg-Cu, Ni-Cu, and Zn-Cu cobalt nano spinel ferrites is shown in Fig. 12 .The activation energy needed for Ni substituted Co-Cu ferrite is the lowest indicating its preference for the system for higher conductivity. Further, these semiconducting ferrites systems have isotropic ferromagnetic nature. As a result, they behave as magnetic semiconductors as defined by Chuwang et al. in 2015 and used for spin transistors. The calculated values of activation energies of the Mg, Zn, and Ni substituted Co-Cu ferrite nanoparticles ferrite samples are shown in Fig. 12. The resistivity variation for the synthesized ferrite nanoparticles was according to Verwey and de Boer's hopping mechanism. Here, electron hopping occurs between Fe2+ ↔ Fe3+ ions during the sintering of the ferrites. If the ferrite's sintering temperature is higher, more Fe2+ ions are produced, thereby accelerating the hopping process. The hopping process is possible in M2+ ↔ M3+ and Cu3+ ↔ Cu2+ Existing together in a system, where M = Ni, Mg, and Zn.The Mg, Ni, and Zn substituted Co-Cu nano spinel ferrite were successfully synthesized with the sol-gel auto combustion method with the use of 1000 °C sintering temperature. X-ray diffraction measurements validated the mono-phase formation and nanocrystalline character of the produced samples. It possesses a monophasic cubic-spinel lattice structure with an average crystallite size ranging from 29.01 to 47.17 nm. The lattice constant determined by XRD analysis falls between 8.374 and 8.474 Å. FESEM micrographs showed grain sizes of 40, 50, and 74 nm. EDS has validated the composition of the resulting ferrite nanoparticles. The magnetic saturation (MS), magnetic moment (MR), and coercivity (Hc) are found to be the highest for the Ni substituted Co-Cu ferrite system. So, the Ni substituted Co-Cu nano ferrite is preferable for electromagnets and recording purposes than Mg and Zn substituted Co-Cu nano ferrites. The aspect ratio between Mr and Ms is 0.5 for all showing their ferromagnetic isotropic nature. The results of the DC resistivity of nanoparticles were increased with increasing temperature indicating their semiconducting nature. So, these ferrites are magnetic semiconductors. We can fabricate dilute magnetic semiconductors using transitional elements rather than electronic active elements. D. Parajuli: Conception and design, or analysis and interpretation of the data; drafting the article or revising it critically for important intellectual content; and Approval of the final version. Correspondence, Yonatan Mulushoa S.: Data curation, Investigation, Validation, Venkateswara Rao: Data curation, Investigation, Validation, N. Murali: Data curation, Investigation, Validation, Correspondence, K. Samatha: Approval of the final versionThe 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 sol-gel auto-combustion procedure was used to synthesize M (M = Mg, Ni, and Zn) substituted Co0.5Cu0.2M0.3Fe2O4 micro ferrites. X-ray diffraction measurements validated the mono-phase formation and nanocrystalline character of the produced samples. It possesses a monophasic cubic-spinel lattice structure with an average crystallite size ranging from 29.01 to 47.17 nm. The lattice constant determined by XRD analysis falls between 8.374 and 8.474 Å. The morphological characteristics were shown using FESEM analysis, which revealed the almost spherical form of grains with sizes ranging from 40, 50, and 74 nm. The compositional verification was undertaken with the help of the EDS analysis which ensured the presence of elements in desired proportions. IR spectra confirm their spinel spectra with the help of the two absorption bands lying between 1200 and 400 cm−1 resembling their two sites: tetrahedral and octahedral. The magnetic saturation (MS), magnetic moment (M)R, and coercivity (Hc) are found to be the highest for the Ni substituted Co-Cu ferrite system, i.e., MS = 49.5 emu/g. The ratio MR/MS for all ferrites is less than 0.5 indicating their ferromagnetic isotropic nature thereby acting as magnetic semiconductors. The electrical properties were measured using the ‘two probe technique’ with varying temperatures which indicates their semiconducting nature. Ni substituted Co-Cu nanoferrite is preferable for electromagnets and recording purposes the magnetic and electrical behavior of Ni substituted Co-Cu nanoferrite shows its relevance for recording and electromagnetic applications.

Data will be made available on request.Persistent Organic Pollutants.Metal-organic frameworks.persistent, bio-accumulative, and toxic.dichlorodiphenyltrichloroethane.polybrominated diphenyl ethers.Organochlorine.Persulfate.polychlorinated biphenyls.hexachlorocyclohexane.polyfluoroalkyl substances.brominated flame retardants.pharmaceutical and personal care products.United Nations Environmental Program.Canadian Environmental Protection Agency.advanced oxidation processes.Reactive Oxygen Species.metal-to-ligand charge transfer.ligand-to-ligand charge transfer.Water quality is of universal interest, and the UN Member States have acknowledged its importance in sustainable development in the Agenda for Sustainable Development. To be delivered by 2030, the Sustainable Development Goal (no.6) seeks to “improve water quality by reducing pollution, eliminating dumping and minimizing release of hazardous chemicals and materials, halving the proportion of untreated wastewater, and substantially increasing recycling and safe reuse globally” (Nations (UN), 2015). Water reclamation and reuse have equally gained widespread attention in recent years due to water scarcity caused by climate change and inadequate water resource management. Likewise, constant water quality degradation, resulting from ineffective wastewater management and treatment, has been observed to limit water provision in growing population countries and be an ever-increasing issue in an expanding global economy (Zendehbad et al., 2019). In response to water quality issues and associated limitations to clean water accessibility, one of the attractive solutions is wastewater reuse and reclamation to ensure sustainable water development and management. However, decades of industrial and agricultural products with growing urbanization and anthropogenic activities have resulted in multiple water pollutants, including persistent organic pollutants. POPs arise concerns that can be challenging as they could still be present in treated water.POPs are carbon-based chemical compounds resistant to environmental degradation and have been continually discharged into the environment. Due to their poor biodegradability, POPs can cause severe harm to wildlife and human beings. On closer inspection, POPs inhibit the immune system's natural reaction while also lowering the body's viral resistance (S. Zhang et al., 2015; L. Zhang et al., 2015). Besides, several studies have revealed that organisms exposed to POPs might lead to congenital malformations and reproductive disorders (Lee et al., 2014; Nadal et al., 2015; Tartu et al., 2015). Several POPs have been identified to be carcinogenic and potential endocrine-disrupting substances (Fång et al., 2015; Ong et al., 2018). Therefore, to ensure that water is devoid of POPs, advanced treatment solutions are crucial.Historical interest in the construction of organic/inorganic hybrid compounds dates back to the 1830 s, with a report on the first organometallic platinum species by Zeise (1830)). In Zeise’s report, the challenges with both synthesis and characterization of the later-called “Zeise’s Salt” (K[PtCl3(C2H4)]H2O) were detailed. This Pt complex indeed marked a remarkable triumph in experimental characterization; it initiated a whole new field of organometallic chemistry and, more broadly, the interest in reactivity occurring at the metal-organic interface.Over the following two centuries, various breakthroughs in chemical physics, quantum mechanics, and optics allowed advanced complex analytical methods that developed synthetic curiosities beyond compositional and structural elucidation to targeted application and function. However, similar to Zeise’s salt, most of the 1900s cutting-edge chemistries were zero-dimensional (or molecular), homogeneous compounds. Driven by industrial motives for heterogeneous materials and academic interests in forming more sophisticated multidimensional compounds, researchers started to investigate physical characteristics that could only appear by extending chemical connectivity into higher dimensions (1D, 2D, and 3D) in both amorphous and crystalline structures (Ozin, 1992).Intrinsic porosity emerged as both a goal and a result of greater dimensionality. Despite being purely inorganic, siliceous zeolite appeared to be a milestone that exhibited how novel chemical characteristics could be experienced by harnessing both the porosity and the capacity to attach heterogeneous catalytic sites into the scaffold (Breck and Breck, 1973; Davis, 1993). Still, the chemical compositions of the zeolitic structure type materials were chiefly limited to aluminosilicates, which are capable of accommodating only minor quantities of transition metals, mainly as defects (Hunger et al., 1987; Yamagishi et al., 1991). Progression to bulkier chalcogenides, bigger organic anions, and metal replacements beyond group IV elements resulted in novel and isostructural topologies with, until now, unseen chemical connectivity. With the later inclusion of organic bridging ligands, the multidimensional porous coordination complexes shaped their own distinctive family: metal-organic frameworks (MOFs).As with any artificial material, synthetic methods and compositions are vital for targeting specific properties (e.g., selectivity, pore aperture). Numerous attempts have been devoted to the synthesis of MOFs and applying MOFs for water treatment. MOF materials provide several paths to targeting different pollutants in aqueous environments ( Fig. 1). In this review, we explore the application of MOFs and MOF-based materials in removing POPs from water, focusing on the most recent advances.This review looks into the properties, impacts, and degradation methods of POPs found in water/wastewater. The focus of this study is adsorption and photocatalytic degradation. Aside from adsorption, which is one of the most essential methods for water purification by MOFs, improved oxidation techniques may be used to remove contaminants, converting these compounds into H2 and harvesting the chemical energy trapped in their bonds. Each method's merits, limitations, and process improvements are discussed in detail. To detect relevant findings and articles, key search terms such as persistent organic pollutants (POPs), metal-organic frameworks (MOFs), adsorptive removal of environmental contaminants, and photocatalytic degradation were utilized in several resource banks such as SciFinder, Elsevier, Springer Link, and Google Scholar.POPs are a group of natural or synthetic chemical compounds which enter the nature cycle either intentionally or unintentionally. These compounds have become a prevalent contaminant and are worldwide pollution due to their ability to travel large distances via air circulation (wind) and ocean currents (water), as well as stored in snow and ice (cryosphere), plants, sediments, and soil. Numerous documents have been accumulated to illustrate that regions far from the predominant POPs sources, like the Arctic and high mountains (the so-called grasshopper effect), have been widely contaminated by POPs (Ma et al., 2016). The rate at which POP particles disperse into the atmosphere is affected by temperature. When the temperature drops, POPs accumulate on the soil surface, while rising temperatures lead them to evaporate into the atmosphere. (X. C. Wang et al., 2016; T. Wang et al., 2016; X. Wang et al., 2016). POPs do not easily degrade into less harmful forms due to their stability and can thus persist in the environment for decades or centuries. Therefore, over time they are expected to accumulate, e.g. in fat-rich tissue and bio-concentrates (Ashraf, 2017). Exposure to POPs has been linked to endocrine disruption, cancer, neurobehavioral problems, reproductive and immunological malfunction, according to several studies. Rachel Carson established POPs as the cause of reproductive failure in birds exposed to dichlorodiphenyltrichloroethane for the first time (DDT) (Carson and Neill, 1962). Studies have shown that POPs may change female and male reproductive processes in mammals, birds, reptiles, fish, and snails and lead to deterrent strength to generate viable oocytes and create and retain a pregnancy (Buck Louis et al., 2011). For instance, clinical studies in humans have shown that polybrominated diphenyl ethers PBDEs may decrease semen quality (Abdelouahab et al., 2011), change reproductive testosterone and progesterone metabolism in both genders in humans (Gao et al., 2016; Makey et al., 2016; Meeker et al., 2009).To reduce the impact of POPs on living beings and the environment, a global treaty was signed in 2008 in Stockholm, Sweden, by more than 90 countries. The Stockholm Convention introduces 12 POPs as a ‘dirty dozen’ chemical that has destructive effects on the environment (B. J. Sun et al., 2020; B. Sun et al., 2020). Presently, there are 26 POPs; however, with urbanization and industrialization developing, the POPs are rapidly produced (Jeong et al., 2020; Mouly and Toms, 2016). These contaminant compounds have been identified into four categories listed in Table 1 (Ma et al., 2016), containing various compounds. First: agricultural POPs, including pesticides, insecticides, and herbicides such as Aldrin, Dieldrin, Chlordane, DDT, Endrin, Heptachlor, Mirex, Toxaphene, etc. However, Organochlorine (OC) pesticides, introduced in 1940–1950 and banned later, are still used in some countries to aid agricultural practice and industrial sectors (Wagner et al., 2021). For instance, many countries use DDT due to its low cost and high efficiency in anti-malarial activity (C. Wang et al., 2016; T. Wang et al., 2016; X. Wang et al., 2016). Second: industrial chemical and unintentional productions, including polychlorinated biphenyls (PCBs), hexachlorocyclohexane (HCB) which was used as a fungicide in the past, polyfluoroalkyl substances (PFASs), brominated flame retardants (BFRs), dibenzodioxins, and dibenzofurans. However, it is speculated that more will be evaluated in the future (Praetorius et al., 2012). Third, pharmaceutical and personal care products (PPCPs) like norfloxacin, carbamazepine, diclofenac, and ibuprofen. The last group contains organic dyes such as methylene blue and methyl orange.Persistence and half-life features in the individual medium, such as air, water, soil, and sediment, were identified as the crucial parameters to introduce POPs chemicals in the Stockholm Convention. For example, the United Nations Environmental Program (UNEP) and the Canadian Environmental Protection Agency (CEPA) identified 60 and 180 days of half-lives for chemical persistence in water and soil, respectively. The transformation half-life of a compound in the environment can involve direct or indirect photolysis, redox reactions, biodegradation, and hydrolysis processes that are strongly influenced by environmental factors such as temperature, salt, redox status, microorganism activity, and sunlight exposure. Thus, based on ecological or physiological conditions, the half-life of POPs may individually vary from a few weeks to many years (Bu et al., 2016). Since most POPs have been used for a variety of purposes, including agriculture and medicine, it is critical to investigate appropriate methods that are highly efficient and cost-effective for monitoring and controlling POPs in environments for degradation and preventing them from accumulating in the natural cycle. The food chain is a global concern.Literature reports various methods for removing POPs from the environment, including chlorination, ion exchange, neutralization, oxidation, filtration, membrane filtration, activated carbon, etc. However, these approaches are limited in their application by low effectiveness, high pesticide concentration requirements, high prices, and the formation of harmful by-products, which further complicates their removal (Adithya et al., 2021). The most prominent technologies are divided into three categories: physical, chemical, and biological (Ghoreishi and Haghighi, 2003) ( Fig. 2). The biological process, which includes bioremediation, is one of the most prevalent ways of POPs degradation and is frequently used to remove POPs due to the number of microorganisms and their capacity to function in severe environments. Many of these wastewater treatment technologies have not been widely used in industry due to their high cost and disposal issues. Most research has focused on chemical and mixed chemical–biological treatments to promote dye biodegradation and reduce sludge generation. As toxicity requirements grow more widespread, new approaches for reducing the content of POPs and their breakdown products in wastewater are required. These issues have prompted research into novel solutions.A very promising solution is the combination of photocatalysis and adsorption in water/wastewater treatment. Photocatalytic remediation of POPs offers a highly effective degradation process, simultaneous production of hydrogen and possibly other compounds as renewable solar fuels, while adsorption synergistically complements this with low cost, energy effectiveness, environmental friendliness, and ease of regenerating adsorbents.Any materials utilized in these procedures must meet specified requirements, such as stability of the adsorbents in reaction media, which is a critical issue in photocatalytic applications. Other aspects include thermal and mechanical stability, as well as their toxicity. Traditional materials for water purification, such as carbon compounds, zeolites, aluminosilicates, mesoporous materials, and others, have limited utility due to their low surface area, unsuitable electronic structure (either metallic or insulating), and limited options for chemical modifications (Pi et al., 2018). Most importantly, the materials need to have semiconducting properties with suitable band gaps to both ensure optimum light absorption and efficient redox activity of excited charge carriers (Dörr et al., 2018). So far, the most important photocatalysts for POPs removal from aqueous solutions include transition metal oxide (e.g. TiO2, Fe2O3) and porous organic polymer composite (Doll and Frimmel, 2005).While organic catalysts suffer from low stability and degradation, metal oxides are often limited by their relatively large band gap. For example, TiO2 has a band gap of 3.2 eV for anatase (Eder et al., 2010, 2009, p. 2) that limits light absorption to near UV range. It also suffers from fast e-/h+ recombination. Nevertheless, while at the moment, TiO2 still represents the most used semiconductor for photocatalytic applications, we need to develop catalysts that combine suitable photocatalytic and adsorption properties under solar light irradiation.Photocatalysts of first-generation are made up of single-component materials (e.g., WO3 and ZnS), while second-generation photocatalysts consist of multiple components in a suspension (e.g., RGO/CdS and g-C3N4/BiOI). Among the most promising candidates are metal-organic frameworks (MOFs). Photocatalysis based on MOFs can be considered third-generation photocatalysts immobilized on solid substrates (Hayati et al., 2021).MOFs are a class of porous crystalline material consisting of metal ions or clusters and organic linkers (Kitagawa, 2014). Yaghi and Li, (1995) were the first to synthesis these compounds, which were then studied for a variety of uses as well as wastewater treatment. MOFs are made up of metal ions or metal-oxo clusters (called secondary building blocks, SBUs) that are connected by organic linkers into highly porous crystalline networks ( Fig. 3). A variety of pore sizes and shape flexibility, tunable functionality, decent thermal stability, and record surface areas render MOFs ideal candidates for the removal of a variety of pollutants. MOFs mainly synthesis via solvothermal routes. This synthesis method became the most popular way to obtain grams of MOFs in laboratories. In this process, inorganic salt and organic linker solutions are mixed in a sealed reactor vessel and heated to stimulate the formation of insoluble frameworks that precipitate as fine crystals. Moreover, synthesis methods such as non-solvothermal processes include: conventional electric (CE) heating, microwave (MW) heating, electrochemistry (EC), mechanochemistry (MC), and ultrasonic (US) techniques are also of interest in the synthesis of some MOFs (Stock and Biswas, 2012). The possible compositions and structures of MOFs are nearly infinite. MOFs can be functionalized at the organic or inorganic linker, or catalytic units can be accommodated in their pore space to generate catalytic activity ( Fig. 4). MOFs large surface area and pore volume enable active guest species to be introduced into the pores/cages/channels and facilitate access of substrates to the internal active sites. The active sites of MOF catalysts can be categorized as follows: (i) coordinatively unsaturated metal (CUM) centers and functional linkers, (ii) functional groups attached to the linkers and/or metal centers by direct synthesis or by post-synthesis modification (PSM), and (iii) active guest species such as metal nanoparticles (MNPs), complexes, and polyoxometallates (POMs) encapsulated in the pores (Y.-B. J. Huang et al., 2017; Y.-B. Huang et al., 2017; Gascon, 2013).Photocatalysis involves redox reactions induced by the light energy that typically occurs on the surface of a semiconductor (Zango et al., 2020b). Among the various reactions, advanced oxidation processes (AOPs) or pollutants have been intensely investigated in the last decades. The Key is the utilization of photoexcited holes that can directly oxidize organic compounds. Alternatively, these holes, but also photoexcited electrons can transform water into reactive hydroxyl radicals and superoxide species that, in turn, can drive the desired degradation reaction. In addition to faster degradations, the advantages of photocatalysis in the removal of pollutants also include operation under ambient conditions, tunable selectivity towards the desired oxidation products, and utilization of non-reactive reactants, such as water and CO2. (Russo et al., 2020). The catalytic performance of a semiconductor is guided by an optimum band gap that allows to absorb of a large portion of light photons and provides long-living free charge carriers with suitable potential energies (defined by the HOMO-LUMO levels for holes and electrons, respectively) to drive the desired reactions. Effective exciton separation and enhanced charge transport properties, as well as direct reactant access to a large number of active sites with suitable adsorption chemistries without kinetic limitations (e.g. by reactant diffusion) are also crucial to maximize photocatalytic efficiency.MOFs have the potential to fulfill all those requirements. In addition to an unparalleled number of accessible active sites, the tunability of both SBUs and organic linkers, offers a powerful tool to adjust the chemistry of active sites (e.g. Lewis acidity) as well as the optoelectronic properties (i.e. band gap, charge absorption, and transfer pathways) (Choi et al., 2009; L. Zhang et al., 2015; S. Zhang et al., 2015).With the development of water/acid-resistant MOF materials in recent years (T. Wang et al., 2016; X. Wang et al., 2016; C. Wang et al., 2016), an increasing number of light-responsive MOFs for photocatalytic pollution removal have been reported.In the last years, reviews have been published on various aspects of MOFs in photocatalytic environmental applications (Wang et al., 2020). The majority of photoactive MOFs were used to degrade pollutants in water using UV light. Thus, the stability of MOFs in both aqueous solutions and UV irradiation becomes a crucial parameter and has to be considered for each reaction and set of process conditions.Chemical stability refers to MOFs’ durability under various chemical conditions, including moisture, solvents, acidic, basic, or aqueous solutions containing coordinating anions. On the other hand, thermal and mechanical stability refers to MOFs' capacity to maintain structural integrity when exposed to heat, pressure, or vacuum treatment conditions. MOFs generally have good thermal stability, and some MOFs can even be stable at 500 °C (Cavka et al., 2008). The majority of known MOFs, on the contrary, have low stability in aqueous environments. This has become a significant disadvantage that has hampered the actual application of MOFs (Ding et al., 2019). Water could break the coordination and displace the attached ligand, block the binding site, prevent other target molecules from adsorbing, and potentially lead to a collapse of the MOF structure (K. F. Tan et al., 2015; K. Tan et al., 2015). Based on Low et al. research (Low et al., 2009), there are two possibilities for degradation mechanisms of MOFs in water vapor and liquid water: (1) ligand substitution by water and (2) hydrolysis. These mechanisms were established by computer simulations. Some of the most water-stable MOFs reported are listed in ( Fig. 5). It appears that the stability of most of the MOFs remains limited to just a few hours to days at most.In the majority of cases, water molecules break up the coordination between SBUs and the organic linkers. At first, water substitutes the organic linker and forms aqua ligands that often undergo H-bond formation with the former linkers. (Eq. (1)). (1) M−L + H2O→M−(H2O)⋯L This may be followed by deprotonation to yield hydroxo ligands without interactions with the former linkers, thus resulting in a structural collapse (F. Tan et al., 2015; K. Tan et al., 2015). (Eq. (2)). (2) M−L + H2O→M−(OH) + LH Experimental studies have confirmed that adding hydrophobic functional groups to the organic ligands can improve the water stability of MOFs. For example, by combining several water-repellent functional groups in the frameworks, Wu et al. were able to create a novel MOF with water resistance in boiling water for up to one week(Wu et al., 2010). The downside of this process, however, is that this modification also changed the adsorption chemistry as well as the optoelectronic properties.Another need for an effective photocatalyst, as previously noted, is stability under irradiation conditions since otherwise, owing to degradation of the photocatalytic sites, its activity would diminish with time. Several publications indicate that MOFs utilized as photocatalysts, particularly those based on terephthalic acid (BDC), are stable under irradiation conditions (Dhakshinamoorthy et al., 2018; Jing et al., 2017; P. W. Wu et al., 2017; P. Wu et al., 2017). However, this conclusion is mainly based on relatively short time experiments, typically only a few hours. However, some of the most effective photocatalytic MOFs, such as UiO-66, MIL-101, and MIL-125, include aromatic carboxylates as linkers, and several of these compounds have been demonstrated to be photolabile (Mateo et al., 2019).The Photocatalysis process enables the degradation of refractory organic compounds into by-products by the in-situ generation of reactive oxygen species (ROS), such as superoxide (•O2 -), hydroperoxyl (HO2•), alkoxyl (RO•), sulfate (SO4•−) and chlorine (Cl•) radicals (depending on the catalyst or the oxidant used). Hydroxy radicals (•OH), which are the most appealing among the others, are usually generated from reactions involving oxidants such as hydrogen peroxide, ozone, or catalysts including metal ions and semiconductors under UV-Vis irradiation or other sources of energy. Depending on the catalysts phase, different AOP methods can be categorized as heterogeneous and homogenous photocatalysis (Antonopoulou et al., 2021; Bedia et al., 2019; Russo et al., 2020; Zango et al., 2020b).In heterogeneous photocatalysis, ROS are produced after charge separation caused by semiconductor irradiation (photocatalyst) (Bedia et al., 2019). Due to their existence in solid form, heterogeneous catalysts are very easy to separate from solution. However, the initial rate (turnover frequency) in heterogeneous photocatalysis is very low as the active sites are not well-defined. Hence product conversion is not uniform and leads to secondary intermediates, products. In comparison, homogenous processes have limitations in terms of catalyst loss and non-recovery (Lu et al., 2021). In addition, the catalyst deactivation is quite fast. Therefore, compared to homogeneous processes, heterogeneous photodegradation is an interesting option for wastewater treatment since the catalyst can be separated from the reaction media and reused, reducing costs and environmental issues (Bedia et al., 2019). However, the accessibility of reactive sites is an important factor. Diffusion of reactants and products to and from the catalytically active sites is often the rate-limiting step (Gascon, 2013).An essential criterion in the choice of MOFs for photocatalytic applications is the ability of the MOFs to harvest and channel light energy (Zango et al., 2020b). The high flexibility of the MOFs framework allows elaborate design and tailoring of their structure to enhance photocatalytic activity (Y. X. Li et al., 2016; Y. Li et al., 2016; Wen et al., 2021a, 2021b; Wu et al., 2019; Younis et al., 2020). Some MOFs containing Fe, Cr, Zr, and Ti metal ions can harvest and channel solar energy. They usually possess a small bandgap that enables visible light excitation; hence, they are considered promising candidates for photocatalytic degradation of organic pollutants (Pi et al., 2018; Zango et al., 2020b). The presence of organic linkers in MOFs allows them to have a tunable absorption spectrum and an efficient charge separation enabling lifetimes in the microsecond range. However, a few MOFs have been reported showing photocatalytic activities under visible light. Heterogeneous photocatalysis principal is illustrated in Fig. 6.A photodegradation reaction is initiated when a catalyst absorbs energy equal to its band gap energy (Bedia et al., 2019; Russo et al., 2020; Younis et al., 2020). After photoexcitation, and generation of electron (e-)-hole (h+), (Eq. (3)), electron transfer takes place from the catalyst surface to the adsorbed acceptor (A) molecules and from the adsorbed donor (D) molecules to the catalyst, (Eqs. (4) and (5)): (3) Photoexcitation : M O F ⟶ h υ e − + h + (4) Reduction of acceptor : A + e − → A − (5) Oxidation of donor : D + h + → D + These chain reactions continue until final oxidation products are formed, while recombination of electron-hole leads to photoelectric energy dispersion (Eq. (6)). (6) Termination reaction : e − + h − → N + E where E is the energy released in the form of heat or light and N is the neutral center resulting in a reduction of the photoexcitation process efficiency (Russo et al., 2020).The electron e--h+ charge transition can be described by four mechanisms ( Fig. 7): (1) ligand-to-metal charge transfer (LMCT), (2) metal-to-ligand charge transfer (MLCT), (3) metal-to-metal charge transfer, node-localized excitation, (MMCT), (4) ligand-to-ligand charge transfer, linker-localized excitation, (LLCT) (Wu et al., 2019). The LMCT is the most effective charge transfer mechanism to prevent e--h+ recombination because MOFs can achieve better e--h+ separation via ligand-to-metal charge transfer (Y. X. Li et al., 2016; Y. Li et al., 2016; Wen et al., 2021a, 2021b; Wu et al., 2019; Younis et al., 2020). The LLCT is generally rare because the linkers in the MOF structure are separated by metal clusters (Wen et al., 2021a). The LMCT mechanism's efficiency depends on the energy required to transfer photo-generated electrons from the LUMO of ligands to the LUMO of metal nodes, ELMCT (Y. X. Li et al., 2016; Y. Li et al., 2016; Wen et al., 2021a, 2021b; Wu et al., 2019; Younis et al., 2020). Also, Eg and Eabs are other fundamental parameters that affect LMCT. Eg is the band gap between HOMO and LUMO of ligands. Eabs is absorption energy, the energy required to excite the semiconductor (Wen et al., 2021b). The Eabs is the sum of Eg and ELMCT. Therefore, for a constant Eg, the zero or negative values of ELMCT lead to the smaller values of Eabs that favor the absorption of visible light. MOFs composed of ligands with high-energy lone pairs and metals with low-lying empty orbitals are more desirable due to the favorable ELMCT (Y. X. Li et al., 2016; Y. Li et al., 2016; Wen et al., 2021a, 2021b; Wu et al., 2019; Younis et al., 2020). Alvaro et al. (2007) reported on the semiconducting characteristics of MOF-5 as an early example of MOFs as photocatalysts. They synthesized MOF-5 by precipitating a mixture of two DMF solutions with triethylamine at room temperature. MOF-5 material contains clusters of Zn4O located at the corners of the structure that are connected orthogonally to six units of terephthalate. According to the author's findings, absorption of solar irradiation (from 350 to 400 nm) by the MOF-5 framework will produce charge separation and generate photoexcited electrons. The author tested the activity of MOF-5 in the photodegradation reaction of phenol in comparison with Degussa P25 (TiO2 nanoparticles) and another semiconductor, ZnO. The result revealed that MOF-5 behaves as a semiconductor and undergoes charge separation (electrons and holes). The remarkable photocatalytic activity of MOF-5, compared with that of ZnO and TiO2, also confirmed that MOF-5 degraded under catalytic conditions, forming ZnO. Hence, these ZnO nanoparticles could be responsible for the catalytic process (Alvaro et al., 2007).When dyes are released into the environment, they frequently pose severe ecological hazards, such as harming aquatic life, impeding plant growth, and posing various forms of toxicity to humans, such as genotoxicity, reproductive toxicity neurotoxicity, and other forms of the disease (Zango et al., 2020b).The majority of photocatalytic degradation studies using MOFs were focused on the photodegradation of organic dyes. Numerous MOFs have been used for the photodegradation of dyes, and some relevant studies are summarized in Table 2. As can be concluded from the literature review, Methylene blue (MB, cationic dye), Methyl orange (MO, anionic dye), and Rhodamine B (RhB) have been studied as target pollutants in most articles mainly because they are easy to analyze. However, the number of researchers who examined sunlight for dye removal is rare. The sunlight absorption capacity of pure MOFs is low; therefore, a lower rate of conversion and removal can be achieved when using sunlight in comparison with UV light sources. The studies focused on the synthesis and application of MOF composites instead of pure MOFs to enhance light harvest capacity. The results of the literature review revealed that the mechanism for photodegradation of organic dyes is the formation of electron (e-)-hole (h+) pair in MOF structure under a light source as the initial step of the photocatalytic dye degradation process. After absorption of energy by the MOF, the e- was excited from the valence band (VB) and entered into the conduction band (CB), leaving the h+ in the VB. Cd (II)-imidazole MOFs, used under UV light for the MB and MO degradation demonstrated that the Cd (II) complexes formed with the linkers were excited by UV light, generating the photogenerated charges that are required for dyes degradation (B. Liu et al., 2014; L. Liu et al., 2014). Zn (II)-imidazolate MOF, named ZIF-8, also revealed a high activity for MB degradation under UV light due to the excitation of the Zn-complexes (Zhang et al., 2016). Different Fe-based MOFs have been tested for the photodegradation of dyes under visible light. Among them, MIL-88B (based on terephthalate linkers) showed higher activity for R6G degradation than other MIL-100 and MIL-101 MOFs, which was associated with the visible light absorption of the Fe-oxo-clusters that formed the structure so as the photogenerated electron is transferred to Fe3+ from O2 - (Laurier et al., 2013).The transformation of traditional into modern agricultural practice requires insecticides, herbicides, pesticides, and fertilizers. The discharge of increasing agrochemicals to water sources poses potential hazards to the ecosystem (Mon et al., 2018; Zango et al., 2020b; Wen et al., 2021a; Zendehbad et al., 2022). Pesticides contain various chemicals such as herbicides, insecticides, and fungicides and are commonly applied in agriculture (Khodkar et al., 2019; Wen et al., 2021a). Herbicides are primarily produced to inhibit weeds that compete with the plant’s growth (Oladipo, 2018; Zango et al., 2020b). At the same time, insecticides are aimed at repelling or mitigating insects and other pests from attacking agricultural products (Zango et al., 2020b). Herbicides account for about 62% of the total pesticides used in agriculture in the United States. During the past two decades, the five most-used herbicides were glyphosate, atrazine, metolachlor-(S), 2,4-D, and acetochlor (Wen et al., 2021a). Traditional materials such as TiO2, WO3 and their modified derivatives have shown good performance on the photocatalytic degradation of pesticides like diazinon, acephate, omethoate, (4-nitrophenol (4-NP) and methyl parathion herbicides like isoproturon and triazine, etc. (Pi et al., 2018). However, the number of literatures on photocatalytic degradation of these agricultural materials using MOFs is rare. The most recent literature related to herbicides and pesticides is reviewed here.The photodegradation of 4-nitrophenol (4-NP) was investigated by (Samuel et al., 2018) using [Zn(BDC)(DMF)] crystal which was synthesized via ultrasonic irradiation and solvothermal method. The synthesized MOF [Zn(BDC)(DMF)] exhibited high photocatalytic activity in the presence of NaBH4 under natural sunlight irradiation, and the reduction of 4-NP to 4-aminophenol (4-AP) was completed within 10 min. The catalyst showed higher catalytic activity even after ten cycles, with an efficiency above 95%. The 4-NP degradation reaction data were fitted to the first-order kinetic plot with a rate constant of 0.6008 min−1. After irradiation and generation of electron-hole pairs, electron transfer takes place from BH4 - ion to 4-NP. In the presence of MOF[Zn(BDC)(DMF)], BH4 - ion gets adsorbed on its surface, and discharge of electron from BH4 - ion takes place through the metal oxide to the acceptor 4-NP. The aqueous medium consists of H+ ions for complete reduction of 4-NP to 4-AP.Photocatalytic removal of 4-nitrophenol (4-NP) under visible LED light irradiation was investigated by using a MOF/CuWO4 composite system which was synthesized by (Ramezanalizadeh and Manteghi, 2018). [CoNi(μ3-tp)2(μ2-pyz)2] MOF was prepared with the utilization of hydrothermal approaches. The comparison between PL spectrum of the pure CuWO4, pure MOF, and MOF/CuWO4 composite structures demonstrates that the MOF/CuWO4 composite has the minimum recombination of electron-hole pairs. This behavior leads to a high lifetime for the charge carrier’s species and increases the photocatalytic performance. Photocatalytic destruction in the presence of pure MOF was found to be 24%, while photodegradation efficiency was enhanced to 81% in the presence of MOF/CuWO4 (1:1) composite. There was no data on the kinetic of photodegradation 4-NP using the as-synthesized composite.Motivated by the phenomenon of utilizing two types of the organic linker and assisting in extending the lifetime of charge carriers, (Surib et al., 2018) constructed a Cu-MOF photocatalyst They employed H4btec (1,2,4,5-benzenetetracarboxylic acid) and 1,4-bimb (1,4-bis(imidazole-1- methylbenzene)) as ligands, and the framework was constructed through a facile and simple hydrothermal route. Photocatalytic degradation of 2-chlorophenol was examined by undertaking solar light. After four hours of irradiation, 100% photodegradation for this pollutant was achieved and photodegradation of 2-chlorophenol under solar light was fitted to the first-order kinetic. The light activation excited the electron from HOMO to LUMO. The excited electrons at LUMO are utilized by the surface adsorbed O2 and converted to reactive •O2 - state and in turn to •OH. Meanwhile, the removal of one e- from HOMO has altered the stability of the molecule that, forces the HOMO to react with the same surface adsorbed water molecule and convert it into active •OH. This chain reaction continues until the irradiation is terminated, resulting in the generation of a higher concentration of ROS. These reactive species continuous the reaction further with the available substrate (2-CP) to its simplest nontoxic form. Ramezanalizadeh and Manteghi (2017) prepared a photocatalyst by immobilization of mixed cobalt/nickel metal-organic framework on a magnetic BiFeO3. The 2.18 eV band gap of BiFeO3 makes it attractive to visible light harvesting. However, due to the fast electron-hole pairs regeneration, it needs to be combined in the composite form with a foreign high surface area structure. The photocatalytic performance of the prepared samples was evaluated for the photodegradation of 4-NP. The visible light source was a 5 W LED lamp (λ = 440 nm). With the aid of visible-light irradiation, MOF and BiFeO3 absorb the photons and produce the electron-hole pairs. Afterward, the photogenerated charge carriers are separated under the influence of internal electrostatic fields in the heterojunction areas. As a result, the regeneration of the electron-hole pairs in the MOF/BiFeO3 (1:1) effectively decreased, and this trend leads to higher photocatalytic performance for this system. The degradation of about 70% was obtained by MOF/BiFeO3 (1:1) as shown in Fig. 8. The data on kinetic of photodegradation 4-NP using the as-synthesized composite was not available in their study.Removal of diazinon and atrazine pesticides under visible light illumination with a high-pressure mercury-vapor lamp (400 W and λ = 546.8 nm) was studied by Fakhri et al. (2020). They synthesized a core-shell structured magnetic graphene oxide@MIL-101(Fe) which acts as a photo-Fenton catalyst to degrade the mentioned pesticides. After 105 min irradiation, the photodegradation efficiency was about 100% and 81% for diazinon and atrazine, respectively. However, TOC analysis reveals that the mineralization of diazinon and atrazine was approximately 84.0% and 62.0%, respectively. The difference between mineralization and photocatalytic efficiency is due to the presence of intermediates during photodegradation, which allows the ongoing process to mineralize for a longer time. According to this study, the Fenton reaction was initiated through the photo-reduction of Fe3+ to Fe2+ in the MIL-101 (Fe) component. Thus, the photo-produced electrons transfer to the cores of Fe3O4 through graphene oxide (GO) layers. GO assists in increasing the transformation of Fe3+ to Fe2+, which is ideal for producing more OH radicals. The photodegradation process occurs by the attack of these active radicals on pesticides’ structure and subsequent transformation into H2O and CO2.Recently, Hayati et al. (2021) prepared a novel Ag (I) coordination complex [Ag(p-OH-C6H4COO)2(NO3)]n (1) using the laying as well as sonochemical irradiation methods (Hayati et al., 2021). The synthesized catalyst was used for simultaneous photodegradation of 2,4-D and MCPA under sunlight irradiation. These herbicides belong to the phenoxy acid herbicide group. The decline in TOC (total organic content) of irradiated samples confirms the mineralization of the herbicides to CO2 and H2O. The TOC was reduced from 2.20 to 0.10 mg/L with the irradiation of sunlight. The synthesized photocatalyst's FOM (figure of merit) was the best among other photocatalysts, such as TiO2, BiVO4/Ag3VO4, FTO/BiOBr, and Steel/TiO2–WO3. The mechanism of photodegradation of 2,4-D is not discussed in their study. Khodkar et al., (2018) prepared a magnetic photocatalyst α-Fe2O3 @MIL-101 (Cr)@TiO2 via the sol-gel method to remove paraquat from an aqueous solution . A 125 W medium UVC lamp was used as a light source. The response surface methodology, RMS, was applied to optimize the effective parameters such as contact time, pH, catalyst dosage, and herbicide concentration. They utilized potassium dichromate analysis for measuring COD to determine the number of organic compounds before and after photodegradation. The catalyst dosage and paraquat concentration significantly affected photocatalytic degradation, while pH and contact times were not significant factors. The maximum photocatalytic degradation and COD reduction were achieved at 87.46% and 90.09% at optimal conditions with paraquat concentration of 20 mg/L, catalyst dosage of 0.2 gL−1, pH 7, and contact time of 45 min. The major role of the improved photocatalytic elimination of paraquat was due to an increase in the total active surface area of the catalyst. Their data for photodegradation was well fitted with pseudo-second-order kinetic (R2 = 0.9986 and k = 0.0053 g mg−1 min−1). The photodegradation mechanism was not discussed in their research. Mei et al. (2019) modified MIL-53 (Fe) by regulating the electronegativity and BET surface area of Fe-O clusters in MIL-53 (Fe) (Mei et al., 2019). The synthesized MOF was used for photocatalytic removal of Thiamethoxam (TMX) under the LED light source. They confirmed that the excellent photo-Fenton-like catalytic activity could be obtained through modifying MOF by Xylitol and D-sorbitol ( Fig. 9). Their findings revealed that the Fe2 + and photogenerated electrons in the system respectively reduce persulfate (PS) and O2 to produce the SO4•− radicals and the •O2 - radicals, similar to those observed by homogeneous Fe-containing complexes by Fenton-like reaction. Activation of as-synthesized MOF by visible light initiates charge separation, in which electrons are captured by Fe3+ to form •O2 - radicals and Fe2+. The as-formed Fe2+ and the original Fe2+ can reduce PS to produce SO4•− radicals, and the Fe2+ is oxidized back to Fe3+ by PS. The more electrons are consumed by Fe3+, accelerating the separation of electrons (e-) and holes (h+) pairs effectively, the more holes (h+) corresponding to it are in a free state. And the free holes (h+) and other radicals (•O2 -, SO4•−, •OH) are united together, contributing to the degradation of TMX.Taking advantage of the catalytic properties of MOFs and fast carrier mobility of 2D materials, MIL-125 (Ti)/BP as the first 2D/MOF photocatalyst for diazinon removal from water under UV light was synthesized by (Hlophe and Dlamini, 2021). The removal mechanism of diazinon was based on Ti3+-Ti4+ intervalence electron transfer (Wang et al., 2021), which generates electron-hole pairs as shown in Fig. 10. The composites were prepared by varying the weight loadings of MIL-125 (Ti) while keeping the amount of FLBP constant and denoted 4%BpMIL, 6%BpMIL, and 12%BpMIL (m/m). According to their findings, the 4%BpMIL at neutral pH had the best removal efficiency (96%) after 30 min irradiation.Other studies on the photodegradation of agricultural products and their main results have been summarized in Table 3. As can be concluded, the number of agricultural products studied has been limited to several special chemicals. Also, the single component photocatalyst suffers from quick charge recombination and limited activity. Therefore, the research in recent years focused on constructing heterostructure MOF composites.Zr-based MOFs (i.e. UiO-66) and other MOFs such as MIL-53 (Fe) and MIL-100 (Fe) have high chemical and thermal stability in water (Ahmad et al., 2019). Fe-based MOFs are found to be a better choice compared to Zr-based MOFs for efficient photo-Fenton activity owing to their Fe-metal clusters. However, weak visible light response remains an unsolved issue when using MOFs as photocatalysts. Therefore, a combination of MOFs with materials such as BiOBr (Xue et al., 2018) and WO3/graphene oxide (Fakhri and Bagheri, 2020) is possibly an effective way to enhance photocatalysis efficiency. Synergistic effect of these materials in combination with MOF improves the photocatalytic activity. One example is the nanocomposite WO3/graphene oxide, where the conductive band (CB) of UiO-66 (−0.53 eV vs. NHE) is more negative than the CB of the W sample (+0.4 eV vs. NHE). It appears that after visible light irradiation, the photoexcited electron move from UiO-66 to CB of W, and holes move from the valence band (VB) of UiO-66 (+3.5 eV vs. NHE) to VB of W (+3.2 eV vs. NHE). Consequently, graphene oxide acts as an acceptor for photogenerated electrons, thereby limiting recombination of charge carriers. Furthermore, the •O2 - and •OH radicals can be easily formed by the photo-excited electrons and holes. These active radicals subsequently attack pollutant molecules that are adsorbed on the photocatalyst surface, and oxidize them to H2O and CO2.The different materials are widely used in making useful industrial products such as petrochemicals and plastics. Material such as phenol and its derivatives are used as a precursor in pharmaceuticals, dyes, herbicides, pesticides, detergents, epoxy resins, and polycarbonate plastics.Polycyclic aromatic hydrocarbons (PAHs) comprise benzene ring molecules, forming highly stable structures (Zango et al., 2020b, 2020a, 2020c). They can originate from natural processes (e.g., volcanic eruptions, bush fires) (Zango et al., 2020c) and more from oil exploration, exportation, effluents discharged from petroleum refineries and petrochemical industries, oil spillage, etc. (Zango et al., 2020b, 2020a, 2020c).Perfluorinated compounds (PFCs), which are called ‘’forever chemicals’’ are another group of POPs that have been widely used for the production of industrial (e.g., surfactants) and consumer products (e.g., non-stick coatings) (Zango et al., 2020b). These chemicals are composed of a fluorinated carbon backbone terminated by a carboxylate functional group (Chen et al., 2016). The most toxic of these groups in the perfluoroalkyl carboxylates (PFCAs) and perfluoroalkyl sulfonates (PFAS) (Zango et al., 2020b).In recent years MOF based materials have gained attention as an excellent candidate for treating industrial residues. However, to the best of our knowledge, there is no report on the application of MOFs in the photocatalytic removal of PAHs and PFASs. Few papers have been published on the adsorption removal of these chemicals. Because of their lipophilic character, they resist photo- and bio-degradation (Zango et al., 2020d). Most studies related to industrial pollutant removal are limited to photocatalytic removal of phenol and its derivatives which are discussed in this section. Ahmad et al. (2019) synthesized MIL-100 (Fe) loaded with ZnO nanoparticles using the solvothermal technique (Ahmad et al., 2019). The photo-Fenton activity of as-prepared M.MIL-100 (Fe)/ZnO NS hybrid has been evaluated degradation of phenol and bisphenol-A under LSH-500 W Xe arc lamp irradiation. A certain amount of H2O2 was added to the mixture of catalyst and pollutants to increase photo-Fenton activity. The photodegradation pollutants under various reaction conditions like initial pH value and H2O2 concentration were investigated. Their analysis confirms ligands–to-clusters charge transfer (LCCT) mechanism, suggesting the strong bonding of ligands oxygen with Fe (III) atoms. The synthesized photocatalyst shows much lower photoluminescence (PL) intensities compared to MIL-100 (Fe), and MIL-100 (Fe), which are associated with recombination inhibition and a longer lifetime of photoinduced charge carriers.Taking advantage of the sulfate radical (SO4•−)-based AOPs (SR-AOPs), Lv et al. (2020) prepared two novel MOFs/COFs hybrid materials with nitrogen-rich building blocks for the first time (Lv et al., 2020). The pristine MOFS were MIL-101-NH2 and UiO-66-NH2. The photocatalyst was fabricated by a feasible step-by-step assembling method and used for the photodegradation of BPA (bisphenol-A). The possible photocatalytic mechanism was explored through radical species quenching experiments. Their results also reveal that the composite photocatalyst displayed a lower PL intensity, implying the recombination of electron-hole pairs was efficiently inhibited. In the case of MIL-101-NH2, results suggested that the h+ and •OH were responsible for the degradation process of BPA, while h+ was involved in photocatalytic degradation over UiO-66-NH2.Another study by Lin et al. (2020) confirmed the synergic effect between MIL-88 (Fe) and PS for the photodegradation of BPA under visible light (Lin et al., 2020). The results revealed efficient charge carrier separation ability in the presence of PS, which could scavenge conduction band electrons and prevent the recombination of electrons and photo holes. The reaction of photogenerated electrons with PS can result in abundant production of SO4 - as shown in Fig. 11. In addition, the excitation of PS could also produce other active species, such as O2 - and OH radicals.To enhance the Fenton-like performance of MIL-88B-Fe, (S. Zhang et al., 2019; X. Zhang et al., 2019; Y. Zhang et al., 2019; H. Zhang et al., 2019) studied the incorporation of carbon nanotubes (CNTs) into the MOF structure. CNTs, with electron-rich oxygen-functional groups on the surface, facilitate the reduction of Fe (III) to Fe (II) to initiate the Fenton reaction. Phenol, BPA, and 2,4-dichlorophenol were entirely removed within 30 min using this photocatalyst. Also, in the presence of H2O2, the coordinatively unsaturated metal sites in the MIL-88B-Fe, which behaved as Lewis acid sites, are prone to adsorb the H2O2 molecules (a Lewis base). As a result, the generated Fe (II) directly reacted with H2O2 to form •OH, attacking pollutant molecules. C. Wang et al. (2019); Q. Wang et al. (2019); H. Wang et al. (2019) investigated the coupling of NH2-MIL-125 (Ti) with CdS to photo-degrade organic pollutants such as BPA under visible light (H. C. Wang et al., 2019; Q. Wang et al., 2019; H. Wang et al., 2019). The synthesized composites showed higher photo-degradation activities compared with pristine CdS and NH2-MIL-125 (Ti) nanocrystals. Following other research, the results confirmed a lower recombination rate of photo-generated electrons and holes in synthesized photocatalysts rather than pristine nanocrystals.The NH2-MIL-125 (Ti)/Bi2MoO6 core-shell heterojunction was synthesized by H. Zhang et al. (2019); X. Zhang et al. (2019); Y. Zhang et al. (2019); S. Zhang et al. (2019) to accelerate the transformation of photogenerated electrons and to improve the photocatalytic performance of pure Bi2MoO6 and NH2-MIL-125 (Ti) (S. H. Zhang et al., 2019; X. Zhang et al., 2019; Y. Zhang et al., 2019; S. Zhang et al., 2019). The electrons are transferred from the organic linkers and Bi2MoO6 to the Ti-oxo clusters of NH2-MIL-125 (Ti). Generally, they suggest that three features can improve the photocatalytic performance of the as-synthesized structure. The formation of heterojunctions leads to the rapid transfer of electrons and holes. The pores of NH2-MIL-125 (Ti) inside the heterojunction can provide a channel for electron transfer and extend the lifetime of an electron and help to scatter and reflect the light. Also, the presence of surface defects and smaller contact angles can effectively enhance photocatalytic ability.The increasing consumption of pharmaceutical drugs, cosmetics, and household chemicals in modern societies releases enormous amounts into the ecosystem (Mon et al., 2018). Photocatalytic removal of antibiotics which belong to pharmaceutical groups has been widely investigated. MOFs of the MIL family, such as MIL-100 (Fe), MIL-101 (Fe), and MIL-53 (Fe), are used as photocatalysts for tetracycline (TC) removal. (Jiang et al., 2019) prepared a MIL-53 (Fe) using a solvothermal method. Photocatalytic activity of as-synthesized photocatalyst was evaluated by TC removal under visible light. A 300 W Xe lamp was used as a light source. They also investigated the effect of the HCl modulator on the functional group of MIL-53. Their findings reveal that more uncoordinated ligands are present in the MOF structure due to the absence of the metal clusters dissolved in HCl. Compared with MIL-53, the photocatalytic activity of acid-regulated MIL-53 increased by 1.5 times due to the easier separation of photogenerated carriers.Photocatalytic activity of three Fe-based MOFs, including Fe-MIL-101, Fe-MIL-100, and Fe-MIL-53, was studied by Wang et al. (2018) through TC removal under visible light (Wang et al., 2018). Their results show that Fe-MIL-101 has much higher activity than the other two kinds of Fe-MILs under visible light exposure. The degradation rate was not very fast in the first hour under irradiation. However, 96.6% of the total TC was removed after three hours of irradiation. Fe-MIL-101 also exhibited high stability and recyclability for the photocatalytic degradation of tetracycline. Extensive iron oxo clusters cross-linked by ligands make this MOF responsive to visible light. As per their results, the photo-excited electrons on the valence band (VB) of Fe-MILs move to their conduction bands (CB) under visible-light irradiation. Consequently, the formed active species such as •O2 -, •OH, and H+ could effectively degrade the tetracycline into small intermediates or end products directly.In another study, Rasheed et al. (2018) investigated the effect of using carbon aerogel (CA) in MIL-Fe structure through TC removal (Rasheed et al., 2018). The synthesized MIL-100 (Fe) was combined with Fe3O4 and CA. The performance of ternary MIL-100 (Fe)@Fe3O4/CA photocatalysts was compared to the CA, Fe3O4, Fe3O4/CA, and MIL-100 (Fe)/CA. According to their results, 85% removal was achieved by MIL-100 (Fe)@Fe3O4/CA, which confirms the highest performance for ternary systems. The coupling of CA considerably accelerated the transfer of photo-generated charge carriers and enhanced the performance of MIL-100 (Fe)/Fe3O4.Fe-based MOFs MIL-53 (Fe) as an integrated photocatalytic adsorbent was synthesized by (Gao et al., 2017), and the photocatalytic activity of MIL-53 (Fe) was tested by the photocatalytic decomposition of clofibric acid (CA)and carbamazepine (CBZ) with visible light illumination According to their findings, the MIL-53 (Fe) became highly active when a small amount of H2O2 was added in acid conditions ( Fig. 12). The conversion efficiencies of CBZ in MIL-53 (Fe)/Vis, MIL-53 (Fe)/H2O2, and MIL-53 (Fe)/H2O2/Vis system were 6.8%, 4.9%, and 90.1%, respectively, while for CA in MIL-53 (Fe)/vis, MIL-53 (Fe)/H2O2, and MIL-53 (Fe)/H2O2/vis system were 19.8%, 2.2%, and 98.2%, respectively. In this system, Fe (III) on the defect sites of MIL-53 (Fe) was exposed and could catalyze the decomposition of H2O2 to produce •OH by the Fenton-like reaction. On the other hand, since H2O2 can serve as an efficient scavenger that captures the photo-induced electrons in the excited MIL-53 (Fe), more •OH can be formed, further enhancing photocatalytic activity.The photocatalytic activity of the Pd@MIL-100 (Fe) has been firstly evaluated through the degradation of theophylline and ibuprofen under visible light irradiation by Liang et al. (2015) (Liang et al., 2015b). The results show that under visible light irradiation, Pd@MIL-100 (Fe) was able to degrade about 7.3% of theophylline after 150 min (without H2O2), which may contribute to the direct holes oxidation pathway. Pd@MIL-100 (Fe) becomes highly active by adding a certain amount of H2O2. Moreover, after visible light illumination for 150 min, the Pd@MIL-100 (Fe) exhibits much higher activity than MIL-100 (Fe). The Pd@MIL-100 (Fe) exhibits higher photoactivity than original-MIL-100 (Fe) in ibuprofen degradation. According to their results, with the decrease of the pH value, the degradation efficiency of theophylline was increased. However, an opposite tendency was observed for the Pd@MIL-100 (Fe)/ibuprofen system. S. Li et al. (2019); N. Li et al. (2019) synthesized Fe3O4/MIL-100 (Fe), via microwave in 30 min, for photocatalytic removal of Diclofenac sodium (DCF) (N. Li et al., 2019; S. Li et al., 2019). As the author claimed, it was the first in-situ fabricated Fe3O4 @MIL-100 (Fe) nanostructures using a microwave (MW) assisted synthesis method. In this catalytic system, Fe-O clusters in Fe3O4 @MIL-100 (Fe) were activated by visible light, and an electron-hole pair on its surface was formed. The electrons reacted with H2O2 and O2 to generate •OH and •O2 -radicals. The holes reacted with H2O to produce hydroxyl radicals which could directly oxidize DCF molecules. Based on their results, more than 99.4% of DCF was removed at the Fe3O4 @MIL-100 (Fe) dosage of 0.1 g/L. Their finding revealed the positive effect of the MW synthesis method on photodegradation of DCF, as shown in Fig. 13. The MW synthesis can be an effective way in MOFs synthesis because of its unique superiorities such as fast crystallization and narrower particle size distribution.MOFs such as MIL-88 (Fe), MIL-125 , and UiO-66 are also used for photocatalytic removal of PPCPs. Table 4 summarizes the results of applying these MOFs for the removal of pharmaceuticals.To promote photocatalytic performance, strategies such as functionalization, deposition of metal nanoparticles, combination with semiconductors in different types of heterostructures, and sensitization with dyes are adopted. MOF functionalization can be carried out by in-situ functionalization or post-synthetic modification (PSM) (Bedia et al., 2019; X. Li et al., 2016; Y. Li et al., 2016; Subudhi et al., 2018; Younis et al., 2020).In-situ functionalization is based on using functionalized organic linkers to synthesize MOF (Bedia et al., 2019; Subudhi et al., 2018). Metal nodes, organic linkers, and external functionalized groups can act as active photocatalytic sites (Subudhi et al., 2018). Various ligand functionalized groups (e.g., —CH3, —NH2, —NO2, —OH, Br-, Cl-, and —SH) have been applied to shift the photon absorption from the UV to the visible region. In particular, amine functional (electron-donating) groups have been broadly employed due to their interactions with π∗— orbitals of the ligand benzene ring, which increase electron density around the antibonding orbitals (formation of sp2) in the aromatic carbon ring for enhanced visible photon absorption (Younis et al., 2020).Post synthetic methods consist of functionalizing parent MOF by guest molecules (Bedia et al., 2019; C. Liu et al., 2019; W. Liu et al., 2019; N. Liu et al., 2019; Younis et al., 2020). Some of these strategies are shown in Fig. 14. Metal ion doping is an effective tool to enhance photocatalytic activity. Partial substitution of the metal centers can improve their semiconducting properties by electron-hole separation (Bedia et al., 2019; Younis et al., 2020). The design of bimetallic MOFs with targeted band gap structures is another PMS strategy for photocatalytic performance (Younis et al., 2020). In general, Au, Ag with strong plasmonic absorption properties and Pt, Pd, Ni, and Cu nanoparticle (MNPs) with excellent catalytic activity have commonly been used as active co-catalysts to improve electron transfer (Guo et al., 2021). These metal nanoparticles are expected to increase electron-hole pair separations and enhance photocatalytic activity (Wen et al., 2021b; Younis et al., 2020). Co-doped UiO-66 nanoparticles and Cu NPs MOF-based materials are among the less expensive alternatives than Pt NPs MOF-based photocatalysts (Bedia et al., 2019). There are different strategies for the synthesis of MNP/MOFs. Adsorption of MNPs by MOF substrates simply by mixing them in one of these methods. In this method, MNPs and MOFs are synthesized separately. The adsorption strategy is simple and easily scalable for practical industrial production, but the interactions between MNPs and MOFs need to be strengthened by adding binders to generate stable performances in long-term photocatalytic applications. MNPs also can be introduced into the matrix of preformed MOF substrates. One-pot, bottle around-ship, and sandwich structure strategies are among the other methods for synthesis of MNP/MOFs photocatalysts. Among all, the one-pot strategy is obviously the simplest and most straightforward one to obtain MNP/MOF composites, but the success of this strategy is far away from research satisfaction due to the very limited selections of specific metal precursors and MOF ligands as well as rigorous synthesis conditions including solvent selection, temperature programming, reaction time and etc. Until now, different MNP/MOFs structure such as Au/MIL-125, Pd/MIL-125, Pt/MIL-125, Pt/NH2-MIL-125, Au@ZIF-8, Au@ UiO-66 (NH2) has been synthesized (Guo et al., 2021). MNP/MOF composites not only combine both advantages of individual MNPs and MOFs also take advantage of the synergistic effect arising from the strong interaction between different components.However, there are unsolved issues regarding MNP/MOF development. Energy efficiencies obtained by MNP/MOF composites are still far lower than the basic criteria required by industry applications. A comprehensive understanding of mechanism of photocatalytic process requires advanced characterizing tools. Other challenging issue is constructing MNP/MOF composite with multifunctionality to expand their photocatalysis reaction scope (Guo et al., 2021) significantly.In recent years, heterojunction photocatalysts (MOF-based composite materials) have been introduced as the most common strategy for improving MOF activity under visible and solar light (Bedia et al., 2019; Wen et al., 2021b; Younis et al., 2020). Various semiconductors such as TiO2, BiVO4, and In2S3, Bi2MoO6 have been incorporated into MOFs to improve the charge transfer (Bedia et al., 2019; Wen et al., 2021b; Younis et al., 2020). Also, the synthesis of ternary systems to take advantage of the synergy between different heterojunctions has received significant attention (Bedia et al., 2019). Other hybrid structures such as MIL-100 (Fe)/Fe3O4/CA (Rasheed et al., 2018), Ag/AgCL/MIL-88A (Fe) (W. J. Huang et al., 2018; L. Huang et al., 2018; W. Huang et al., 2018), Ag/Ag3PO4/HKUST1 (Shen et al., 2013), Pd/GO/MIL-101 (Cr) (Wu et al., 2015), etc., are prepared, and their photocatalytic performance was examined through the removal of organic pollutants.Despite the accelerating progress in applying effective strategies to improve MOFs' photoactivity, a complete analysis is necessary till MOFs can offer remarkable opportunities for natural waste water purification. Contaminant concentrations varied widely from study to study, but it is unclear how relevant the contaminant concentrations are to a specific wastewater stream. Another aspect is the limited diversity of target contaminants (mostly organic dyes) despite the wide range of existing pollutants. Also, in most studies, synthetic waste samples (i.e. simple matrix) are considered. A complete discussion of photocatalysis mechanism is still needed. Furthermore, identifying possible decomposition products and byproducts (other than CO2 and H2O) needs to be studied.Adsorption techniques have been widely applied as an alternative wastewater remediation process and presented as a solution to the challenging task of incomplete extraction of pollutants during wastewater processing. Organic contaminants, in particular, tend to be more resistant to various types of water treatment due to their hydrophobicity and lower molecular weight. In the process of adsorption, pollutant molecules are drawn onto the adsorbent materials' surfaces through the process of diffusion from the solution bulk to the adsorbents' active pores (Kumar et al., 2019; Siipola et al., 2020). Generally, the mechanism occurs due to intermolecular forces of attraction, such as physisorption (e.g., van der Waals, π–π interactions and hydrogen bonding) and chemisorption (e.g., ionic interactions) (Fu et al., 2019; Lv et al., 2019; Ahmed et al., 2022). The adsorption process has been highlighted by the adsorbent materials' unique properties, such as large specific Brunner Emmett Teller (BET) surface area, high porosity, water, and thermal stabilities, high selectivity for the contaminant of interest, low cost and availability, ease of regeneration, etc. (Xin et al., 2017; Zhan et al., 2020).The flexibility to synthesize a variety of frameworks from numerous clusters of metal ions with organic linkers allows for an unlimited range of crystalline MOFs with microporous or mesoporous structures. Furthermore, diverse functional groups in the metal node and organic linkers serve such as adsorption centers for different organic pollutants (Hasan and Jhung, 2015). In general, the adsorption of POPs on MOF-based nanomaterials (Shan and Tong, 2013; Song and Jhung, 2017) is mainly determined by the materials' chemical properties and physical structure, as the pore structure, specific surface area, and surface functional groups might affect the capacity of adsorption directly. In general, the process of adsorption can occur through physical or chemical interaction. Determination factors are molecular size, solubility, chemical composition, surface charge, reactivity, and hydrophobicity. The interaction between contaminants and adsorbent surfaces, commonly, may occur through electrostatic interaction, complexation/coordination, electrostatic interaction, ion exchange, oxidation. Some of the mechanisms governing the adsorption of MOFs in aqueous solutions are illustrated in Fig. 15.Parameters influencing adsorption behavior of MOFs, including adsorbent dosage, adsorbate concentration, pH, and contact time ( Fig. 16). It is essential for identifying optimized conditions leading to study these parameters for each type of MOFs in order to maxize adsorption performance and to better understand the mechanisms of interaction between adsorbates and adsorbents. The following section of this review paper will discuss the reports on using MOF-based adsorbents in the removal of persistent organic pollutants (POPs) from water and wastewater.The broadly investigated absorbent materials, so far, mainly involve activated carbon (AC), organo-zeolites, alkylsilane modified silica, polymeric resins, etc. (Liu et al., 2013; Lemić et al., 2006; Westerhoff et al., 2005; Groisman et al., 2004; Bagheri and Mohammadi, 2003; Masqué et al., 1998b, 1998a) have recently been explored as promising adsorbents for the removal of organic molecules from aqueous solutions. MOFs diverse chemistry, good chemical and physical stability, and tunable pore structure possibility of introducing functional groups make them very attractive candidates for the adsorption of POPs such as pesticides (Feng et al., 2018; Ghanbari et al., 2020; Li et al., 2018).Most MOF-based water purification researches focus on the adsorption capacity and regeneration. MOFs’ adsorption performance (e.g. selectivity) is governed and influenced by the interactions between MOFs’ active sites (or functional groups) and targeted contaminants.To remove pesticides from aqueous solutions, common MOFs like Chromium-based MIL-53 and MIL-101; Zinc-based ZIF-8 and Cobalt-based ZIF-67; Zirconium-based UiO-67 and UiO-67/GO; and Cu-BTC have shown to be particularly efficient. MIL-53 (Cr) was employed for the first time in a study by Jung et al. (2013) to adsorb 2,4-D weed killer (2,4-dichlorophenoxyacetic acid), a commonly used herbicide in agriculture (Jung et al., 2013). Compared with AC or USY zeolite, Cr-based MIL-53 exhibited a much larger adsorption capacity and substantially faster adsorption (within 1 h). The adsorption of 2,4-D weed killer was highly effective, particularly at low 2,4-D concentrations in solution. The adsorption process was greatly favored by π − π stackings and electrostatic interactions between the Cr-based MIL-53 and 2,4-D weed killer. After four cycles of use, the MOFs were still effective. Smedt et al. (2015) also demonstrated that Fe-based MOF-235 had an adsorption advantage over zeolite and AC (De Smedt et al., 2015). The researchers compared the adsorptive performance of Fe-based MOF-235 with that of AC and zeolite in the removal of clopyralid, bentazon, and isoproturon from water. In terms of kinetics and capacity, MOF-235 was the most effective at removing these three pesticides. The MOF-235, on the other hand, had poor reusability due to its low water stability. Mirsoleimani-azizi et al. (2018) studied the mesoporous Cr-based MIL-101 as an adsorbent in a fixed-bed system for the constant removal of diazinon (Mirsoleimani-azizi et al., 2018). At 150 mg/L diazinon concentration, the MIL-101 was reported to have 92.5% diazinon removal from aqueous solutions. In another study for the adsorptive removal of pesticide 14C-ethion, which is considered carcinogenic and toxic, a Cu-based MOF, Cu-BTC, was used and exhibited reliable removal efficiency (Abdelhameed et al., 2018). The adsorption capacity of Cu-BTC at 150 min was calculated to be approximately 122 mg/g for a 14 C-ethion concentration of 75 mg/L, and the MOF demonstrated stability for up to six cycles. The adsorption followed first-order kinetics, confirming the chemisorption process through the coordination of 14 C-ethion to the Cu (II) atom of Cu-BTC MOF. Abdelhamid et al. (2019) also conducted a study comparing the adsorption of two widely-used pesticides (ethion and prothiofos) on Co-based ZIF-67 and Zn-based ZIF-8 (Abdelhameed et al., 2019). The maximum adsorption capacity of Co-based ZIF-67 for ethion and prothiofons was calculated to be 211 and 261 mg/g, respectively, while Zn-based ZIF-8 was 279 and 367 mg/g, respectively. The different calculated capacity of adsorption was reported to be attributable to the weaker coordination of ethion and prothiofos to the Co metal ions compared with that to the metal ions of Zn. Despite this, in the adsorption of ethion and prothiofos onto both MOFs, hydrogen bonding (HB) played a significant role.Zr-based MOFs have been widely explored in removing different pesticides via adsorption due to their high-water stability, high surface area, optimal pore size, and Zr-OH moiety presence. In this sense, to remove methylchlorophenoxypropionic acid (MCPP) from water, a Zr-BDC, Zr-based UiO-66, was studied. The results showed that the UiO-66's adsorption capacity was ∼7.5 times, and the kinetic constant was ∼ 30 times that of AC (Seo et al., 2015). In another study to remove glyphosate (GP) and glufosinate (GF), a Zr-based UiO-67 - which appeared to have high stability, suitable pore size, and strong affinity - was investigated, and results showed a large adsorption capacity for GF (360 mg/g) and GP (537 mg/g) (Zhu et al., 2015, p. 67). Pankajakshan et al. suggested that, since the metal nodes are the same in both MOFs, it is likely the larger pore diameter of the NU-1000 (Zr), which makes its adsorption more effective than UiO-67 (Zr). They also observed a higher adsorption energy for GP (−37.63 kJ/mol) for NU-1000 (Zr) compared to UiO-67 (−17.37 kJ/mol). The more negative adsorption energy indicates a stronger interaction of GP with the Zr clusters of the MOFs. This is in line with the shorter Zr···O–P interatomic distance during the polar interaction in the case of NU-1000 (4.2 Å) compared to UiO-67 (4.6 Å). (Pankajakshan et al., 2018). In a comparative study by Akpinar and Yazaydin (2018), Zr-based MOFs of UiO-66 and UiO-67 and Zn-based ZIF-8 were applied for the atrazine (ATZ) removal from aqueous solution and compared with commercial AC (F400) (Akpinar and Yazaydin, 2018). UiO-67 (Zr), ZIF-8 (Zn), and F400 were able to remove the ATZ up to 98%; however, UiO-66 appeared ineffective. Due to its larger pores, UiO-67 displayed stronger and interestingly faster adsorption performances (removing ∼98% ATZ within 2 min, while UiO-66's hydrophilic characteristics rendered it ineffective for the ATZ adsorption even through surface interactions. ZIF-8 was also revealed to be an efficient ATZ adsorbent due to its hydrophobic characteristics, which inhibited considerable amounts of water adsorption, thus favoring the adsorption of ATZ.Another study revealed that only Ze-based NU-1000 out of eight Zr6-based MOFs exhibited a 100% ATZ removal effectiveness within 5 min (Akpinar et al., 2019). This study also suggested that the linker structure might significantly influence the adsorption characteristics of the MOFs, as their performance improved with the number of carboxylic acid groups and aromatic rings in the linker. The presence of a pyrene-based linker in the MOF structure, which provides enough sites for π–π interactions, might explain the significant adsorption of NU-1000 (Zr). Additionally, a negligible efficacy loss was reported after three cycles of regeneration in the adsorbent. Jamali et al. (2019) (Jamali et al., 2019) used Zr-based MOFs UiO-66 and UiO-67 for the adsorptive removal of metrifonate and dichlorvos (OPP) from an aqueous solution in another investigation. The UiO-67 outperformed the UiO-66 in OPP removal, with greater removal rates of 99.8% versus 97.8%, respectively. UiO-67 (Zr) adsorption capacity for the metrifonate and dichlorvos removal was also measured to be 379 and 571 mg/g, respectively.Due to their high porosity and sufficient adsorption sites, pure MOFs are particularly efficient in pesticide removal from aqueous solutions. The findings of the adsorptive removal of pesticides using pristine MOFs are summarized in Table 5.Abdelhameed et al. have validated the efficiency of Al-based MIL-53 in eliminating organophosphorus (OPP) pesticide dimethoate from an aqueous solution (2021) (Abdelhameed et al., 2021b). The surface area of MIL-53 was increased as a result of functionalization by Al, resulting in a rise in the number of adsorption sites and a higher absorption capacity.To remove herbicides such as alachlor (ALA), diuron (DUR), gramoxone, and tebuthiuron, a Cr-based MIL-101 was recently modified with thiophene or furan. The resulting Cr-MIL-101 (Cr)-C(1−5) samples showed enhanced adsorption efficiency, which was ascribed to HB interactions and π–π stacking (Y. Q. Yang et al., 2019; Y. Yang et al., 2019). In another study for GP adsorption, Feng and Xia (2018) used a modified Cr-based MIL-101 loaded with a urea group or amino and studied the effects of ionic strength and pH on the adsorption process (Feng and Xia, 2018). Due to ESI between the MOFs' surface with positive charges and the GP anions, the urea- and amino-functionalized MOFs demonstrated the highest adsorption capability at pH > 6 and pH = 3, respectively. The researchers witnessed lower adsorption performance of urea-functionalized MIL-101 (Cr) compared to NH2-MIL-101 (Cr) due to steric hindrance. In a recent study by Wu et al. (2020), UiO-66 (Zr)-NMe3+, a MOF functionalized with cationic sites, was tested for the removal of 2,4-D, exhibiting a significant capacity of adsorption (279 mg/g), which may be attributed to the existence of π–π conjugation and electrostatic interactions (ESI). Furthermore, the modified MOF demonstrated high reusability and an acceptable adsorption rate within 2 h of contact time (Wu et al., 2020).Development and use of MOF-based composites in different applications are gaining momentum amongst researchers of other disciplines. In this section, several classes of MOF-based composites that are being used for POPs removal will be discussed.In a recent study, a Cu-BTC@cellulose acetate composite was prepared by Abdelhameed et al. (2021) and used for the removal of dimethoate (Abdelhameed et al., 2021a). An increase in the cellulose acetate's adsorption performance was observed when adding 40% Cu-BTC. The coordination between the composite's active site and dimethoate molecules, HB, and ESI might explain the adsorption of dimethoate via Cu-BTC@cellulose acetate composite in this study. In another study by the same research team, a Cu-BTC@cotton composite extracted ethion (Abdelhameed et al., 2016). By removing 97% ethion from water, the adsorption capacity was calculated to be 182 mg/g. Langmuir model was found to be the best-fitted adsorption model, and the adsorption process remained almost unchanged for up to five cycles of regeneration. The composite's adsorption performance was explained by the interaction of the sulfur atoms of ethion with the copper atom in the MOF through a coordination bond and the availability of cellulose functional groups for HB. Also, several other pesticides, including 4-chlorophenoxyacetic acid, dicamba, 2,4-D, and 2-(2,4-dichlorophenoxy) propionic acid, were preferentially absorbed by using the composite adsorbent Cotton@UiO-66 (Zr) in a packed column, in a study by Su et al. (2020) (Su et al., 2020).By depositing Zr-based UiO-66 on ionic liquid-modified chitosan (ILCS), Huang et al. (2020) prepared ILCS/U-X, a powder adsorbent, for the removal of organic herbicides from an aqueous solution (Huang et al., 2020). High maximum capacity for the adsorption of 2,4-D (893 mg/g) was reported for the contact time of 1 h, attributable to the presence of O-containing groups in Zr-based UiO-66 and through ESI and HB.To remove GP, Y. Yang et al. (2017); Q. Yang et al. (2017) prepared a highly effective nanocomposite constructed from Zr-based UiO-67 combining graphene oxide (GO) (Q. Y. Yang et al., 2017; Q. Yang et al., 2017). Compared with the other GO-based adsorbents, a higher capacity of adsorption (483 mg/g) was reported for the produced composite, which was attributable to the interaction of the GP molecule with the Zr-OH functionality UiO-67 through the formation of Zr–O–P bonds. Liang et al. (2021) used a composite made up of Zr-based MOFs (ZIF-8 or UiO-66–NH2) on a carbon nanotube aerogel (MPCA) to remove herbicides (chipton and alachlor) (Liang et al., 2021). The resultant UiO-66–NH2 @MPCA composite demonstrated significant chipton adsorption capacity (227.3 mg/g) and was reported to be reusable for up to 5 cycles. The exposed active site explained the chipton and alachlor adsorption onto the MOF@MPCA composite. The MPCA's micron-size pores expanded the interaction between pesticides and MOF nanoparticles and enhanced the adsorption of the pesticides. The influence of pH was also investigated, and it was discovered that UiO-66–NH2 @MPCA displayed the highest chipton adsorption capability at pH= 4 since chipton exists mainly in anionic forms at pH> 3.1. The surface charge of UiO-66-NH2 is positive at pH＜5.1. As a result, the adsorption of chipton could be due to electrostatic interaction. However, despite occurring charge repulsion at pH> 5.1, the adsorption capacity was good, revealing that there might also be other interaction forces like π–π stacking and HB interactions. Furthermore, for the nonionic pesticide alachlor, UiO-66–NH2 @MPCA also showed good adsorption capabilities, implying that π–π or HB interactions were the significant contributors to adsorption.Amongst various MOF-based composites, magnetic derivatives have equally been confirmed to enhance the MOFs' adsorptive performances due to their magnetic properties (Kumar Gupta et al., 2017; Wang et al., 2014). For instance, using iron oxide-GO-β-cyclodextrin (Fe4O3-GO-β-CD) and a Cu-based MOF (resulting in a rise in the BET surface area and thus boosting its adsorptive performance), Liu et al. (2017) produced a magnetic nanocomposite (M-MOF). Acting as magnetic cores, the M-MOF extracted neonicotinoid insecticides (i.e., imidacloprid, thiamethoxam, nitenpyram, acetamiprid, dinotefuran, thiacloprid, and clothianidin) from water (Liu et al., 2017). Such results were observed due to the M-MOF's large surface area and hydrophobic inner pore. The pesticides' adsorption mechanisms were also described by HB, ESI, hydrophobic interactions (HPI), and π–π stacking interactions attributable to hydrophobic and N-containing groups delocalized π electrons from the benzene rings in the molecules of the adsorbate. Through coordination polymerization, Liu et al. (2018) prepared a functional magnetic hybrid composite (M−M−ZIF−8) by depositing particles of Zn-based ZIF-8 on magnetic multiwalled carbon nanotubes (Liu et al., 2018, p. 8). The provided composite was found to be an effective adsorbent for the removal of eight different OPPs (i.e., diazinon, triazophos, phosalone, methidathion, profenofos, ethoprop, isazofos, and sulfotep) from water and soil. Additionally, the Freundlich bimolecular adsorption model was reported to be best fitted, and the OPPs' adsorption proceeded by electron exchange between the adsorbents' vacant active sites and molecules of OPPs. For the detection and adsorption of glyphosate, Yang et al. (2018) applied a layer-by-layer assembly method to prepare magnetic hybrid materials Fe3O4 @SiO2 @UiO-67 (Zr) (Yang et al., 2018). The authors reported a high adsorption capacity (257 mg/g) and good regeneration functionality, attributable to the Zr–OH moieties' effective interaction with the phosphate group and Fe3O4 magnetic core. A double-layer magnetic MOF (M-ZIF-8 (Zn)@ZIF-67 (Zr)) prepared by B. Li et al. (2020); Z. Li et al. (2020); T. Li et al. (2020) exhibited excellent capability in removing fipronil and its derivatives from aqueous solution (T. B. Li et al., 2020; Z. Li et al., 2020; T. Li et al., 2020). The established large pores in the composite's bilayer structure allowed for 70.9–99.7% fipronil removal from water and cucumber.Providing sufficient adsorption sites, MOF composites and functionalized MOFs have proved to be highly efficient in eliminating pesticides from water. Notably, for removing anionic/cationic pesticides, MOF fabricated with anionic/cationic (respectively) active sites can be favorable because of electrostatic interactions. Composite MOFs have also been shown to efficiently remove nonionic pesticides due to active sites for coordination, π–π stacking, and HB interactions. MOFs incorporating magnetic materials have also shown excellent reusability since used adsorbents may be easily removed from an aqueous solution using a magnetic field. The application of magnetic strategy to the field of MOFs is justified for a number of reasons. On the one hand, the potential for incorporating porosity into these magnetic coordination polymers presents an appealing way to create multifunctional materials in which the magnetism can be tuned by the presence of molecules in the pores. The chemical pressure created by the guest molecules, the intermolecular interactions between the guests and the framework, or changes in the MOF's electronic properties can all be attributed to these systems as potential explanations for the magnetic behavior. Due to their magnetic response, these characteristics might be useful for detecting the molecular species trapped in the pores. On the other hand, the presence of magnetic centers within a crystalline MOF's nodes or pores makes it possible to organize these magnetic centers into nanostructures while maintaining their spatial segregation. As quantum technologies require a controlled disposition of magnetic moieties in the space, such a feature might be of potential interest (Espallargas and Coronado, 2018). Table 6 summarizes the results of the adsorptive removal of pesticides from aqueous media using MOF-based composites and functionalized MOFs.Due to their nanoporous structure, high surface area, and unique physical/chemical properties; MOF-derived carbonaceous materials have received enormous attention in a variety of applications, including catalysis, adsorption, and energy conversion and storage (Borchardt et al., 2017; Islamoglu et al., 2020; Duan et al., 2020; Mondol and Jhung, 2021).Prepared by carbonization of Fe (III)-modified Zn-based MOF-5, a MOF-derived magnetic porous carbon-based sorbent (MPC) was employed to remove ATZ from an aqueous solution (Chen et al., 2017). The enhanced adsorption capacity for ATZ removal was reported, attributed mainly to adequate hydrogen bonding between MPC and ATZ with the best-fitted isotherm model of Dubinin-Ashtakhov for this investigation. The synthesis of a magnetic nanoporous carbon (MNPC) via carbonization of Co-based ZIF-67 MOF was performed to extract multiple neonicotinoid insecticides (i.e., acetamiprid, thiacloprid, thiamethoxam, and imidacloprid) from an aqueous solution (Hao et al., 2014). Exhibiting 15 cycles of regeneration, the Co-MNPC confirmed an excellent adsorption efficacy for removing the compounds. Bhadra et al. (2020) synthesized a Zn-based MOF-74-derived porous carbon (CDM-74) to investigate its removal performance of DEET insecticide from an aqueous solution (Bhadra et al., 2020). Due to the high acid site density and mesopore volume, the CDM-74 demonstrated a high adsorption capacity (340 mg/g). Additionally, the adsorbent performance remained unchanged for up to 4 cycles of DEET removal, confirming the derived MOFs' efficacy in removing pesticides from water. Through the carbonization of a multifunctional β-cyclodextrin MOF, another carbonaceous material, β-CD MOF-NPC, was prepared and investigated to remove amide herbicides (i.e., Pretilachlor, Acetochlor, Alachlor, Metolachlor) (C. N. Liu et al., 2019; W. Liu et al., 2019; C. Liu et al., 2019). The results exhibited that the resultant microporous structure with a high potassium content and large surface area effectively removes amide herbicides via ESI, HB, and π–π interactions.The sufficient adsorption sites and high porosity of MOF-derived carbonaceous materials have been reported as highly effective factors in removing pesticides from water. Table 7 summarizes the research findings on using carbonaceous materials produced from MOFs or MOF composites in the adsorptive removal of pesticides.Acting as intermediates for industrial production, phenols (a class of organic compounds) are broadly applied by chemical and allied industries in manufacturing valuable products, such as petrochemicals, plastics, pesticides, herbicides, dyes, cosmetics, pharmaceuticals, epoxies, detergents, among others. Non-feasible and non-economical conventional treatment methods that failed to address the persistent problems and limited applicability of photocatalytic degradation due to the semi-conductive properties of the materials have turned the attention towards the adsorptive removal of phenolic compounds as an effective removal technique.To date, there have been several reports on the adsorptive removal of phenolic compounds using MOFs and MOF-based composites. Park et al. (2013) extracted bisphenol A (BPA) from water employing Cr-based MIL-53, which outperformed AC and ultra-stable Y zeolite in terms of the maximum adsorption capacity (Park et al., 2013). The π–π interactions and HB between the hydroxyl groups of MIL-53 (Cr) and BPA were identified as the particular interactions that favored adsorption in MOFs. Qin et al. (2014) also successfully used MIL-100 (Fe) and MIL-101 (Cr) to remove BPA from an aqueous solution (Qin et al., 2015). MIL-101 (Cr) absorbed BPA more significantly than MIL-100 (Fe) and faster than AC. These findings were ascribed to MIL-101(Cr)'s favorable textural characteristics, including a higher surface area and bigger pore apertures. Furthermore, the interactions between BPA and MIL-101 (Cr) have been proposed to be π–π interactions and HB. In another work, MIL-100 (Cr), MIL-100 (Fe), NH2-MIL-101 (Al), and an AC were used to remove phenol and p-nitrophenol (PNP) from aqueous solutions (B. L. Liu et al., 2014; B. Liu et al., 2014). In terms of adsorption of phenol, the AC samples outperformed the MOF materials. The poor performance was attributed to MIL-100 (Cr) and MIL-100 (Fe) having high binding energies with water. In terms of PNP adsorption, the performance order was as follows: NH2-MIL-101 (Al) > AC > MIL-100 (Fe) ∼ MIL-100 (Cr). According to the authors, the impact of metal sites in MOFs was insufficient to explain the PNP and phenol adsorption from water. Instead, HB between the PNP nitro groups and the MOF amine groups was ascribed to the enhanced adsorption of PNP by NH2-MIL-101 (Al). To remove BPA from the aqueous solution, two Al-based MOFs, MIL-53 (Al) and MIL-53 (Al)-F127, were synthesized, and the results showed a fast BPA removal (90 min for MIL-53 (Al) and 30 min for MIL-53 (Al)-F127) (Zhou et al., 2013). The maximum removal on MIL-53 (Al) and MIL-53 (Al)-F127 was measured at 329.2, and 472.7 mg/g, respectively, and the BPA sorption kinetics data followed the pseudo-second-order model. π–π Bonds and HB explained the BPA sorption. The well-known Co-based MOF, HKUST-1, was investigated for the removal of p-nitrophenol (PNP) and demonstrated a high adsorption capacity of 400 mg/g, with maximum removal accomplished within 40 min (Andrew Lin and Hsieh, 2015). The HKUST-1 high adsorption capacity for the PNP was attributable to the MOF's high interactions with the NO2 compounds of the PNP via HB. In another study, Cu- and Zr-based MOFs, MOF-199, and ZIF-8 were used to remove phenol and PNP (Giraldo et al., 2017). The adsorption capacity for phenol and PNP on MOF-199 was higher (Phenol: 79.55% and PNP: 89.3%), while ZIF-8 was measured to be 65.5% for phenol and 77.0% for PNP. Adsorption of phenols followed the Langmuir isotherm model, and kinetics was fit to pseudo-second-order. A hexagonal MOF of NH2-MIL-88B was prepared to remove 2,4,6-trinitrophenol (TNP), and the maximum capacity of adsorption based on the Langmuir isotherm was reported to be 163.66 mg/g (Guo et al., 2018). The NH2-MIL-88B adsorption mechanism for TNP was attributed to HB interaction and complexation between unsaturated Fe (III) on the surface of NH2-MIL-88B and -OH in TNP. The crystalline and water-stable Zr-based MOF, NH2-UiO-66, was also examined for the adsorptions of 2,4-dinitrophenol, 2,4,6-trinitrophenol, 2,4-dinitrotoluene, and 2,4,6-trinitrotoluene in an aqueous solution (Xu et al., 2017). Higher equilibrium adsorption capacities were achieved with the formation of HB between the atoms of NH2-UiO-66 (Zr) and the pollutants. In a study for removing bisphenol A, Luo et al. (2019) observed an enhanced removal performance by employing a sodium alginate-chitosan-based aluminum MOF composite, Al-MOF/SA (Luo et al., 2019). The capacity of adsorption was calculated to be 136 mg/g. The experimental findings demonstrated greater adsorptive performance when compared to the counterparts of Al-MOF/SA. The most important processes involved in adsorption mechanisms were reported to be cation-interaction, HB, and π–π stacking. After five consecutive batch cycles, as-synthesized beads were recycled and regenerated with over 95% adsorption efficiency. Abazari and Mahjoub (2018) prepared a Zn (II)-based MOF, [Zn(TDC)(4-BPMH)]n·n(H2O), and investigated its performance for the removal of 2, 4-dichloropheno from wastewater (Abazari and Mahjoub, 2018). With a 60-ppm concentration, the 2, 4-dichloropheno removal efficiency (after 180 min) was observed up to 94.5%. The removal of 2, 4-dichloropheno over this MOF follows the first-order reaction kinetics.The SiO2 @MIL-68 (Al) composites were prepared, and their adsorption efficiency was tested for the removal of aniline (Han et al., 2016). The results demonstrated a high adsorption capacity of 532 mg/g of the composite towards aniline. The aniline adsorption on SiO2 @MIL-68 (Al) obeyed the Langmuir model. The adsorption equilibrium was reached in only 40 s with good reusability up to 5 cycles. HB interaction between μ 2-O in adsorbent and −NH2 in aniline and the π–π interaction between the framework and the benzene rings of aniline ascribed to the process of adsorption. A carbon nanotube (CNT)@MIL-68 (Al) composite was synthesized and compared with pure MIL-68 (Al) to investigate their performances in phenol removal from an aqueous solution (Han et al., 2015). CNT@MIL-68 (Al) showed the highest capacity of adsorption (257 mg/g), 119% higher than that of pristine MIL-68 (Al) (117.6 mg/g), possibly attributable to the expansion of small micropores introduced by the incorporation of CNTs. The high adsorption capacity is caused by the HB between –OH in phenol and –COO− in adsorbent and the π–π interactions between the benzene rings of phenol and the composites. Another comparison between pristine MOF and its composite was made in a study to remove PNP using MIL-68 (Al) and MIL-68 (Al)/GO (Wu et al., 2016). The adsorption capacity of PNP calculated from the Langmuir model was 332 mg/g for MIL–68 (Al)/GO, which was higher than the MIL-68 (Al) (271 mg/g) due to the higher surface area of the composite. The adsorption process fitted well with the pseudo-second-order model. This performance is attributed to the HB and π–π dispersion interaction between the composite and PNP. Cu-BDC MOFs decorated over GO and CNT hybrid nanocomposites, namely Cu-BDC@GO and Cu-BDC@CNT, were synthesized by Ahsan et al. (2019) for BPA removal from water (Ahsan et al., 2019). The hybrid nanomaterials demonstrated a high adsorption capacity of 182 and 164 mg/g toward the BPA removal for BDC@GO and Cu-BDC@CNT, respectively, which was much higher than that of Cu-BDC MOF itself (60.2 mg/g). The results confirmed that the Freundlich model describes the experimental data best, and the kinetics data were best fitted to the pseudo-second-order kinetic model. The π–π interactions between the nanomaterials and BPA played a crucial role in the BPA adsorption process. A biocomposite of laccase@HKUST-1 was also prepared by Zhang et al. (2020) (Zhang et al., 2020). The adsorption performance of the material for BPA was tested, and the BPA removal after four h by laccase@HKUST-1 was reposted to be 74.2%.Combining the unique surface activity of MOFs with the excellent carry nature of the porous polymers can result in unexpected adsorption performances. A MOFs/polymer composite membrane, MIL-68 (Al)/PVDF, was fabricated and tested to remove PNP from an aqueous solution (Tan et al., 2019). The maximum adsorption capacity of PNP on MIL-6/PVDF reported 183.49 μg/cm2 (94%), and the material was reusable for up to 6 cycles. The Langmuir isotherm model characterized the adsorption process. The coordination bonding formed between Al (III) of MIL-68 (Al) and NO2 of PNP was responsible for the PNP removal.Pharmaceuticals and personal care (PPCPs) items are other significant areas of emerging pollutants. They comprise various chemicals such as medications, cosmetics, and veterinary treatments; all considered necessary life elements. Unfortunately, conventional wastewater treatment procedures such as coagulation, sedimentation, and flocculation, frequently followed by chlorination, are not optimally suited to remove these potentially hazardous PPCPs efficiently. Many conventional treatment approaches (e.g., adsorption, photocatalytic degradation, separation, thermal decomposition, hydroxylation, biodegradation, coagulation, flocculation, sedimentation, ozonation, and advanced oxidation processes) account for partial contribution to the removal of PPCPs (Westerhoff, 2003; Li et al., 2015; Wang and Wang, 2016; Jin et al., 2020; Reyes et al., 2021). Among the methods described above, the adsorptive elimination of PPCPs is one of the most effective approaches. Zhao et al. (2019) used UiO-66 MOFs for the removal of three broadly applied PPCPs: 2,4-Dichlorophenoxyacetic acid (2,4-D), diclofenac sodium (DCF), an anti-inflammatory drug, and clofibric acid (CLA), an extensively employed herbicide (Zhao et al., 2019). The results showed that UiO-66-NH2 had the highest capacity for absorption of these PPCPs, which was 3–4 times larger than the UiO-66-COOH counterpart, which had the lowest absorption capacity. The structure-function analysis revealed that HB, ESI, and interfaces between MOFs and PPCP molecules significantly impacted the adsorption process. Despite UiO-66's high PPCPs removal capacity, its usage has been limited by difficult separation and recovery from aqueous environments.In a study on the adsorptive removal of tetracycline, a method was proposed to make hierarchical pore-structured UiO-66 (Yuxi Jie Zhang et al., 2018; Ying Zhang et al., 2018; Yuxi Zhang et al., 2018). These mesostructured MOFs demonstrated a significant boost (up to 430%) in adsorption capacities compared with microporous UiO-66 (i.e., 667 and 126 mg/g, respectively). The pseudo-second-order kinetic model and Langmuir isotherm model could well describe the adsorption of tetracycline on H-UiO-66 MOFs. Authors attributed the high tetracycline removal capacity of H-UiO-66 MOFs to the π–π stacking interaction between the aromatic ring of tetracycline and the benzene ring of H-UiO-66 ( Fig. 17). The enhanced adsorption capability of UiO-66 with hierarchical pores demonstrates that adsorbate and adsorbent size matching is critical.A new mechanism for the adsorption of p-arsanilic acid was proposed in another study using amine-modified UiO-67 (C. H. Tian et al., 2018; C. Tian et al., 2018). The synthesized MOFs were comprised of UiO-67-NH2 (1) (the ligand has one amino group) and UiO-67-NH2 (2) (the ligand has two amino groups), and pristine UiO-67. The order of maximum adsorption capacity was UiO-67 > UiO-67-NH2 (1) > UiO-67-NH2 (2); however, the order was reversed following normalization by surface area. Three adsorption mechanisms were discovered using DFT calculations: π–π stacking, As-O-Zr coordination, and NH∙∙∙O HB, which is a novel characteristic in the adsorption of p-arsanilic acid. Both UiO-67–NH2 and UiO-67 maintained the adsorption performances up to 4 cycles of adsorption. Peng et al. (2019) also used MIL-101 (Cr) to remove sulfonamide-containing antibiotics such as sulfamonomethoxine, sulfadimethoxine, and sulfachlorpyridazine (Peng et al., 2019). The authors reported high removal efficiency of the pharmaceuticals over the MIL-101 (Cr) (196.08, 588.24, and 142.86 mg/g, respectively). The adsorption proceeded according to the pseudo-second-order kinetic model. The key mechanisms for three sulfonamides adsorption on MIL-101 were found to be ESI and HB. Chai et al. (2019) tested the application of MIL-101 (Cr)–SO3H on the efficient adsorption of moxifloxacin and gemifloxacin (Chai et al., 2019). MIL-101 (Cr)–SO3H showed maximal adsorption capacities of 493 and 535 mg/g for the respective fluoroquinolones. These results are significantly greater than those for MIL-101 (Cr) adsorption of gemifloxacin and moxifloxacin. The MOFs were able to be regenerated for up to 4 cycles. The adsorption behavior of MOF was shown to follow the Langmuir isotherm and pseudo-second-order models. B. Li et al. (2020); T. Li et al. (2020); Z. Li et al. (2020) also used the same MOF material (MIL-101 (Cr)–SO3H) to remove ciprofloxacin, a frequently used antibiotic from the fluoroquinolone family (Z. B. Li et al., 2020; T. Li et al., 2020; Z. Li et al., 2020). The adsorption kinetics were best fitted to the pseudo-second-order model. Furthermore, the equilibrium adsorption data followed the Langmuir model. The maximum capacity of adsorption (564.9 mg/g) was found to be much higher than that of pure MIL-101 (Cr) (113.2 mg/g).Same as other fluoroquinolones, ciprofloxacin adsorption was facilitated by the ESI of SO3−. MIL-101 (Cr) was also used by Shadmehr et al. (2019) to eliminate diazinon from aqueous media (Shadmehr et al., 2019). This chemical is a thiophosphoric acid ester insecticide designed to replace DDT; however, it is harmful to humans and animals. The maximum adsorption capacity was 75.04 mg/g (96.1%), higher than activated bentonite and activated carbon. The adsorbent was reusable for up to 5 successive cycles for adsorption of diazinon. The Langmuir model and pseudo-second-order model were found to be more consistent with the experimental data.Gao et al. (2019) prepared MIL-53 (Al), MIL-53 (Cr), and MIL-53 (Fe) with narrow and large pore sizes (Gao et al., 2019a). The large pore conformation of MIL-53 (Al) and MIL-53 (Cr) exhibit a high removal efficiency of sulfamethoxazole (451 and 469 mg/g, respectively), while the narrow pore-sized MIL-53 (Fe) was found to be inefficient for the adsorptive removal of sulfamethoxazole's large molecules.In a study to capture nitroimidazole antibiotics, MIL-53 (Al) demonstrated high performance in the adsorption of dimetridazole (DMZ) (Peng et al., 2018). The adsorption capacity was 467.3 mg/g, with a fast adsorption rate of 10 min. The strong van der Waals interactions and the HB between –NO2 of DMZ and μ 2-OH of the MOF were suggested to be the interaction mechanism between DMZ and MIL-53 (Al). The data were best fitted by the Langmuir isotherm model and pseudo-second-order model. After four cycles, MIL-53 (Al) retained over 98% of its initial DMZ adsorption capacity.An amino-functionalized In-based MOF, MIL-68 (In)–NH2, was also prepared to capture p-arsanilic acid in an aqueous media and compared with pristine MIL-68 (In) in terms of the capacity of adsorption (Lv et al., 2018). Accordingly, the adsorbents exhibited a maximum capacity of 402 mg/g and 340 mg/g, respectively. The adsorption obeyed the pseudo-second-order model and was better characterized by the Langmuir model than the Freundlich model. The removal performance of MIL-68 (In)–NH2 was attributed to π–π interaction, ESI, and HB between the functional groups of p-arsanilic acid and organic linkers of the adsorbents. S. Li et al. (2019); N. Li et al. (2019) used ZIF-8 to remove two common antibiotics, oxytetracycline hydrochloride (OTC) and tetracycline (TC), simultaneously (N. S. Li et al., 2019; N. Li et al., 2019). Resulting of the synergistic interaction of TC and OTC, the removal efficiency of a mixture of the two antibiotics was higher than that of single pollutants. The maximum capacities for adsorption of OTC and TC were reported to be 312 (89%) and 303 mg/g (95%), respectively. The adsorption of both pollutants followed pseudo-second-order kinetics and was better explained by the Langmuir adsorption model. π–π interaction of multiple phenolic hydroxyl groups and benzene ring structures of TC and OTC with the imidazolate rings ZIF-8 attributed to the high adsorption capacities.ZIF-67 was also used by Dehghan et al. (2019) for the adsorption of tetracycline from aqueous solutions (Dehghan et al., 2019). ZIF-67 synthesized by cobalt acetate (ZIF-67-OAC) showed the highest capacity of adsorption for tetracycline (447 mg/g; with a removal efficiency of 93.7%) among other synthesized MOFs using different cobalt sources (e.g., sulfate, chloride, nitrate, and acetate). At the end of cycle 4, the tetracycline removal performance of the regenerated adsorbent was maintained relatively unchanged. Freundlich model and pseudo-second-order kinetic model could better explain the adsorption by ZIF-67-OAC. The adsorption could be attributed to Van der Waals forces attraction and HB between polar tetracycline groups and groups on the surface of ZIF-67-OAC. Large organic contaminants can be challenging to remove from the aqueous media via microporous MOFs.Antibiotic chloramphenicol is commonly prescribed for treating a variety of bacterial illnesses. The widespread usage of such a beneficial medication, on the other hand, can result in significant water contamination, causing aplastic anemia and bone marrow depression in humans. To remove chloramphenicol, PCN-222 (Zr porphyrinic MOF) was successfully used (Zhao et al., 2018). Compared to various other porous materials, including carbon nanotubes, mesoporous sol-gels, and ordered mesoporous carbon, PCN-222 showed quicker sorption and a significantly higher adsorption capacity of 370 mg/g. Two interactions between the adsorbate and the adsorbent were reported, in which the structural characteristics of PCN-222 play a key role in the high chloramphenicol removal efficiency; ESI, which was facilitated by the charge differences between chloramphenicol (negatively charged) and PCN-222 (positively charged), and HB via hydroxyl groups provided by the Zr cluster of PCN-222 and different organic groups in the molecules of chloramphenicol (e.g., –OH, –NO2, and –CO). When compared to varying MOFs like MIL-101 (Cr), MIL-68 (Al), and MIL-53 (Al), PCN-222's large surface area and pores resulted in a high diffusion rate and chloramphenicol removal performance. As coexisting cations and anions, various common inorganic salts (e.g., KCl, NaCl, Na2SO4, and NaNO3) were used to test PCN-222's adsorption ability in actual water. PCN-222 showed an adsorption capability of 250 mg/g in solution with coexisting salts.Gao et al. (2019), in a study for degradation and adsorption of diclofenac, prepared PCN-134 (a mixed-ligand Zr-MOF), comprised of TCPP, (tetrakis(4-carboxyphenyl)porphyrin), as pillar ligand and BTB (benzene-1,3,5-tribenzoate) ligands in the 2D layer (Gao et al., 2019b). The adsorptive removal studies showed a maximum adsorption capacity of 604 mg/g for diclofenac via PCN-134, considerably more significant than that of MIL-101 (Cr), activated carbon, and anion exchange resin Amberlite IRA 67. Sun et al. (2019) synthesized composite microspheres of calcium alginate/MOFs and tested its adsorption performance toward levofloxacin, a widely prescribed antibiotic, from water (Sun et al., 2019). The observed adsorption capacity (86.43 mg/g) was found to be considerably higher than that of calcium alginate or individual UiO-66. The reusability testing, following five cycles of levofloxacin adsorption, revealed more than 70% levofloxacin adsorption.In an investigation by Jun et al. (2019), a new MOF named Basolite A100 outperformed commercial activated carbon in removing ibuprofen and CBZ fitting pseudo-second-order kinetics (Jun et al., 2019). The potential adsorption processes of MOFs were ascribed mostly to hydrophobic interactions, with contributions from HB and ESI. Furthermore, the recycling and regeneration properties of the MOF were examined for four continuous cycles to confirm its feasibility in wastewater remediation. C. Liu et al. (2019); N. Liu et al. (2019); W. Liu et al. (2019) used a three-dimensional porous and water-stable Cu (II)-based MOF to investigate the adsorption ability of three common personal care items and pharmaceutically active medicines, including chlorpromazine hydrochloride (CLF), amodiaquine dihydrochloride (ADQ), and diclofenac sodium (DCF) (W. C. Liu et al., 2019; N. Liu et al., 2019; W. Liu et al., 2019). As reported by the authors, the Cu (II)-based MOF effectively eliminated 650 mg/g DCF from the water sample. In comparison, only 67 and 72 mg/g of adsorption was reported for CLF and ADQ, respectively. The small-sized DCF molecules enable it to enter the Cu (BTTA) pores, where it interacts weakly with the open metal sites of Cu2 + and the N-atoms of the triazole ring. The adsorption of diclofenac sodium over the MOF followed the Freundlich model and pseudo-first-order kinetics.Synthetic dye pollution is becoming a rising environmental concern, as many dyes are hazardous to humans and aquatic life. More than 10,000 tons of dyes are used by textile companies globally each year, with around 5000 tons of these dyes and 3600 tons of various wastes containing high concentrations of dyes being discharged into water streams (Murugesan et al., 2021; Renita et al., 2021; Yagub et al., 2012).Many dyes present in industrial wastewater are poisonous, carcinogenic, and teratogenic (Liang et al., 2018).Various processes have been used to remove dyes from industrial effluents, including adsorption, coagulation, advanced oxidation, and membrane separation (Wong et al., 2020).Along with its ease of use and great efficiency, adsorption is regarded as one of the most influential modern wastewater treatments for the removal of toxic organic contaminants such as dye in effluents (Abhinaya et al., 2021; Akpinar and Yazaydin, 2017, p. 67; Sharma et al., 2021; Ullah et al., 2021). ( Fig. 18).MOFs have been used as adsorbents for dye contaminants in liquid-phase extractions (Ahmed et al., 2017; Khan et al., 2018; Uddin et al., 2021; Yu et al., 2018; Y. H. Zhang et al., 2019; S. Zhang et al., 2019; X. Zhang et al., 2019; Y. Zhang et al., 2019), solid-phase extractions (Liu et al., 2021), solid-phase micro-extractions (González-Hernández et al., 2021), and high-performance liquid chromatography (Aqel et al., 2021). Due to a large number of recent studies on the use of metal-organic frameworks for dye adsorption, this review refrains of deep-diving into the MOF-based dye adsorption processes. Table 8, however, summarizes some of the works that have been reported on using MOFs for the removal of different.This review has explored recent advances and discussed the potential of using MOFs and MOF-based composites for the removal of POPs from contaminated water and wastewater streams. From the literature discussed in this review, it can be concluded that MOFs, due to the tunability of their structural and electrical properties, are considered promising materials for the removal of POPs. They can be used not only as effective adsorbents but also as very efficient catalysts for removing and degrading POPs in contaminated water. The pore size of MOF can be tuned to accommodate targeted contaminant molecules. MOFs can also be functionalized to improve electrostatic, acid-base, π–π interactions, or hydrogen bonding. They can be coupled with metals, inorganic semiconductors, or organic linkers to increase their photoexcitation rate and decrease electron-hole recombination, resulting in composites with high photocatalytic efficiency.Photocatalysts hold the promise of low-cost, environmentally friendly pollution control that required little to no energy as they do not release harmful residues or change organic contaminants from one phase to another. However, photocatalysts are still limited by their inherently low chemical activity, which is not always offset by a large number of active sites. This limitation has two significant consequences: photocatalysts may produce undesirable byproducts due to an incomplete reaction or during degradation processes, in addition to having a limited ability to remove pollution.Although studies have documented that photocatalytic oxidation does not always completely oxidize organics to CO2 and H2O, only a few studies have looked at the byproducts. According to (Selishchev et al., 2017) research, CO was produced during the photocatalytic oxidation of VOCs in the presence of CO2 and H2O. Mo et al. (2009) discovered that humidity and VOC content affected the production and concentration of byproducts in a study of toluene degradation. S-methyl-methanethiosulfonate and methane thiol were synthesized from sulfur-containing precursors by Yao and Feilberg (2015). Wang et al. (2011) investigated the decomposition of DMS on TiO2, identifying the products using chromatography without mentioning CH2O concentrations. Formaldehyde synthesis from S-doped TiO2 photocatalysts was observed, but no yield was reported. It must be noted that some byproducts of photocatalytic degradation of POPs can be harmful chemicals. Therefore, the toxicity of the (by)products should constantly be tested, if possible. Given the significance of the subject matter, it is obvious that more future research should study byproducts produced by the catalytic photodegradation of POPs (Yu et al., 2021).MOFs for photocatalysis:Based on the literature reviewed in this study, MOFs show great promise for photocatalysis. In general: i. MOFs are great but sometimes limited by low activity despite having a large surface area/a large number of sites. ii. Absorption can be tuned by ligand; One of the benefits is the use of ligands to render the MOFs active under visible light, however it seems that most literature mainly focused on UV irradiation. Moreover, the performance of MOFs still subpar than most inorganic semiconductors because of their low photo responsivity, low efficiency of visible light utilization, and fast recombination of photo-generated electron-hole pairs. iii. Low charge recombination and low chemical activity can be addressed by adding MNPs as guest compounds and by designing MOF composites MOFs are great but sometimes limited by low activity despite having a large surface area/a large number of sites.Absorption can be tuned by ligand; One of the benefits is the use of ligands to render the MOFs active under visible light, however it seems that most literature mainly focused on UV irradiation. Moreover, the performance of MOFs still subpar than most inorganic semiconductors because of their low photo responsivity, low efficiency of visible light utilization, and fast recombination of photo-generated electron-hole pairs.Low charge recombination and low chemical activity can be addressed by adding MNPs as guest compounds and by designing MOF compositesTherefore, MOF-based hybrid (composite) materials have gained more attention in recent years. Efforts have been made to increase MOF photocatalytic activity and solar light-harvesting capacity through different methods, including combing MOFs with other semiconductors (heterojunction photocatalysts). Moreover, MOFs unrivaled adaptability provides a variety of ways to alter and regulate pure MOFs for improved photocatalytic activity in visible light. Integration of MOFs with light-harvesting semiconductor materials such as CdS and Fe3O4, gCN and In2S3 to form hybrid materials has been recognized as an efficient way for the production of effective photocatalyst with good light absorption and photocatalytic efficiency. Compared with MOFs themselves, the hybrids display significant benefits because of their synergistic effect (Ding et al., 2017; Huo et al., 2019; H. Zhang et al., 2019; S. Zhang et al., 2019; Y. Zhang et al., 2019; X. Zhang et al., 2019).MOFs for adsorption:According to the literature reviewed in this paper, MOF-based adsorptive processes for the removal of POPs are very promising, particularly because our knowledge of making water-resistant MOFs with high adsorption capacity is advancing quickly. So far, the MOF-based adsorption technique is very promising in treating small to medium volumes of water with low contamination levels. That said, however, given the cost associated with manufacturing such MOFS is still high, there is still a long path toward commercializing such technologies. While working toward the development of inexpensive yet efficient MOFs should be the focus of future research in this field, researchers should work on the development of facile regeneration techniques to recover and reuse saturated MOF-based adsorbents. It is noteworthy that poor selectivity of some MOF-based adsorbents for removal of some organic contaminants is a challenge that needs additional research. The suitability of the applications in a batch reactor or fixed-bed column reactor, reusability, lifespan of materials, the cost of regeneration of spent materials, the ease of post-treatment, and the environmental impact of exhausted/used adsorbents/catalysts should be considered in the design of larger-scale treatment processes (Li et al., 2022).Despite the fact that scientists have had significant advancements in the manufacturing and modification of MOF-based materials as adsorbents and catalysis, however, synthesis of many highly efficient MOFs are still a very energy-intensive and expensive process. Therefore, more research is needed on optimizing the manufacturing processes of MOF-based compounds using inexpensive materials and methods. In addition, strong hydrodynamics, water turbulence, and water flow scouring, especially for unshaped powder-like MOFs, could cause leakage of nano-sized MOFs currently in use. Similar to the well-known high toxicity of nanoparticles (such as silver nanoparticles), nano-sized MOFs may be harmful to humans and other living things (Li et al., 2022).This report also discussed MOFs as adsorbents that allow easy separation and constant recycling in water remediation. Although the adsorption technique has numerous benefits over photocatalysis, theoretically, photodegradation is a better approach as it results in the entire pollutant elimination and the need for subsequent treatment.Challenge for both applications: i. Kinetic restrictions by reactant diffusion can be managed by introducing mesoporous, but ligand chemistry is also important, as hydrophobic/hydrophilic ligands will affect access of POPs to the active sites. Strategies such as selective ligand removal (Naghdi et al., 2022), (e.g. via selective oxidation, thermal degradation or dissolution) to design novel hierarchical microporous-mesoporous MOFs that facilitate reactant diffusion as well as induce uncoordinated centers as potential catalytic sites and new adsorbent sites for water purification can provide a new tool for the purposeful engineering of hierarchical MOFs with advanced applicability in liquid media. ii. Low stability can be solved by mixed-ligand approach (refer to Section 2.1.1). Kinetic restrictions by reactant diffusion can be managed by introducing mesoporous, but ligand chemistry is also important, as hydrophobic/hydrophilic ligands will affect access of POPs to the active sites. Strategies such as selective ligand removal (Naghdi et al., 2022), (e.g. via selective oxidation, thermal degradation or dissolution) to design novel hierarchical microporous-mesoporous MOFs that facilitate reactant diffusion as well as induce uncoordinated centers as potential catalytic sites and new adsorbent sites for water purification can provide a new tool for the purposeful engineering of hierarchical MOFs with advanced applicability in liquid media.Low stability can be solved by mixed-ligand approach (refer to Section 2.1.1).Despite the MOFs' promising prospects, various concerns must be addressed, especially in large-scale and real-life practical situations such as developing strategies to use thermodynamically water-stable MOFs with a stronger coordination bond or kinetically water-stable MOFs with a better hydrophobic coat, MOFs with redox-active metals and/or organic functionalized ligands, MOFs for adsorptive removal of gaseous pollutants, understanding of the mechanism of adsorptive reduction of POPs, the influence of pH, temperature, and solute ions, the MOFs effectiveness against a wide range of organic pollutants found in real-life applications, focus on low-bandgap MOFs or employing different techniques (e.g., doping, nanocomposite formation) to improve the bandgap of the MOF to make it more appropriate for absorbing visible light. Although there are several publications on water-resistant MOFs, such as the UiO-66, MIL-125, MIL-101, and ZIF series, the stability under extreme circumstances (strong acidic and alkaline pH) is still lacking.Lastly, another challenge comes from the vast number of available MOFs. Which makes it difficult to wisely choose the most promising ones for the respective POP and removal process.To conclude, innovative material design, quality control, and environmental considerations are essential with an ever-expanding potential in MOFs’ applications. Consequently, interdisciplinary research among scientists from different disciplines such as chemistry and chemical engineering, environmental science engineering, and computer science (modeling) is undoubtedly significant to developing MOF science and technology and commercializing MOF-based processes.POPs are among the most hazardous materials released, intentionally or unintentionally, to water sources due to human activities and resist environmental degradation. Due to harmful impact of POPs on wildlife and human beings it is crucial to explore and develop effective strategies for POP removal from water/wastewater. MOFs possess unique features which make them a proper potential for this purpose. Photocatalytic degradation and adsorption of POPs by MOFs has drawn scientific attraction as a potential solution. Resealing scientific results discussing the advances, challenges and improvement strategies in this field could open up a new door for future researches.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.D. Eder and S. Naghdi acknowledge the support of the Austrian Science Fund (FWF, I 5413-N) H. Kazemian and H. Djahaniani acknowledge the support of the Natural Sciences and Engineering Research Council of Canada (NSERC) through Discovery Grant [RGPIN-2019-06304].

The presence of persistent organic pollutants (POPs) in the aquatic environment is causing widespread concern due to their bioaccumulation, toxicity, and possible environmental risk. These contaminants are produced daily in large quantities and released into water bodies. Traditional wastewater treatment plants are ineffective at degrading these pollutants. As a result, the development of long-term and effective POP removal techniques is critical. In water, adsorption removal and photocatalytic degradation of POPs have been identified as energy and cost-efficient solutions. Both technologies have received a lot of attention for their efforts to treat the world's wastewater. Photocatalytic removal of POPs is a promising, effective, and long-lasting method, while adsorption removal of persistent POPs represents a simple, practical method, particularly in decentralized systems and isolated areas. It is critical to develop new adsorbents/photocatalysts with the desired structure, tunable chemistry, and maximum adsorption sites for highly efficient removal of POPs. As a class of recently created multifunctional porous materials, Metal-organic frameworks (MOFs) offer tremendous prospects in adsorptive removal and photocatalytic degradation of POPs for water remediation. This review defines POPs and discusses current research on adsorptive and photocatalytic POP removal using emerging MOFs for each type of POPs.

Massive dearth for industrialization, globalization and civilization has brought annihilation of natural resources to liberal extents. Fresh water, fossil fuels, atmosphere and forests encounter rapid pollution due to man-made actions. Discharge of untreated/poorly treated textile wastewater into water bodies is one such event that causes catastrophic alarms. More than 15% of the dye produced globally is found to be lost in the dyeing processes and escape through the drains [1]. In addition to these unfixed dyes, the effluents are also concentrated with huge amounts of inorganic salts, surfactants and alkalis [2]. Their contrary impact on chemical and biological oxygen demand (COD and BOD), total organic carbon (TOC), salinity, turbidity, suspended solids, color intensity, toxicity and pH pose severe threat not only to mammalian and aquatic lives but to the entire ecosystem [3]. Myrna Solís et al. [4] professed that every year nearly 280,000 tons textile dyes are liquidated in the form of industrial wastewater. This marked the textile industry to be one of the foremost contributor of enviro-pollution.Commercially used synthetic dyes are categorized into three types (i) Cationic (basic dyes), (ii) Anionic (direct, disperse, metal complex, acid & reactive dyes) and (iii) Non-ionic dyes. They can also be classified in terms of their chemical structure as azo, anthraquinone, triaryl methane, sulfur linkages etc [5,6]. Among these the azo dyes containing azo (-NN) and sulphonic (-SO3-) linkages have fascinated the textile industries with its excellent structural stability, amplified conjugation, stronger color intensity and high aromatic nature. Consequently, 60–70% dyes manufactured around the world are only azo dyes. On the contradictory side the wastewater discharge from azo dyeing processes are found to have negligible biodegradability owing to its high molecular weight and complex structure [7] thus leading to certain ecological conflicts. They increase the turbidity of water and obstruct the sunlight from entering into the water bodies thereby interrupting the growth of flora and fauna. This deficiency of sunlight induces the micro-organisms and hydrophytes to decompose and dissipate foul smell to the atmosphere. Also, heavy metals like Pb, Cu, Cd, Cr, As and Ni dispersed in the contaminated water bio accumulate and absorbed by aquatic life which then directly affect humans over consumption. Carcinogenic and mutagenic effluent-contaminated water triggers several health issues over persistent exposure and ingestion. Some major conditions are cancer, kidney failure, metabolic stress, heart diseases, hepatocellular carcinoma, dermatitis, splenic sarcoma, emphysema, dysfunction of central nervous system and reproductive system etc [8,9].Since industrial dyes are tailored to withstand harsh environments like sunlight, humidity, chemical corrosion, wear and tear resistance etc they cannot be removed or degraded into non-toxic fragments naturally. Along decades there are three different methods of approach for the removal of dye from the effluent wastewater: (i) Biological techniques (algae degradation, enzyme degradation, aerobic & anaerobic remediation, fungal decolorization microbial cultures etc), (ii) Physical techniques (adsorption, reverse osmosis, membrane filtration, coagulation & flocculation, nano & ultra-filtration etc) and (iii) Chemical techniques (ozonation, Fenton's oxidation, electrochemical destruction, electro-kinetic coagulation, photochemical etc) [10,11]. Amidst the various techniques, photocatalytic degradation takes a crucial advantage over the other techniques by its efficient, sludge-free, chemical-free and ecofriendly process for the removal of toxic dyes from industrial effluents. Photocatalytic degradation technique works on the basis of photo-inducing reactive oxygen species (ROS) such as superoxide radicals, singlet oxygen and hydroxide radicals by means of a photocatalyst to strike the organic dye moiety and break it to harmless end-products like water and carbon dioxide [12].Over the years semiconductor photocatalyst materials have gained augmented attention in enviro-remediation. Some frequently used photocatalysts are TiO2, WO3, NiO, SnO2, CuO, Cu2O, ZnO, Bi2O3, Al2O3 and Fe2O3 [13]. Munir Ahmad et al. [14] fabricated ZnO and gold decorated ZnO nanoparticles via green synthesis involving pecan nut leaf extract as the reducing agent. The enhanced Au-decorated ZnO nanoparticles showed 95% degradation against rhodamine B. Hematite (α-Fe2O3) nanoparticles were created in two forms (nanorods & nanocubes) and enhanced with gold decoration by Emre Alp et al. [15]. These nanoparticles exhibited almost complete degradation of rhodamine B in addition to its lethal effects on E.coli bacteria. Owing to their non-toxic nature, tunable bandgap, chemical stability, versatile structure and cost effectiveness TiO2 has proved to be a proficient candidate for photocatalysis. But due to its narrow absorption in the visible light region the need for doping it with other metal oxides is a necessity to extend its applicability in sunlight [16,17]. Researchers are engaging consistent efforts in tuning TiO2 to widen its bandgap range and enhance the photocatalytic activity. Ravi Kumar Mulpuri et al. [18] co-doped zinc and boron to TiO2 nanocatalyst via sol–gel route and incorporated it for the photodegradation of acid red 6A (AR6A). Zhongming Liu et al. [19] put forth the template-assisted synthesis of xylan/PVA/TiO2 composite and its efficiency in photocatalytic degradation of ethyl violet and astrazon brilliant red 4G dyes with 94% degradation rates.This work emphases in embracing a unique solvent system acknowledged as the first-ever deep eutectic solvent (DES) discovered by A. P. Abbott et al. [20], choline chloride:urea in 1:2 ratio. DES's are well-known for their biodegradability, high polarity and dielectric constant and exclusively for assisting material synthesis. We report a conventional solid-state synthesis mediated with DES to prepare calcium-based TiO2 composite ceramics and further doping it with lanthanum to study the effect of rare-earth dopant on photocatalytic activity. Furthermore, adaptation of the ceramics as photocatalyst under simulated solar radiation against reactive black 5 (RB5), reactive red 198 (RR198) and reactive yellow 145 (RY145) will be reconnoitered.Choline chloride, urea, calcium oxide, lanthanum oxide and titanium dioxide were all purchased and used without further purification from Sigma. Industrial dyes reactive black 5, reactive red 198 and reactive yellow 145 were also procured from Sigma and their general data is listed in Table 1 . Doubly distilled water with neutral pH was used for making up and diluting the dye solutions.Firstly, the DES solvent medium was prepared by mixing 1:2 ratio of choline chloride and urea at 80 °C. Then, a finely ground mixture of the metal oxide precursors titanium dioxide and calcium oxide was added to the DES solution under simultaneous stirring. The resultant homogenous mixture was calcined at 800 °C for an hour to achieve calcium titanate. For the lanthanum doped photocatalyst, lanthanum oxide precursor was added to the above mixture in adequate amount. Finally the obtained pure and La-doped calcium titanate photocatalysts (LaxCa1-xTiO3, x = 0.0 & 0.5) were labelled as pure CTO and La-doped CTO followed by vacuum storage.The diffraction patterns of pure and La-doped CTO photocatalysts were studied using X-ray diffraction and are as shown in Fig. 1 . The pattern obtained for pure CTO was indexed to the orthorhombic crystal phase of CaTiO3 which matched well with the standard JCPDS 78–1013. Major diffraction peaks occurred at 2θ 23.3°, 33.2°, 47.6°, 59.5° and 69.7° were expressive of the planes (101), (100), (050), (042) and (242). Whereas, pattern of the La-doped CTO exposed the advent of some peaks in addition to that of the pure CTO peaks referring to the fusion of lanthanum in the CTO lattice. These peaks were characteristic to the presence of monoclinic La2Ti2O7 in accord with JCPDS 27–1182. Some prominent peaks of La-doped CTO were 2θ 27.6°, 37.8°, 38.7°, 40.7° and 48.1° pertaining to (400), (031), (220), (022) and (−104). The peak at 2θ 47.6° (pure CTO) has partially split and gave rise to two peaks at 2θ 47.4° and 48.1° (La-doped CTO), which showed the successful incorporation and formation of lanthanum in the pure CTO lattice. Both the patterns had sharp intense peaks exhibiting their crystalline nature. Also, the presence of precursor materials like CaO, TiO2 and La2O3 were not detected on both the photocatalysts showing its high degree of phase purity. Based on the Debye–Scherrer's formulations the crystallite size of pure and La-doped CTO photocatalysts were found as 45 and 22 nm. The inclusion of lanthanum ions has restricted the grain growth and also led to the increment of lattice strain from 0.00216 to 0.00252 [21]. This parallel behaviour of decreasing crystallite size and increasing lattice strain demonstrates the dopant's presence in the lattice.FT-IR spectrum exposed the type of functional groups and bonds present in pure and La-doped CTO photocatalysts and are displayed in Fig. 2 . The peaks at 490 and 643 cm−1 were indicative of the Ti–O and O–Ti–O bonds present in pure CTO photocatalyst which were also expressed at 478, 584 and 740 cm−1 for the La-doped CTO photocatalyst [22]. The act of doping has transformed the broad peaks into refined peaks when compared to the pure photocatalyst. Likewise, both the photocatalyst had a peak at 3420 cm−1 pertaining to O–H stretching of water molecules adsorbed on the surface of the photocatalyst. Notably, none of the peaks related to organic residues from the DES solvent medium were detected ensuring the photocatalyst purity.Absorption edge and optical response of the photocatalysts were studied and are presented in Fig. 3 . Pure CTO had an optical absorption at 310 nm while that of the La-doped photocatalyst was around 360 nm. They simultaneously extended their edges into the visible wavelength region. This showcases the ability of the photocatalyst to absorb incident light not only in UV region but also in visible region and hence approves their applicability on a wide spread spectrum. Bandgap (Eg) was calculated using the Kubelka–Munk function modified Tauc relation which then gave the direct (n = 1/2) and indirect (n = 2) bandgap values of the photocatalysts [23] (Fig. 4 ). The bandgaps of pure CTO (3.2 & 3.1eV) had decreased to (2.95 & 2.8eV) illustrating the incidence of Burstein-Moss shift [24]. This decline in Eg values of both the direct and indirect bandgaps depicted the increased generation of electron–hole pairs thereby the greater photocatalytic activity of La-doped photocatalyst.Photoluminescence emission, electron–hole recombination rate and optical properties of pure and La-doped CTO were studied using photoluminescence spectroscopy technique. The studies were conducted at an excitation wavelength of 320 nm and is exhibited in Fig. 5 . It was clearly observed that the emission spectra of pure CTO was elevated in comparison to the La-doped CTO photocatalyst. This lowered photoluminescence emission intensity of the La-doped CTO directly indicates the depressed electron–hole recombination rate of the photocatalytic material which occurs due to the instance of doping lanthanum. The low recombination rate results in increased mobility of the electrons which attacks dye molecules through the formation of hydroxyl radicals. Hence La-doped CTO photocatalyst can effectively ease the process of dye degradation.Surface characteristics of the photocatalyst were analyzed using BET analysis technique. Their adsorption–desorption isotherms are as exposed in Fig. 6 . Based on the Brunauer's classification of surface properties pure and La-doped CTO are categorized under the type II and type III isotherms [25]. Surface area of pure CTO was 6.28 m2g-1 while that of the La-doped photocatalyst was 32.84 m2g-1. Whereas the pore size decreased from 4.035 nm to 1.564 nm and exposed the transition from macroporous to non-porous nature [26]. This inverse proportionality of the surface area and pore size is a rare phenomenon where the non-porous La-doped CTO photocatalyst compensates its low porosity with its higher surface area and oxygen vacancies. Thus La-doped CTO photocatalyst proves to be a superior candidate for photocatalytic activity.Transmission electron micrographs showed the absolute morphological representation of the photocatalysts (Fig. 7 ). Pure CTO was found to form pillar-like structures whereas the La-doped CTO had broken pillar-like morphologies. This clearly explained the advent of doping a larger ion (La3+) with ionic radius 1.032 Å in the A-site has brought about destruction in the structural level causing an increase in the photocatalytic active surface.Pure and La-doped CTO were examined for photocatalytic activity towards 50 ppm RB5 solution with 1 mg catalyst/ml loading. Solar light imitated in a photoreactor aided as the UV–Vis light source to trigger the photocatalytic activity of the catalysts. A time-dependent spectrum was taken to analyze the degradation of RB5 when kept in contact with the photocatalysts. In a continuous experiment of 120 min duration aliquots were collected at an interval of 10 min and were centrifuged to get rid of any micro particles present. The absorbance spectra displayed in Fig. 8 pictured the superior degradation of RB5 in La-doped CTO catalytic surface in comparison to that of pure CTO. This can be attributed to the increased surface area, porosity and surface defects which arose due to the La3+ infusion in the lattice sites of the La-doped CTO. As a result La-doped CTO is evidenced to be enhanced and effective among the two catalysts and was selected for further studies.When exposed to UV–Vis light an electron (e - ) from the valence band of the photocatalyst is excited to the conduction band leaving a positive hole (h+) behind. Consequently, the credible mechanism of photocatalytic degradation (Fig. 9 ) can take place in two ways: photo-induced splitting of water molecules to form hydroxyl radicals (OH - ) by means of h+ and reduction of surface adsorbed oxygen molecules to form superoxide anions (O2 .- ) which in turn forms hydroxyl radicals by means of e - . However, the highly reactive hydroxyl radicals formed either way exclusively attacks the organic dye moiety forcing it to breakdown into smaller intermediates and finally to non-hazardous carbon dioxide (CO2) and water (H2O) [27]. Also, it is worth mentioning that due to the absence of chemical stimulants like hydrogen peroxide, sodium hypochlorite, ozone etc the possibility of chemical oxidation is completely eliminated. Thus, UV–Vis light stands as the only external source to trigger the photocatalytic degradation of the dyes.La-doped CTO catalyst was tested for its photocatalytic activity on RB5 in a wide range of dye concentration to find out its effectiveness over highly concentrated real-time textile effluents. Degradation efficiency was estimated using the formula: % Degradation efficiency= (C0–C/C0) x100 where C0 is the absorbance at time = 0 and C is the absorbance at time = t. Figure 10 showcased the gradual decrement of photocatalytic efficiency from 85.6 to 59.4% when the dye solution was concentrated from 25 to 100 ppm. This occurrence at higher concentrations is due to the increasing hindrance of dye molecules which obstruct the photons from reaching the catalytic surface and therefore depressing the generation of hydroxyl radicals. Since only fewer hydroxyl radicals are generated to breakdown more number of dye molecules a drop in efficiency takes place [28]. However, La-doped CTO had a moderate efficiency (72.6%) over 50 ppm RB5 and was preferred for forthcoming scrutiny with dyes of similar properties.The photocatalytic efficiency of La-doped CTO photocatalyst was efficaciously examined with 50 ppm RB5 and was also evaluated with two more reactive anionic dyes, RR198 and RY145. From Fig. 11 it is clear that the degradation efficiency increased in the order of RB5˂RR198˂RY145. Additionally, the efficiency didn't reach a plateau which resembled that the catalyst is capable of progressing towards complete destruction of the dye provided extra duration. The catalytic surface expressed 86.2% efficiency in RR198 whereas 97.8% efficiency in RY145 solution. Exclusively, the catalyst was able to degrade almost complete RY145 in just 60 min in spite of its higher molecular weight. This confirms that La-doped CTO is highly favored for RY145 remediation from the environment.The order of the degradation process and its kinetic parameters were found by applying the relation: ln (C0/C) = k∗t where C0 is the initial concentration, C is the concentration at time t of the dye and k∗ is the rate constant of the reaction. The linear fit to the plot of ln (C0/C) verses time (Fig. 12 ) resulted in a linear graph proving that the reaction kinetics falls under the pseudo first order kinetics category. The rate constant, correlation coefficient (R2) and the calculated concentration of dye at equilibrium time (Ceq cal) found from the graph are charted in Table 2 . Since La-doped CTO catalyst has the ability of degradation beyond 120 min a stable saturation level has not been attained in all the three dye degradations. Hence estimating the experimental concentration of dye at equilibrium time (Ceq exp) is not feasible. But the Ceq cal values attest that while attaining equilibrium the concentration of the dyes would be very negligible. Thus complete degradation of dye is once more substantiated strongly.pH of the dye solution plays a chief pivotal role in the photocatalytic degradation process. It can either serve as a driving force or an impeding force towards the degradation mechanism. Degradation taking place at neutral pH varies drastically when the pH of the medium is altered towards acidic or basic scale. Influence of pH over dye molecules mainly depend on functional groups, molecular weight and the type of dye being degraded besides the nature of the catalyst. Also it is noteworthy that the three anionic dyes are weak acids containing number of sulphonate functional groups (-SO3 - ) which favors the dye molecules to be negatively charged [29]. The variation in degradation efficiency of RB5, RR198 and RY145 with respect to the pH of the solution is disclosed in Fig. 13 . Overall, in all the three dyes it is noticeable that in comparison with the neutral pH the efficiency is amplified in acidic medium and declined in basic medium. As La-doped CTO is a titanium dioxide based material it can have amphoteric nature. Due to this fact, at acidic pH the catalyst is positively charged and has greater electrostatic affinity towards the anionic dye species while at basic pH the catalyst becomes negatively charged leading to repulsion in contact with the anionic dye. This explicates the change in efficiencies at different pH [30].The reusability of a catalyst after several number of photocatalytic cycles is a crucial property of a proficient catalyst. It reinforces the existence of ample number of surface active sites in the catalyst and the possibility of rapid seizing/relinquishing dye molecules in the porous cavities of the catalytic surface. Due to such significance, stability and recyclability of La-doped CTO catalyst was verified with RB5, RR198 and RY145 solutions (Fig. 14 ). Degradation efficiencies remained almost constant up to five cycles for all three dye degradations. The photocatalyst showed 82.4, 90.6 and 92.6% average efficiencies for RB5, RR198 and RY145 respectively ensuring its active usage for more cycles.Collectively in summary, pure and La-doped CTO were synthesized using a conventional solid-state technique which was mediated by a DES medium (choline chloride:urea). The synthesized materials were applied as photocatalysts for aiding the photocatalytic degradation of RB5, RR198 and RR198. The deviation of degradation efficiencies were noted in terms of type of catalyst, concentration of dye, type of dye and pH. Kinetics of the reaction order and the recycling capacity of the catalyst were also examined to strengthen the potential activity of the photocatalyst.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 Deanship of Scientific Research (DSR) at King Abdulaziz University, Jeddah Saudi Arabia has funded this project, under grant number (KEP-40-130-42). This work is sustained by the support rendered by the DST-FIST facility (CRIST), Stella Maris College, Chennai - 600086. The author Ayyar Manikandan is thankful to Tamil Nadu State Council for Science and Technology (TNSCST), DOTE Campus, Chennai for the financial support (S&T Project: TNSCST/ STP-PRG/AR/2018-2019/9307). The authors are also thankful to Ms. Jeena N Baby and Dr. Raghu Subashchandrabose for their research assistance and support.

Water serves as a key component in the basis of life to function and evolve unceasingly. Excessive deterioration of fresh water resources through discarding enormous amounts of dye effluents is becoming a problem of concern. In this work we report a conventional solid-state technique improved with choline chloride: urea DES medium in 1:2 ratio for synthesizing pure calcium titanate (CTO) and lanthanum doped calcium titanate (LaxCa1-xTiO3; x = 0.0 and 0.5). Both the materials were subjected to photocatalytically degrade reactive dyes RB5, RR198 and RY145. La-doped calcium titanate showed effective degradation efficiency when compared to the pure calcium titanate. On increasing the concentration of the dye, the efficiency of degradation declined gradually. Also, the efficiency was assessed to be greatest for RY145 followed by RR198 and lastly RB5. Efficiency dependence on different pH exposed the enhanced degradation in acidic medium and comparatively lowered pH in basic medium. Recyclability was highly satisfying and promising to validate the real-time applicability of the catalyst.

Zeolites are crystalline and microporous materials built mainly of silicon, aluminum, and oxygen. The occurrence of Brønsted acid sites inside micropores allows to use this group of aluminosilicates as catalysts in numerous industrial processes [1,2]. However, due to a limited accessibility of the active sites and slow transport of reactants in micropores, which can lead to rapid deactivation of catalyst, the application of this groups of minerals as catalysts is not fully satisfying [3-5]. The formation of mesopores via alkaline treatment of zeolites seems to resolve this problem [6-8].One of the routes in the synthesis of mesoporous materials is the application of ultrasonic irradiation. This technique allows to reduce the duration procedure, the use of milder pressure and temperature conditions and may cause the limitation (or even elimination) of the utilization of expensive and toxic reagents, which can also result in the decrease of synthesis costs. The phenomenon of the replacement of conventional desilication of microporous zeolites (in alkaline media) with ultrasonic technique is based on the cavitation mechanism of ultrasounds propagation and the formation of local “spots” of ultrahigh temperature (5000 K) and pressure (1000 bar). The cavitation leads to collisions of particles moving at high velocities, promoting the formation of radicals triggering sonochemical reactions [9,10].The application of ultrasounds indicated beneficial effects in syntheses of zeolites: CHA [11], A [12], NaP [13], bilikalite [14], MCM-22 [15] and RHO-type zeolites [16]. Generally, ultrasonic-assisted synthesis of zeolites caused the generation of products in a shorter time, which presented an improved crystallinity degree and smaller crystal sizes.Ultrasonic irradiation was also used in the preparation of single metal oxides, such as: ZrO2 [17], TiO2 [18], MnO2 [19], Cu2O [20], CeO2 [21] and ZnO [22]. Furthermore, ultrasonic-assisted procedure of the synthesis of mixed oxides was reported for: BayZr3-yTiO3 [23], NiO/Al2O3 [24], Ce0.5Zr0.5O2 [25], CuO/ZnO/ZrO2/Al2O3 [26] and Ni-Co/Al2O3-ZrO2 [27].Ultrasonic-assisted modification of zeolites was also reported in literature. Hosseini et al. [28] performed dealumination of zeolite Y in ethanol-acetylacetone solution as a chelating agent both in the absence and presence of ultrasounds. It was shown that the sonochemical-assisted modification of zeolite samples resulted in a higher aluminum extraction from zeolite framework than for the materials prepared conventionally.Zhang et al. [29] obtained mesoporous FAU-type zeolites via chemical dealumination (using citric acid and H4EDTA aqueous solutions) and a subsequent ultrasonic-assisted alkaline treatment in aqueous NaOH solutions. The use of ultrasounds accelerated the formation of mesoporosity in respect to the analogues prepared traditionally. Similar observations were reported by Oruji et al. [30], who synthesized mesoporous FAU-type zeolites in sodium form. Based on porosity studies, it was shown that the rising duration of base-wash procedure led to gradual increase of mesoporosity with a simultaneous decrease of crystallinity.In turn, Kuterasiński et al. [31] investigated the effect of ultrasonic-assisted desilication of commercial FAU-type zeolite. The prepared samples have been used as catalysts for the decarbonylation of furfural into furan. It was shown that the application of high-frequency ultrasounds during alkaline treatment procedure caused higher mesoporosity and enhanced catalytic properties in respect to the catalysts modified under conventional desilication conditions.Khoshbin and Karimzadeh [32] prepared mesoporous ZSM-5 zeolite via ultrasonic-assisted desilication of parent zeolite, which was synthesized using a rice husk ash as a silica source and various contents of carbon nanotubes (between 0 and 30 wt% of CNTs) playing a role of hard template. It was evidenced that increasing amount of carbon nanotubes in precursor mixture caused the increase of both external surface area and mesopore volume of such prepared ZSM-5-type zeolite product.In this study, we present an ultrasound-assisted desilication of MFI-type zeolite, which can be used as a catalyst for the dehydration of ethanol into diethyl ether (DEE) and ethylene. DEE can be used in pharmaceutics, explosives and in petrochemistry [33-35]. In turn, ethylene undergoing polymerisation, oligomerisation, hydrogenation, halogenation, oxidation and a lot of other reactions, has also a great meaning in many industrial processes [36-40].In this study, we applied commercial zeolite of MFI type structure (Si/Al = 40) – MFI-40. The physicochemical and catalytic properties of the zeolite-based catalysts prepared in the presence of ultrasounds were compared with those obtained under conventional treatment. Presented research results constitute a precious enrichment of the knowledge concerning the synthesis of hierarchical materials.The parent MFI-type zeolite (Si/Al = 40) from Zeolyst (CBV 8014) was used as a reference sample. Ultrasonic-assisted desilication was performed using 6 g of zeolite and 200 ml of 0.2 M aqueous solutions of the pure sodium hydroxide (NaOH) or mixture of NaOH and tetrabutylammonium (TBAOH) hydroxide, which contained 10 or 70 mol% of TBAOH (i.e. 0.18 M of NaOH and 0.02 M of TBAOH or 0.06 M of NaOH and 0.14 M of TBAOH, respectively) at the same pH values (13.8).Alkaline treatment was performed for 30 min. Whole reaction system (alkaline solution, zeolite and the sonicator probe) was placed in an ice bath, which ensured low temperature. QSonica Q700 sonicator with power of 600 W and frequency of 20 kHz was used as a generator of ultrasounds. The device was equipped with a “1” diameter horn (Church Hill Rd, Newtown, CT, USA). For comparison, the conventional desilication (ultrasonic-free procedure) was carried out for 30 min also in the ice bath using alkaline solution at the same chemical compositions as above. In order to investigate a direct influence of ultrasounds on the demineralization intensity, we changed only one parameter, namely, we introduced ultrasonic irradiation into “zeolite-alkaline solution” at the same temperature of ice bath, chemical composition of alkaline solution, mass ratio of zeolite to mixture and the duration of procedure as for conventional demineralization. After desilication procedure, the suspension was fourfold centrifuged at 4000 RPM and dried overnight at 80 °C.Subsequently, desilicated zeolites were calcined in air with a flow rate of 50 ml/min for 10 h at 525 °C and with a temperature ramp of 1.5 °C/min.Afterwards, fivefold Na+→NH4 + ion-exchange of desilicated zeolites with 500 ml of 0.5 M aqueous NH4NO3 solution was performed at 80 °C for 2 h. In next step, the zeolite samples in ammonium form were centrifuged at 4000 RPM, washed, and dried at 80 °C again. Finally, the samples were calcined in 50 ml/min air flow for 8 h at 450 °C and with temperature ramp of 2 °C/min.The catalysts prepared conventionally or sonochemically were designated by the index “c” or “s”, respectively. Depending on the molar content of TBAOH in 0.2 M NaOH/TBAOH of desilication agent (0 or 10 or 70 mol% of TBAOH), the samples were named as M−0c or M−0 s, M−10c or M−10 s and M−70c or M−70 s, respectively. The parent MFI-type zeolite was denoted as M.ICP-OES chemical analysis of zeolites was carried out by the dissolution of ca. 100 mg of powder in a HF/HCl mixture in a Teflon vessel for one day. In next step, the liquid was diluted to 250 ml and both Si and Al quantitative analyses were performed using Optima 2100DV - PerkinElmer instrument.In order to determine the zeolite crystallinity, X-ray diffraction (XRD) experiments were performed using a PANalytical X’Pert PRO MPD diffractometer (40 kV and 30 mA), equipped with CuKα generator (λ = 1.5418 Å). 2θ angle was at 5–50° with a 0.033° step. The zeolite samples were in the form of powder and were placed in holders. The calculations of the average size of crystallites were performed using PANanalytical X Pert Data Viewer software connected with the diffractometer and were based on Scherrer equation (1). (1) L = λ K / β cos θ Where: ʎ corresponds to X-ray wavelength value (1.5418 Å); K is a dimensionless shape factor (0.9); β means FWHM, i.e. full width at half maximum and θ is the Bragg angle.The status of Al in the investigated samples was determined by the solid-state 27Al MAS NMR method using a Bruker Advance III 500 MHz WB spectrometer operating at 11 T of magnetic field. 27Al MAS NMR spectra were recorded at 130.3 MHz of the basic resonance frequency and at 10 kHz of a spinning rate (in zirconia rotors) with a short pulse length of 0.2 μs (π/12) and a recycle delay of 0.1 s. 1 M aqueous Al(NO3)3 was used as a reference for 27Al MAS NMR chemical shifts. Prior to NMR experiments, the samples were fully hydrated at ambient temperature in the presence of vapor-saturated Mg(NO3)2 solution.The solid-state 29Si MAS NMR spectroscopy was used in order to determine the status of silicon. 29Si MAS NMR spectra were recorded using a Bruker Advance III 500 MHz WB spectrometer operating at a magnetic field of 11.7 T and at the basic resonance frequency of 99.4 MHz, a spinning rate of 8 kHz (in zirconia rotors) with high-power proton decoupling (SPINAL64), at 5.8 μs (π/3) pulses and repetition time of 20 s. The chemical shifts of 29Si MAS NMR spectra were externally referenced to Tetramethylsilane (TMS; >99%).The porosity was determined by the low temperature sorption of nitrogen at −196 °C using Autosorb-1 Quantachrome. Specific surface area (SBET) was determined by BET model, external surface area (Sext) and volume of mesopores (Vmeso) were estimated by the application of Barrett-Joyner-Halenda (BJH) model on the adsorption branch of the isotherm. Micropore volume (Vmicro) was calculated using t-plot method. Prior to each measurement, the sample was outgassed for 20 h at 250 °C in a vacuum.The morphology of the prepared samples was investigated with a JEOL JSM – 7500F Field Emission Scanning Electron Microscope (SEM). Prior to the SEM analysis, the samples were dried for 24 h without the covering of specimens with the coating in order to enable the detailed observation of the surface of the studied materials.Transmission electron microscopy analysis (TEM) of chosen samples was performed using JEOL JEM 2100 HT LaB6 (JEOL USA, Inc., Peabody, MA, USA), with accelerated voltage of 80 kV and the spot size of 1 nm. Prior to TEM analyses, the studied materials were sprayed onto formvar film coated copper grids.The FT-IR measurements were conducted with NICOLET iS10 spectrometer (supplied by Thermo Scientific) equipped with a MCT detector. The IR spectra were recorded at 4000–650 ± 4 cm−1 with 128 scans per each spectrum. FT-IR measurements were preceded by the activation of samples (in the form of self- supporting wafers of ca. 70 mg) for 1 h at 400 °C with a temperature ramp of 5 °C/min under vacuum conditions.Quantitative analysis of acidity was performed by the IR studies of the sorption of ammonia (Air Products, 99.95%) at 120 °C and calculated based on the intensities and the extinction coefficients of the bands assigned to ammonia interacting with Brønsted and Lewis acid sites. The bands of 1450 cm−1 are attributed to ammonium ions and are characterized by the extinction coefficient of 0.12 cm2/μmol, while the maxima at 1620 cm−1 correspond to ammonia interacting with Lewis sites with the extinction coefficient of 0.026 cm2/μmol [31].The acid strength of Si-OH-Al groups was determined via CO sorption at −100 °C, followed by the comparison of the values of frequency shifts between the maxima of free acidic OH groups and OH interacting with CO by hydrogen bonding, according to the procedure described in [41].The dehydration of ethanol into diethyl ether and ethylene as a testing reaction was investigated at 150–290 °C (with a 50 °C step) in a fixed-bed glass microreactor coupled on-line with a gas chromatograph. Analysis of the products obtained in the reaction was carried out in a Perkin Elmer Clarus 580 equipped with Elite-Plot Qcapillary column (with length of 30 m and inner diameter of 0.53 mm) and TCD detector. Prior to each experiment, 100 mg of catalyst (with the granulation of 190–260 μm) was placed on a quartz wool plug in the reactor and exposed to a pure helium flow of 35 ml/min (Air Products, 5.0) at 300 °C for 30 min. Subsequently, He was passed through the liquid ethanol, generating the ethanol in helium flow with a concentration of 2.24 mmol/ml. The total reaction gas mixture stream was 35 ml/min. The weight hourly space velocity was kept at 2.0 gethanol/(gcatalyst·h).Turnover frequency (TOF, s−1) was defined, as follows: (2) TOF = n E t OH 2 / m · c BAS Where: nEt(OH)2 corresponds to the number of transformed molecules of substrate in one second (µmol/s); m is the catalyst mass (g), and cBAS is the protonic acidity concentration (µmol/g). Experimental error was not higher than 5%.Analysis of the ICP-OES results (summarized in Table 1 ) indicated that the treatment of MFI-40 with 0.2 M of NaOH solution in the absence of ultrasounds resulted in a small leaching of both Si (0.5%) and Al (0.1%) from zeolite structure. The addition of ultrasounds caused a significant acceleration in the extraction of both silicon (15.3%) and aluminium (2.6%) from MFI framework, leading to the decrease of Si/Al ratio from 37.7 to 32.4.In case of the application of demineralizing agent containing TBAOH, the introduction of ultrasonic irradiation into system “zeolite - alkaline mixture” also caused elevated removal of Si and Al from MFI-type zeolite framework. For instance, the use of 0.2 M NaOH/TBAOH solution containing 10% of TBAOH led to the increase of Si and Al extraction from 1.0% to 9.4% and from 1.0% to 5.9%, respectively. Analogous situation was found for the NaOH/TBAOH demineralizing mixture including 70% of TBAOH. The leaching of silicon was 2.3% vs. 10.8%, meanwhile the aluminium extraction was 2.3% vs. 8.6% in the absence vs. presence of ultrasonic irradiation, respectively. The alkaline treatment of MFI structure type zeolites with solutions containing TBAOH led to a slight decrease of Si/Al ratio from 37.7 to 36.2–37.5 due to simultaneous extraction of both elements (Si and Al), although the desilication took place more intensively than dealumination process.At first sight, it seems that the alkalinity of 0.2 M NaOH should be stronger than 0.2 M (NaOH/TBAOH) due to the presence of TBAOH playing the role of a protective layer on the zeolitic external surface and being a moderator in the desilication process [42-44]. However, it was found that TBAOH is able to remove aluminum from zeolite framework, which was in line with Sadowska [45] and Abello [46]. This effect rose with the TBAOH mol% content in NaOH/TBAOH demineralizing agent regardless of the way of demineralization procedure (conventional vs. ultrasonic), however, the Al leaching was more intensive under sonochemical conditions. Elevated extraction of aluminum from zeolite framework could lead to the removal of vicinal silicon atoms, which led to the production of holes, followed by the formation of “swiss cheese” type zeolite grains (see TEM analysis, Section 3.4). This effect can be observed only at low temperature (in our case: ice bath conditions). The application of much higher temperature of alkaline treatment always leads to a more intensive desilication, which was reported for the zeolites with MFI [45,47–50], MTW [50], BEA [51] and FER [52] of similar Si/Al ratio (30–50), for which removal of silicon exceeded 50%. So far, however, a direct introduction of the ultrasonic irradiation into high-temperature desilication performed under identical conditions was not found probably due to technical problems, such as the overheating of sonicator.In order to avoid the overheating of the source of ultrasounds and investigate a direct influence of ultrasounds on the demineralization intensity, we just introduced ultrasonic irradiation into “zeolite-alkaline solution” at the same (ice bath) temperature, chemical composition of alkaline solution, mass ratio of zeolite to mixture and the duration of procedure like in the case of conventional treatment.From our ICP-OES results obtained for the desilicated MFI zeolite (Si/Al = 40), it may be concluded that our zeolite was weakly prone to desilication in comparison with FAU-type zeolite of Si/Al = 31 [31]. For faujasite ultrasonically treated with alkaline solutions of various TBAOH content under similar conditions, the percentage amount of Si extracted from framework was 20–40% (for MFI-zeolite was ca 15%). Observed differences in the extraction degree of Si between two types of zeolite structure (FAU vs. MFI) can be explained by a higher stability of MFI zeolite framework than FAU zeolite of elevated Si/Al of 31 (being not natural for the zeolite of this topology, obtained commercially via dealumination of pristine faujasite). Hence, FAU31 zeolite was more sensitive for any modifications (including interaction with aqueous alkaline solutions).The XRD patterns of the investigated catalysts are presented in Fig. 1 . For all samples, the occurrence of MFI-type zeolite phase was detected [53]. Analysis of the XRD reflexes of the studied samples leads to the conclusion that neither ultrasonic irradiation nor chemical composition of NaOH/TBAOH alkaline mixture had a significant impact on the crystallinity of the prepared materials. In all cases, the crystallinity was preserved, which well corresponds to the ICP-OES results (Table 1). It may be explained by a limited extraction of both Si and Al from zeolite MFI, which resulted in minor changes of Si/Al of parent material. Hence crystalline structure of modified zeolites did not undergo a collapse.Based on the analysis of crystallite sizes calculations (Table 1), it can be concluded that the modification of M−40 zeolite with alkaline solutions caused some drops in the crystallite sizes. The average size of crystallites decreased from 501 to 341–448 Å. Registered changes in the size of crystallites did not reveal apparent relationship with the conditions of alkaline treatment procedure.Abello et al. [46], Schmidt et al. [47], Groen et al. [54], Rutkowska et al. [55-57] and Ahmadpour et al. [58] also did not observed significant changes in crystalline structure of desilicated MFI-type zeolites in respect to parent samples of similar Si/Al ratio. 27Al MAS NMR spectra (Fig. 2 ) illustrate the status of Al in the prepared catalysts. For the reference sample (M), the occurrence of the Al signal at ca 57 ppm is attributed to zeolite framework tetrahedral aluminum [45,50]. Simultaneously, very weak signal at 0 ppm originating from extra-framework octahedral Al species was found. The status of Al depended slightly on both the application of ultrasounds during desilication and the chemical composition of 0.2 M NaOH/TBAOH aqueous solution. In the absence of ultrasounds, aluminum did not go from tetrahedral to octahedral positions. According to Sadowska et al. [45], the treatment of zeolite MFI with NaOH did not lead to the formation of extra-framework aluminum species, what means that mild desilication allows to stay all Al atoms in zeolite framework. The lack of the growth of signal at 0 ppm can be also explained by the reinsertion of Al into the zeolite framework (known as realumination), which was previously reported for zeolites with FAU [41,59,60] or MFI-type structure [45,54,61,62]. In case of ultrasonic-assisted procedure of demineralization of MFI-40 zeolite with alkaline mixture containing TBAOH, the removal of aluminum from tetrahedral coordination is accompanied by a slight formation of extra-framework Al species. Fig. 3 shows the 29Si MAS NMR spectra of the prepared samples. In all cases, the occurrence of the Si signals at −112 and at −108 ppm originates from Si(4Si) and Si(3Si), respectively [45,50]. The Si(3Si) signal correspond mainly to Si(1Al) surroundings. The Si(4Si) signal was dominating for all investigated catalysts. For all studied samples, the 29Si MAS NMR spectra given in Fig. 3 and data obtained from the deconvolution of 29Si MAS NMR spectra (Table S1) are similar. Generally, the status of silicon in the studied samples was independent of the route of demineralization (conventional vs. ultrasonic) and chemical composition of the alkaline mixture applied during the procedure of modification. Slight differences in the intensity of Si(4Si) signals correspond to the leaching of Si (and to a lesser content of Al), resulting in slight changes of Si/Al ratios (Table 1). In both series (conventional vs. ultrasonic-assisted), the most intensive Si(4Si) signals were found for the zeolites modified with the NaOH/TBAOH with TBAOH content of 70 mol% due to the highest Al extraction (and Si/Al ratio). Minor changes in the status of either Si and Al also agree with the crystallinity of the prepared samples (Fig. 1).The results of the porosity of the prepared materials are summarized in Table 2 and Figure S1. The parent MFI zeolite is characterized by the presence of both micropores (0.162 cm3/g) and intercrystalline mesopores (0.185 cm3/g) with the average pore diameter of 30.3 Å. Hence, the percentage contribution of mesopores volume was 53.3%.The way of the alkaline treatment influenced the porous structure of the prepared catalysts. The modification of MFI-40 with 0.2 M NaOH under conventional conditions led to a decrease of the total volume of pores from 0.347 to 0.215 cm3/g with simultaneous growth of the mesoporosity to 61.9%. That caused growth of the average pore diameter from 30.3 to 37.3 Å.The use of the ultrasonic-assisted technique led to more significant production of mesoporosity in relation with the conventional method of modification of MFI-40. The volume of micropores decreased from 0.162 to 0.110 cm3/g and the volume of mesopores and average pore diameter rose from 0.185 to 0.230 cm3/g and from 30.3 to 42.1 Å, respectively. That led to the increase of percentage mesoporosity contribution up to 67.6%.The appearance and further rising of TBAOH in the demineralising agent strengthened the formation of mesoporosity in prepared catalysts. Furthermore, all zeolite samples modified sonochemically revealed higher mesoporosity than counterparts prepared classically. The use of 0.2 M NaOH/TBAOH (10 mol% of TBAOH) in the absence or presence of ultrasounds caused the increase of mesopores volume participation from 53.3% to 62.8% vs. 64.8% and the increase of average pore diameter from 30.3 to 37.5 vs. 37.7 Å, respectively. In the case of the utilization of the NaOH/TBAOH alkaline solution containing 70 mol% of TBAOH, a minimal increase of mesoporosity from 53.3% to 63.0% vs. 66.7% with a simultaneous growth of an average pore diameter from 30.3 to 38.9 vs. 41.7 Å were found.It also was found that an alkaline treatment of MFI-40 zeolite resulted in a minimal growth of percentage Sext/SBET ratio from 16.6% to 17.4–22.4%. However, all catalysts prepared ultrasonically demonstrated higher Sext/SBET values (20.1–22.4%) in comparison with the samples modified conventionally (17.4–19.3%).For all studied samples, the appearance of hysteresis loop of IV type can be attributed to the presence of both intercrystalline pores between the MFI crystals (particularly for parent sample “M”) and the formation of mesoporosity.According to the XRD patterns illustrated in Fig. 1 as well as ICP-OES data summarized in Table 1, small changes in the crystallinity are in line with minor changes in porous structure of the studied samples.Our results are in good agreement with, Zhang et al. [29], Oruji et al. [30], Khoshbin et al. [32] as well as with our previous studies [31]. It was evidenced that a sonochemical demineralization procedure enhanced the production of higher mesoporosity in respect to the conventional alkaline-treatment technique. Additionally, it was indicated that the use of ultrasounds during modification resulted in preserved microporosity. On the other hand, observed changes in the porous structure of our MFI-type zeolites are significantly smaller than for MFI-type analogues desilicated by Sadowska et al. [45,49], Abello et al. [46], Shmidt et al. [47], Gil et al. [50] and Groen et al. [62] due to the use of much milder conditions in our experiments (relatively short duration of the procedure, ice bath temperature, and respectively low concentration of desilicating agents) in comparison with the research quoted above.The analysis of the morphology of the prepared MFI-40-based catalysts are illustrated in Fig. 4 (magnification to 50,000x). The appearance of SEM images leads to the conclusion that the particles of all modified samples are of irregular shape with dimensions ranging from 300 to 1000 nm. For the samples treated under conventional conditions (Fig. 4a-c), alkaline treatment led to some fragmentation of grains, which slightly influenced the porosity of this series of samples.In case of the catalysts prepared via ultrasonic-assisted procedure (Fig. 4d-f), treatment of the parent zeolite (M) led to the appearance of cracks, cavities and strengthened the fragmentation of grains in comparison with the analogues modified in the absence of ultrasonic irradiation. Observed changes in microscopy well correlate with crystallite sizes and textural properties given in Table 2. Nevertheless, the registered changes in the morphology of the investigated MFI-type zeolite materials are quite small, particularly when we compare our current samples (MFI-40) with the materials based on FAU-31 zeolite, which we reported in [31]. Relatively stable morphology of our MFI-type catalysts is in line with their preserved crystallinity (Fig. 1).Furthermore, from analysis of the SEM pictures, it may be concluded that the chemical composition of demineralizing mixture (meant as NaOH/TBAOH ratio) both with and without ultrasounds had no apparent impact on the size and the shape of zeolite crystals.For comparison, the TEM images (50,000×) of the variously prepared catalysts are illustrated in Fig. 5 . Analysis of the appearance of crystalline grains led to the conclusion that alkaline treatment of the parent M−40 zeolite caused the production of holes and the formation of “swiss cheese”-type zeolite grains followed by the changes in porous structure of the prepared MFI-based samples. Observed perforation seems to be more apparent in the case of zeolites desilicated in the presence of ultrasounds (Fig. 5 C and E). Nevertheless, observed changes in the appearance of zeolite grains were not very sharp, which was in line with crystallinity, structure, morphology and subtle changes in porosity of the studied samples.The characterization of the acidity of the prepared MFI-type samples was illustrated in Fig. 6 and Table 1. For the parent sample (M), the IR spectrum of the OH groups region indicated the presence of the bands at: 3740 cm−1 attributed to external Si-OH groups with a shoulder at 3730 cm−1 of silanols in the defects, 3670 cm−1 assigned to Al-OH, 3620 cm−1 originating from acidic Si-OH-Al groups and at 3490 cm−1 coming from silanol nests [45,49,50,62].Alkaline treatment with 0.2 M NaOH or NaOH/TBAOH aqueous solutions led to significant changes in the appearance of IR spectra. The disappearance of the band at 3490 cm−1 as well as an apparent decrease of the signals at 3740 and 3620 cm−1 demonstrated the removal of a significant part of OH groups during desilication procedure. Another effect was a slight increase of the band at 3670 cm−1, which suggests that the interaction of MFI-40 zeolite with alkaline mixtures caused the production of Al-OH groups.Information on the strength of the acidic OH groups (in Si-OH-Al) was given in Table 1. The data obtained from CO sorption showed that the modification of parent MFI-40 zeolite with NaOH or NaOH/TBAOH mixtures both in the absence or presence of ultrasounds did not influence the acid strength of Si-OH-Al (Δν3620OH…CO = ca 310 cm−1). In case of M−40 zeolite treated with NaOH solution under conventional conditions (M−0c), the broad band with the maximum at 3670 cm−1 probably overlapped with the band at 3620 cm−1 of Si-OH-Al, making the latter signal undetectable, therefore the strength of protonic acid sites was not determined for M−0c.Our results obtained in the current study are in contrast to other research reported for zeolites with MFI [49] and BEA [63], for which desilication caused the decrease of the acid strength of Si-OH-Al by ca. 10–20 cm−1, which derived from the occurrence of the more and less acidic OH groups. This contradiction may be explained by the fact that our MFI-40 zeolite was modified under milder conditions in comparison with the majority of experimental procedures reported in the available literature. Furthermore, taking into account relative high amounts of Al extracted from zeolite structure in respect to Si, it could be possible that the process of demineralization of our MFI-40 zeolite caused the destruction of Si-OH-Al groups of the highest and the lowest acid strength. Alternative situation might refer to the removal of Si and Al responsible for Si-OH-Al of medium acid strength. Hence in both suggested cases, an average acid strength of the remaining protonic acid sites after MFI-40 treatment under presented conditions did not undergo apparent changes.The results on the quantitative analysis of the acid sites present in the prepared catalysts are given in Table 1. In all cases, the modification of the parent MFI-40 zeolite with NaOH or NaOH/TBAOH mixtures caused a strong decrease of both Brønsted and Lewis acid sites from 228 to 34–70 μmol and from 309 to 53–105 μmol/g, respectively. The observed tendency implies a simultaneous removal of silicon and aluminum from modified MFI-type zeolite. It also was found that the technique of alkaline treatment slightly influenced concentrations of both types of acid sites. Generally, the application of ultrasonic-assisted procedure of MFI-40 modification with the alkaline mixture at concrete chemical composition resulted in a little higher concentrations of acid sites than for counterparts prepared conventionally. Obtained data agrees with the results of ICP-OES analysis, which showed that the presence of ultrasounds during modification procedure of MFI-40 zeolite shifted demineralization towards Si extraction from zeolite framework. For comparison, the interaction between MFI-40 zeolite and alkaline mixture under ultrasonic-free conditions caused practically equal Si and Al removal from zeolite structure (Table 1).Catalytic activity of variously prepared MFI-type zeolite - based catalysts in the dehydration of ethanol reaction was depicted in Fig. 7 A.For all studied catalysts, It was found that catalytic activity rose with the temperature of experiment in a whole range. In the case of a reference catalyst (M), at 150 °C, 170 °C, 190 °C, 210 °C, 230 °C, 250 °C 270 °C and 290 °C, the conversion of ethanol formed the sequence, as follows: 2%, 9%, 23%, 40%, 57%, 70%, 82% and 100%, respectively. It also was indicated that an alkaline treatment of MFI-40 zeolite (independently of the procedure conditions) led to the enhancement of the catalytic properties of the prepared materials. For all zeolites treated with basic solutions, the conversion of ethanol was 100% at 270–290 °C. The best results were found for the catalyst modified under ultrasonic-assisted conditions with NaOH/TBAOH solution (containing 70 mol% of TBAOH (M−70 s), for which the conversion of ethanol was 8%, 26%, 49%, 68%, 76%, and 84% at 150 °C, 170 °C, 190 °C, 210 °C, 230 °C and 250 °C, respectively. For comparison, some worse conversion of ethanol was found for the counterpart prepared in the absence of ultrasounds (M−70c), which was 5%, 20%, 42%, 60%, 72% and 78% at the same temperature range, as above.The presence of ultrasonic irradiation during alkaline treatment of MFI-40 type zeolites also improved catalytic activity in the case of the NaOH/TBAOH solutions containing 10 mol% of TBAOH. At 150 °C, 170 °C, 190 °C, 210 °C, 230 °C and 250 °C, the conversion of ethanol was 0% vs. 5%, 6% vs. 20%, 22% vs. 39%, 47% vs. 59%, 65% vs. 71%, 80% vs. 85% for the catalysts prepared conventionally and sonochemically, respectively.Much weaker effect of the application of ultrasounds during demineralization procedure of zeolites on their catalytic activity was found for the samples modified with NaOH solution. At 150–190 °C, the conversion of ethanol was higher for the catalysts prepared classically (14%-44% for M−0c vs. 6%-41% for M−0 s), while the zeolites treated ultrasonically revealed higher catalytic activity at 210–250 °C (56%-74% for M−0c vs. 63%-80% for M−0 s).The observed effects correspond to the degree of demineralization of zeolites and the production of mesopores, which facilitated transport of the reagents within porous structure of investigated catalysts.Alkaline treatment of MFI-40 zeolite also influenced the reaction rates in the prepared catalysts (Fig. 7B). Analysis of turnover frequency (TOF) results led to the conclusion that the modification of the parent zeolite (M) with alkaline solutions raised TOF values, which agreed with the changes in either porous structure or acidity of the prepared catalysts (Tables 1 and 2) and with the results reported by Verboekend and Pérez-Ramírez [42].It also was found that the application of ultrasonic irradiation during preparation procedure of catalysts (M−0 s and M−70 s) resulted in lower TOF values than for the analogues prepared conventionally (M−0c and M−70c). An opposite tendency was indicated for M−10c and M−10 s. Observed correlations between TOF values implied from catalytic activity as well as from concentrations of protonic acid sites (responsible for this reaction). Actually, the catalysts prepared ultrasonically revealed generally higher conversion of ethanol, but the concentrations of protonic acid sites also were higher for this group of materials (with one exception for M−10 s) in comparison with the catalysts modified conventionally. Direct comparison of the changes in the concentration of Brønsted acid sites (Table 1) and catalytic activity (Fig. 7A) allows us to claim that a quantitative analysis of active centres has a stronger impact of TOF values than the conversion of ethanol at given temperatures.Analysis of the results obtained from catalytic performance illustrated in Fig. 8 A and 8B indicated clearly that both the type and amount of a concrete product is determined by the temperature range. Up to 210 °C, it was possible to manufacture practically pure diethyl ether: for all studied catalysts, selectivity was 93–100%. At higher temperatures, the appearance of ethylene was registered. It is worth to underline that no pure ethylene was produced at 210–290 °C due to both co-existence of diethyl ether (at 210–270 °C) and the coking occurring at the highest temperatures of experiment.Significant selectivity to ethylene was found for all studied catalysts at 250–290 °C. For parent zeolite (M) it was 17–80%. Alkaline treatment of MFI-type zeolites resulted in a visible enhancement of selectivity to ethylene, however, this effect was generally stronger for the samples after ultrasonic irradiation. For instance, in case of MFI-type zeolite sonochemically modified with NaOH/TBAOH alkaline solution containing 10 mol% of TBAOH (M−10 s), selectivity to ethylene was 55–84%, meanwhile for zeolite modified conventionally (M−10c), selectivity to ethylene was 41–80%. Other by-products like acetaldehyde were not detected.The best catalyst was M-70s, for which the highest conversion of ethanol was found. Additionally, M-70s demonstrated very high selectivity to diethyl ether (94-100%) at 150-210 °C and the highest selectivity to ethylene among investigated catalysts (21%, 66% and 84%) at 230 °C, 250 °C and 270 °C.Similar observations were reported by Oliveira [64], who investigated the ethanol dehydration on Cu- and Fe-ZSM-5 catalysts. The production of diethyl ether (70–100%) was found in the temperature range of 180–200 °C on Cu-ZSM-5, while ethylene was formed mainly at temperatures exceeding 200 °C over both pure and Fe containing ZSM-5 catalysts (20–100% depending on the type of catalyst). For comparison, Zhan et al. [65] reported yield of DEE reaching 67% using 2% PHZSM-5 catalyst. In turn, Jinfa et al. [66] obtained yield of DEE exceeding 70% at 180 °C over ZSM-5.Our catalytic results are also in line with Phung, Chiang et al. [33,67–69], who studied the ethanol dehydration on commercial H-FER, H-MFI, H-MOR, H-BEA, H-Y and H-USY zeolites For the investigated systems, diethyl ether was mainly produced at low temperatures, while the production of ethylene was predominant at high temperatures, which was independent of the type of acid sites of the studied catalysts. It was found that at 180 °C, the selectivity to DEE was higher for H-MFI and H-BEA (exceeding 70%) than for other zeolites. On the other hand, at high temperature, almost full selectivity to ethylene was registered for H-FER, H-Y and H-USY, while co-production of higher hydrocarbons took place in the case of H-MFI, H-BEA and H-MOR, which was in line with the data reported by Stepanov et al. [70] and Wang et al. [71]. Rising reaction temperature led to the formation of coke, particularly over H-MOR and H-BEA. The strength of protonic acid sites was found as similar for all studied zeolites, which was in agreement with Xu et al. [72] and did not correlate with catalytic activity and selectivity.It was also shown [33,67–69] that porous structure and morphology of the investigated catalysts influenced the catalytic properties in given reaction. Medium pore zeolites, such as H-MFI, H-BEA and H-MOR demonstrated the highest selectivity to diethyl ether (98%) at moderate temperatures (180 °C). It may be concluded that zeolites of medium size channels are the most suitable for this reaction than counterparts possessing either larger or smaller cavities (faujasites or ferrierite, respectively). That implies from a confinement effect favouring the production of diethyl ether at lower temperatures and the shape selectivity enabling the formation of ethylene at higher temperatures. It may be also explained by the different kinetic behaviour of these two reactions, namely, in higher activation energy and lower ethanol reaction order towards ethylene formation in relation to DEE.Almost full selectivity to ethylene and yield were obtained at high temperature over small-pore H-FER and on large-pore H-Y and H-USY [69]. In the case of medium-pore zeolites like H-MFI, H-MOR and H-BEA, the selectivity to ethylene was limited by the production of higher hydrocarbons and coke.Detailed mechanism of ethanol dehydration on protonic form of zeolites with MOR, MFI and FER type structures was described by Phung, Chiang et al. [33,67]. It was explained that a selective production of diethyl ether from ethanol occurs at lower temperature by the reaction of ethoxy groups with undissociated ethanol. At higher temperatures, the formation of ethylene is going via decomposition of ethoxy groups on catalysts containing active acid sites.Osuga et al. [73] also investigated ethanol dehydration over MFI- and MOR-type zeolites and under different contact time conditions. DEE was found as one of the initial reaction products, influencing the catalytic activity. At the same catalyst weight / ethanol flow ratio (24.3 g·h/mol), for MFI zeolite, DEE was detected as dominating product with selectivity reaching 100%, whereas ethene was produced over mordenite with selectivity reaching 60%. Therefore, it may be concluded that the dehydration of ethanol over zeolites with two different structure types can occur via different reaction routes.According to Madeira et al. [74], both the acidity and porous structure of zeolites determined their catalytic behaviour in the ethanol conversion into hydrocarbons possessing three carbon atoms or more (C3+). It was indicated that high pore size H-FAU and H-BEA type zeolites were characterized by a high yield of ethylene and diethyl ether as well as low contribution of C3+ hydrocarbons in comparison with medium pore zeolite HZSM-5. This observation can be explained by a faster deactivation of large pore zeolites, which was caused by the formation of coke, eliminating protonic acidity, responsible for the transformation of ethylene into higher hydrocarbons. For H-ZSM-5 zeolite, after 16 h of reaction, practically full conversion of ethanol towards C3+ hydrocarbons (including butenes, parafins and aromatics) was found. In turn, in the case of large pore zeolites (H-BEA, H-FAU), high amounts of by-products like more condensed aromatics were detected. Furthermore, it was found that for H-ZSM-5 zeolite, the deactivation was slower and the production of C3+ hydrocarbons was found even under the saturation of the catalyst with coke molecules, thus it may be supposed that for this zeolite, reaction could take place at the pore mouth of the channel.According to Däumer et al.[75], the formation of C3+ hydrocarbons is favoured in the zeolites of relatively narrow cavities, thus the acid sites strength present in micropores plays a supporting role in the production of higher hydrocarbons. It was also reported that the formation of C–C bonds was strongly dependent on the presence of highly strong Brønsted acid sites. In turn, the intermediate species stability in zeolite framework depended on interaction between fragments of confined hydrocarbon and zeolite framework [76]. The size of cavities influences the stability of the confined species defined by an electrostatic and van der Waals interaction between fragments of hydrocarbons and zeolite framework.Our findings are also in line with Gołąbek et al. [77], who investigated the role of pore arrangement of 10-ring zeolites (ZSM-5, TNU-9 and IM-5) on their catalytic properties in ethanol transformation. From obtained data, it was concluded that all studied catalysts were active at 150–300 °C and at atmospheric pressure, leading to the production of diethyl ether (DEE) and ethylene as the products. It was also shown that the conversion of ethanol increased with reaction temperature. ZSM-5-based catalysts did not undergo deactivation and small and uniform ZSM-5 crystals did not affect catalytic lifetime.Based on available literature, a beneficial impact of the application of ultrasonic irradiation in the synthesis of zeolite-based materials on their physicochemical and catalytic properties has been reported. For instance, Oruji et al. [30] performed ultrasonic-assisted desilication of NaY zeolite in order to prepare hierarchical materials with elevated mesoporosity and higher crystallinity than zeolites treated conventionally. Independently of the used technique, the mesoporosity of desilicated FAU-type zeolite gradually increased with the duration of procedure and was higher for the samples modified sonochemically. In the reaction of catalytic cracking of middle distillate at 550 °C, the it was revealed that all sonochemically treated samples demonstrated higher catalytic activity with their high liquid and low gas yields (78–86% and 14–22%, respectively). Coking was practically absent. Furthermore, it was shown that the catalyst lifetime for sonochemically prepared materials was higher than for the samples treated conventionally due to more apparent destruction of microporosity with simultaneous more noticeable generation of mesoporosity in zeolite.Another example of beneficial influence of ultrasounds during catalysts preparation is an ultrasonic-assisted deposition of active phase on zeolite carrier, which was reported in our previous papers [78-80]. Jodłowski et al. [78] reported that sonochemically prepared structured reactors with a deposited copper on ZSM-5 and USY zeolite revealed a full NO conversion and almost constant 100% selectivity to nitrogen in SCR- DeNOx reaction. In turn, Chlebda et al. [79] indicated that ultrasonic procedure of iron containing ZSM-5-based catalysts preparation enhanced catalytic activity in the DeNOx process, with almost full selectivity to N2. Sobuś et al. [80] reported that Cu- and Co containing BEA zeolites revealed the best catalytic properties in the Selective Conversion of Lactic Acid into Acrylic Acid reaction with the selectivity to Acrylic Acid exceeding 60%.In this research, we investigated an ultrasonic-assisted desilication of zeolite with MFI type structure using aqueous NaOH/TBAOH alkaline solutions of various chemical compositions. Subsequently, we compared the physicochemical and catalytic properties of such prepared samples with counterparts desilicated in the absence of ultrasounds.It was shown that the application of ultrasounds during demineralization procedure caused higher both silicon and aluminum extraction in comparison with analogues treated classically. The Si and Al contents leached from MFI-type zeolite framework were in the range of 0.5–15.3% and 0.1–8.6%, respectively. Si/Al ratio of the modified MFI-based samples was in the range of 32.4–37.6 and was only slightly lower in respect to the reference sample (37.7). Alkaline treatment of MFI-type zeolite resulted in the formation of “swiss cheese” - type zeolite grains containing numerous holes inside zeolite crystallites, which was more visible for the zeolites modified under ultrasonic-assisted procedure.Surprisingly, we observed that the conducting of demineralization procedure (independently of the presence/absence of ultrasonic irradiation) did not alter the crystallinity, structure and morphology of the modified materials. Nevertheless, the application of ultrasounds during alkaline treatment procedure led to the production of higher mesoporosity, which enabled better mass transfer of reagents in porous structure and therefore caused enhanced catalytic properties in the reaction of the dehydration of ethanol in relation to the catalysts obtained under conventional demineralization conditions.Furthermore, it was indicated that independently of the alkaline treatment technique (conventional vs. ultrasonic), a notable decrease of both protonic and Lewis acidity corresponded to a simultaneous leaching of Si and Al from the structure of MFI-type zeolite.The analysis of the results presented above led to the conclusion that ultrasonic assisted demineralization of MFI-type zeolite led to the production of attractive catalysts with easily accessible active sites for the ethanol processing. Ł. Kuterasiński: Conceptualization, Data curation, Formal analysis, Methodology, Software, Validation, Supervision, Visualization, Writing - original draft, Writing - review & editing. U. Filek: Formal analysis, Investigation, Methodology. M. Gackowski: Formal analysis, Investigation, Methodology, Visualization, Writing - review & editing. M. Zimowska: Formal analysis, Investigation, Methodology. M. Ruggiero-Mikołajczyk: Formal analysis, Investigation, Methodology. P.J. Jodłowski: Formal analysis, Investigation, Methodology, Visualization, 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 study was financed by the statutory funds from the Jerzy Haber Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences.Supplementary data to this article can be found online at https://doi.org/10.1016/j.ultsonch.2021.105581.The following are the Supplementary data to this article: Supplementary data 1

In this paper, the ultrasonic-assisted desilication technique was reported as an attractive and efficient way for the preparation of hierarchical zeolites with MFI structure type. The prepared materials were used as active catalysts for the dehydration of ethanol into diethyl ether and ethylene. For all catalysts, the selectivity to diethyl ether was ca 95% or higher up to 210 °C, with catalytic activity in the range of 40–68%. In case of desilicated zeolites, at 270–290 °C, the conversion of ethanol was full with selectivity to ethylene ca 80%. MFI-type commercial zeolite was treated with a sodium and/or tetrabutylammonium hydroxide aqueous solutions (NaOH or NaOH/TBAOH) for 30 min. In the case of the application of ultrasounds, a QSonica Q700 sonicator (60 W and 20 kHz) equipped with a “1” diameter horn was used. In all cases, desilication was performed in an ice bath in order to keep the procedure conditions at low temperature. It was indicated that the use of ultrasounds during desilication procedure caused higher extraction of silicon and aluminum, which was connected with an elevated mesoporosity in relation to the samples modified in the absence of ultrasounds. Ultrasonic-assisted treatment of MFI-type zeolite caused also an apparent formation of numerous holes inside zeolite grains, resembling the look of “swiss cheese”. Furthermore, it was indicated that the samples prepared using ultrasonic irradiation exhibited enhanced catalytic properties in the dehydration of ethanol. For instance, MFI-type zeolite treated with NaOH/TBAOH alkaline mixture containing 10 mol% of TBAOH in the presence of ultrasounds (M−10 s) demonstrated higher both conversion of ethanol (59% vs. 47%) and selectivity to diethyl ether (95% vs. 93%) in comparison with zeolite modified conventionally (M−10c). The best catalyst was zeolite ultrasonically desilicated with NaOH/TBAOH solution of 70 mol% of TBAOH (M-70s). Generally, this catalyst indicated the highest conversion of ethanol, very high selectivity to diethyl ether (94-100%) at 150-210  °C and the highest selectivity to ethylene among investigated catalysts (21%, 66% and 84%) at 230  °C, 250 oC and 270  °C.

The C–C coupling reactions have attracted significant attention since their discovery and have been one of the most important research areas [1]. These reactions typically require the use of precious metal catalysts (e.g., Pd [2,3], Au [4] and Ru [5]) and sometimes significant excess of reductants. In this regard, the economical and less toxic Ni catalysts have drawn increasing attention [6–10]. The design of catalysts and related techniques to overcome these drawbacks is thus of utmost importance. Electrochemical transformations have a high potential to address these issues [11–15]. Several remarkable electrochemical coupling reactions for the C–C [16–20], C–O [21,22], C–N [23–25], C–S [26], N–S [27] and N–P [28] bond formation have been developed, but to the best of our knowledge, less attention has been paid to the homocoupling reactions using cyclic voltammetry methods [29–33].Different pathways are possible for metal-mediated radical formation from aryl and benzyl halides. Among them, single electron transfer pathways, either outer-sphere or inner-sphere mechanism, have been proposed [34–36]. Electron transfer proceeds from metal to the aryl/benzyl halides to form radical, followed by C-X bond cleavage [10,37,38]. Recent DFT calculations on several Ni0 and NiI systems propose the concerted halogen-atom-abstraction pathway and most of the transition-metal catalyzed coupling reactions reported so far require metal center reduction to its formal zero valent state [39–41]. In the case of Ni-based systems, the commonly found reduced metal species is Ni0. In very few systems, the coupling goes through a formal NiI species, probably because it is difficult to control the reduction by using metal powders. On the other hand, while low-valent NiI species is difficult to isolate due to its inherent instability, such type of species can be generated by electrochemical methods and employed in coupling reactions.We have developed various PN3(P) pincer ligand architectures and have demonstrated that their metal complexes often offer some unique reactivities because of their different kinetic and thermodynamic properties compared to those of analogs with CH2 spacers [42–49]. Herein, we report a PN3P–Ni complex (1) (Fig. 1 ) for benzyl halide homocoupling reactions under electrochemical conditions. We present electrochemically, for the first time, that two different formal oxidation states (0 and ​+1) are catalytically active in the presence of benzyl chloride and bromide, respectively.[Ni(LPN3P) (Cl)]Cl (1) was synthesized by mixing the PN3P ligand and NiCl2 in THF and recrystallized in MeOH (Fig. 1) [48]. The purity was confirmed by NMR, HRMS, elemental analysis, and UV–vis (Figs. S1-S5, Tables S1 and S2 in Supporting information). The crystal structure shows that the NiII center is bound to the N atom (pyridine ring), two P atoms of two di-tert-butylphosphine arms and one chloride anion (the other noncoordinating chloride anion resides outside the metallo-ligand to balance charge) to give a slightly distorted square planar geometry, where the metal center has no deviation from the plane (d Ni ​= ​0 ​Å; d plane ​= ​0 ​Å). As we have previously showed that POCOP–Ni complexes could catalyze the homocoupling reaction of a series of benzyl halide derivatives by using zinc dust as a reductant at 115 ​°C (Scheme 1 ) [50], complex 2 was chosen for benchmarking the activity under electrochemical conditions.Cyclic voltammograms (CV) of complexes 1 and 2 using glassy carbon as a working electrode displayed a quasi-reversible reduction wave at −1.49 ​V (ΔE p ​= ​66 ​mV) and −1.33 ​V (ΔE p ​= ​80 ​mV) vs. Fc+/0 in DMF, respectively (Fig. 2 A), which corresponded to a NiII/I redox couple, consistent with values reported for other Ni complexes [51–54]. Additionally, a quasi-reversible reduction wave was observed at a more negative potential at −2.06 ​V (ΔE p ​= ​110 ​mV) vs. Fc+/0 for 1 [53,55]. No additional wave was observed for complex 2 when further scanning was done to more cathodic potentials, indicating that the associated redox process is kinetically hindered (Fig. S6 in Supporting information). The event observed at a more negative potential of complex 1 can be attributed to a NiI/0 redox couple, suggesting the PN3P ligand may stabilize the Ni0 oxidation state. Coulometric study of the reduction of 1 confirmed that the first reduction wave corresponds to a one-electron reduction process. The CV of the NiII/I process was obtained at different scan rates to reveal a linear relationship, confirming a diffusion controlled process with the diffusion coefficient determined to be 1.5 ​× ​10−7 ​cm2/s by the Randles-Sevcik equation (Fig. S7 in Supporting information).Spectroelectrochemical data supported the generation of a single-step one-electron transfer at −1.60 ​V, and the cathodic potential induced a gradual decrease (by ~25%) in the LMCT band at 318 ​nm, while the bands around 459 ​nm shifted to 474 ​nm (Fig. S8 in Supporting information). The depletion of the optical band at 318 ​nm, along with a decrease of the 274 ​nm band and the formation of a new band at 355 ​nm ​at the same solution, suggest a possible one-electron reduction of 1, attributed to the reduction of NiII to NiI.In order to elucidate this reduction process and to address the limited solubility of complex 1, a counterion exchange reaction with NaBPh4 was conducted to afford complex 3 (~95%) (Scheme 2 and Fig. 3 , Figs. S9-S12 and Tables S1 and S2 in Supporting information). Complex 3 showed similar redox behaviors in CV that observed for 1 in DMF (Fig. 2A, violet line). The observed similar NiII/I and NiI/0 redox couples suggest that there is no substantial effect of BPh4 to the NiII center. Chemical reduction of 3 with 1 equiv. of cobaltocene in dry and degassed THF led to the formation of complex 4 (Figs. S13 and S14 in Supporting information). The optical spectrum observed after stoichiometric reduction of complex 3 by cobaltocene is analogous to that of reduced 1. The EPR spectrum of complex 4 in acetone was essentially identical to that obtained by electrochemical reduction of 3 in acetonitrile. The X-band EPR spectrum of this species exhibit a rhombic signal, with g x  ​= ​2.21; g y  ​= ​2.11; and g z  ​= ​2.02, which confirms an S ​= ​½ ground state at 100 ​K (Fig. 2B, red line). The spectrum shows distinct splitting from the pincer ligand donor atoms as well as the chlorine [56]. The g-values and splitting constants for complex 4 is summarized in Table 1 . The deviation of the g-values from g e ​= ​2.002 ​at low temperature concludes that the singly occupied molecular orbital (SOMO) is principally Ni-based. Moreover, the computational analysis of the spin density at DFT level with the wB97X-D functional indicates the radical localized at the Ni center (Fig. 2C), more details are deposited in Supporting information).Our further electrochemical investigations on catalysts 1 and 2 in the presence of benzyl halides, showed clear indication of catalysis. Addition of benzyl chloride (BnCl) in the electrolytic solution containing 1.0 ​mmol/L of catalyst 1 showed a gradual increase of catalytic current (up to 20.0 ​mmol/L of substrate addition). The onset potential in this case (−2.07 ​V vs. Fc+/0) overlays with the NiI/0 process (Fig. 4 A), strongly suggesting that the Ni0 state could catalyze the homocoupling of BnCl, independent of the NiII/I process. Very interestingly, in the presence of benzyl bromide (BnBr), a large electrocatalytic current was observed with an earlier onset potential at −1.38 ​V vs. Fc+/0 (Fig. 4B), indicating that NiI state is the active species to catalyze the homocoupling of BnBr, in sharp contrast to that for BnCl. In this context, it is worthy to mention that, the difference in onset reduction potentials (0.69 ​V vs. Fc+/Fc) between BnCl and BnBr is comparable to that of direct electrochemical reduction of these substrates at a GC electrode in MeCN (0.45 ​V vs. SCE) [57]. The catalytic current (i cat) shows a linear correlation with the concentration of catalyst (Fig. S15 in Supporting information), which confirms that the catalytic process is mononuclear.Complex 2 exhibits a large electrocatalytic current in the presence of BnBr with an onset potential at −1.48 ​V vs. Fc+/0 (Fig. 4D), close to the redox couple NiII/I. However, in the presence of BnCl, no enhancement of current was observed with further scanning of reduction potential (Fig. 4C). These observations imply that Ni0 oxidation state is essential to catalyze the homocoupling reaction of BnCl. The catalysis is absent for blank GC electrode at 10 ​mmol/L substrate in electrolytic solution without catalyst (Fig. 4, black line).The catalytic rate to form the homocoupling product with Ni0 or NiI states for benzyl halides can be determined if i cat/i p are analyzed at a fixed substrate concentration (20.0 ​mmol/L), where the catalytic current plateaus. The pseudo-first order rate of the catalysis is determined by the equation: i cat/i p ​= ​0.72 (k[substrate]/ν)1/2 (see Supporting information for details) [58], where i cat is the peak current of catalysis in the presence of the substrate, i p is the peak current of NiI and Ni0 species without the substrate obtained from the CV data, ν is the scan rate in V/s, and n cat ​= ​2 is the number of electrons transferred in each catalytic cycle (Table 2 ). The TOF for BnCl to C–C coupling product was determined from the scan rate dependence study and was found to be 0.43 s−1 (Fig. 5 A and Fig. S16 in Supporting information) with a rate constant of 17.0 ​L ​mol−1 s−1 (Figs. S17 and S18 in Supporting information). Changing the substrate to BnBr resulted in an increase in TOF to 19.5 s−1 (Fig. 5B and Fig. S19 in Supporting information) with a rate constant of 1642 ​L ​mol−1 s−1 (Figs. S20 and S21 in Supporting information) for catalyst 1. Similar method is applied for catalyst 2 (Figs. S22-S24 in Supporting information).Controlled potential electrolysis at −1.1 ​V vs . Ag/AgCl on a graphite electrode was examined with 1.0 ​mmol/L catalyst loading along with BnBr (10.0 ​mmol/L, 0.13 ​mmol) (Fig. S25 in Supporting information). After 3 ​h of electrolysis, the current slightly decreased but maintained a steady state value, consistent with the consumption of the substrate. Bibenzyl formed in 81% isolated yield after bulk-electrolysis (confirmed by 1H NMR, Fig. S26 in Supporting information), with a Faradaic yield of 98%. For BnCl and derivatives, the electrolysis was carried out at −1.7 ​V vs. Ag/AgCl for 3 ​h (Fig. S27 in Supporting information). The stability of our Ni catalyst was checked before and after the controlled potential electrolysis by 31P NMR, HR-MS and UV–vis spectroscopy (Figs. S28-S30 in Supporting information). Electrodes including glassy carbon (GC) after multiple CVs gave no catalytic response in a fresh, catalyst-free electrolyte in DMF, suggesting no catalyst deposition on the surface of GC under these electrochemical conditions. After 3 ​h of electrolysis with graphite rod electrode, no catalyst decay was observed, indicating a robust nature of the catalyst. Indeed, even with the increasing initial substrate concentration (50.0 ​mmol/L), no obvious decay was detected after 14 ​h of electrolysis. However, the optical data shows a significant decomposition of 2 after 3 ​h of electrolysis in the presence of substrate (Fig. S31 in Supporting information). Accordingly, a series of substrates were further studied. A significant improvement in yields was observed when the substrate was changed from BnCl to BnBr under similar conditions (Table S3 in Supporting information). After electrolysis for 15 ​h in the presence of BnCl, the yield of coupling product reached 90%, consistent with experimental results with zinc dust.From the detailed analysis of the cyclic voltammogram and chronoamperometry of 1, evidently PN3P ligand must stabilize the low-valent Ni species, which is necessary for the C-X bond activation of substrate. Based on these observations, a catalytic cycle for the coupling reaction was proposed (Fig. 6 A). The low-valent Ni species (Ni0 or NiI for BnCl and BnBr, respectively) is generated at the cathode at a respective potential which reacts with benzyl halides and abstract a halogen atom from the benzyl halide molecule to cleave the C-X bond to afford a NiI or NiII intermediate along with a benzyl radical which dimerizes to form bibenzyl [59]. The resulting high valent Ni species immediately take one electron from the electrode with loss of a halide ion to regenerate the low-valent catalyst to continue the catalytic cycle. Additionally stoichiometric treatment of 4 with BnBr at room temperature results in complete conversion to bibenzyl within 5 ​min. In sharp contrast, performing the reaction with BnCl did not generate the coupling product. The proposed mechanism is also supported by the trapping experiment of the benzyl radical with TEMPO (Fig. 6B). No bibenzyl product was observed, but TEMPO-Bn (Fig. S32 in Supporting information) [60].In conclusion, we have unambiguously elucidated the role of NiI and Ni0 for successive homocoupling of benzyl bromides and benzyl chlorides. For the first time, two different formal oxidation states of the catalytically active Ni species (0 and ​+1) were determined for BnCl and BnBr, presumably due to different bond strengths of C-X bonds. Alternatively, it is also possible that in the case of BnCl, a more reduced Ni species (Ni0) is required to facilitate electron transfer in order to cleave C–Cl bond. Accordingly, higher yields of the homo-coupling products from BnBr and derivatives were achieved when the reaction time of electrolysis was restricted to 3 ​h. Our findings show that catalyst 1 containing PN3P-ligand is an incredibly effective catalyst for the homocoupling reactions. The onset potential for the coupling reaction of catalyst 1, in the presence of BnBr, is less negative than catalyst 2. CV experiments confirmed that NiI/0 redox couple was not observed under electrochemical condition for catalyst 2, not catalyze the coupling reaction of BnCl. Above all observations of the PN3P system indicate its distinct kinetic and thermodynamic properties compared to their POCOP analogs. We have further demonstrated the reaction kinetics and mechanism. Our finding reveals a strong ligand effect on the reactivity and selectivity of the reaction. Investigations on the potential applications for other classes of substrates are currently ongoing and will be reported in due course.All experiments with metal complexes and phosphine ligands were performed under an argon atmosphere in a glovebox or using standard Schlenk techniques. All the solvents were purified before use. For C–C coupling reaction, all the substrate were degassed before use. Column chromatography purifications were performed by flash chromatography using Merck silica gel 60. All other reagents were used as received. 1H, 13C, and 31P NMR spectra were recorded using Bruker AVIII 400, AVIII 500 or AVIII 600 spectrometers. EPR spectra was recorded using X-band continuous wave Bruker EMX PLUS spectrometer (BrukerBioSpin, Rheinstetten, Germany) equipped with standard resonator for high sensitivity CW-EPR operating with a microwave frequency. A typical Evans measurement was done in a coaxial tube containing the solvent and the internal standard. The absorption spectra were measured in an Agilent 8453 UV–visible spectrophotometer. Chemical shifts in 1H NMR and 13C NMR were reported in parts per million (ppm). The residual solvent peak was used as an internal reference: 1H NMR (DMSO-d 5 in DMSO‑d6 , δ 2.54, chloroform in CDCl3 7.26) and 13C NMR (DMSO‑d6 , δ 40.45). 31P NMR chemical shifts are reported in parts per million downfield from H3PO4 and referenced to an external 85% solution of phosphoric acid in D2O. Multiplicity was indicated as follows: s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), dd (doublet), br s (broad singlet). Coupling constants were reported in hertz (Hz). Elemental analyses were performed at the KAUST analytical core lab using a Flash 2000d Thermo Scientific CHNS Analyzer. PN3P ligand was synthesized using reported procedure [61].The PN3P ligand (1.00 ​g, 2.50 ​mmol) was weighed in a Schlenk flask along with NiCl2 (596 ​mg, 2.50 ​mmol), 15.0 ​mL of dry THF was added, and the solution was stirred overnight. Orange product insoluble in THF was obtained after drying off the solvent under vacuum. It was then washed with THF. Crystallization was done by dissolving orange product in MeOH. Orange solid; 70% yield. 1H NMR (600 ​MHz, DMSO‑d6 ): 7.60 (t, 1H, J ​= ​12 ​Hz), 7.24 (m, 2H), 6.44 (d, 2H), 3.60 (t, 3H, J ​= ​12 ​Hz), 3.36 (s, 12H), 2.30 (s, 1H), 1.51 (t, 36H, J ​= ​12 ​Hz); 13C NMR (151 ​MHz, DMSO‑d6 ): 162.6 (t, NC–NH), 141.7 (s, pyridine ring carbon), 97.9 (d, pyridine ring carbon), 34.1 (d, (CH3)3–C), 28.3 (d, (CH3)3–C); 31P NMR (242.93 ​MHz, DMSO‑d6 ): 99.03 (s); HRMS (ESI): calcd. for C21H41N3P2Cl1Ni, m/z 491.18, found 491.189. Anal. Calcd. for C21H41N3P2Cl2Ni: C, 47.85; H, 7.84; N, 7.97. Found: C, 47.80; H, 7.81; N, 7.95. UV–vis (THF, [1 ​× ​10−4]), λ max, nm (ε, L mol−1 cm−1): 277 (3568), 324 (12,144), 357 (3896), 464 (914).To a solution of 1 (50 ​mg, 0.094 ​mmol) in 10 ​mL of dry THF in glovebox at room temperature was added sodium tetraphenylborate (32.5 ​mg, 1.00 equiv.), and the solution was stirred for 1 ​h. Then, the solution was filtered through a small Celite plug and concentrated under reduced pressure to afford orange solid (73 ​mg, 95%). Complex show broadened signals for t Bu groups of the ligand arm in the 1H NMR spectra due to hindered rotation. Orange solid; 70% yield. 1H NMR (400 ​MHz, CD3OD): 7.32 (br), 6.98 (br), 6.83 (br), 6.30 (br), 3.75 (br), 3.33 (br), 1.88 (br), 1.56 (br); 13C NMR (125.75 ​MHz, CD3OD): 163.80 (m, BPh4), 142.46 (s, NC–NH), 135.39 (s, pyridine ring carbon), 121.37 (s, BPh4), 125.15 (s, BPh4), 97.95 (s, pyridine ring carbon), 27.39 (s, (CH3)3–C), 24.99 (s, (CH3)3–C); 31P NMR (202.45 ​MHz, CD3OD): 98.72 (s); Anal. Calcd. for C45H61N3P2Cl1B1Ni: C, 66.65; H, 7.58; N, 5.18. Found: C, 66.60; H, 7.50; N, 5.10. UV–vis (THF, [1 ​× ​10−4]), λ max, nm (ε, L mol−1 cm−1): 267 (sh, 10,626), 274 (sh, 8023), 291 (sh, 5292) 317 (13,077), 353 (5278), 467 (1530).Method 1: Chemical reduction using CoCp2. To a solution of 3 (20 ​mg, 0.024 ​mmol) in dry THF was added cobaltocene (5 ​mg, 1.00 equiv.) at room temperature. The solution turns red, and after 2 ​h of stirring, the mixture was filtered through a Celite plug. The THF was evaporated to afford a red solid (8.9 ​mg, 80% yield). Complex in dry acetone was characterized by EPR spectroscopy (see main text). 1H NMR (600 ​MHz, C6D6): 8.09 (br s), 5.96 (br s), 1.12 (br t), −50.83 (br s). UV–vis (THF, [1 ​× ​10−4]), λ max, nm (ε, L mol−1 cm−1): 275 (sh, 15,755), 321 (sh, 11,199), 342 (13,410), 397 (2638), 470 (1112); Method 2: Generation of NiI complex by electrochemical reduction. NiI complex was prepared via controlled potential electrolysis at −1.55 ​V vs. Fc+/0 in acetonitrile. A constant voltage was applied using graphite rod working electrode, a platinum wire counter electrode and an Ag/AgCl/1 ​mol/L KCl reference electrode to orange solution of 3 (1.0 ​mmol/L of complex using 0.1 ​mol/L n Bu4NPF6 as the electrolyte in dry and degassed CH3CN). The voltage was applied for half an hour until the current measures was less than 1% of the starting current. At this point, the solution had become red. The nature of the product formed was confirmed by comparing its low-temperature EPR spectra and g tensor values with those from the chemically reduced 4.The change in the oxidation states of complex 1 in organic media was investigated via spectroelectrochemical technique. Chronocoulometric experiments were performed on compound 1, where it was reduced electrochemically at −1.60 ​V vs. Fc+/0 over time, and the gradual change in its optical spectra was monitored. The electrochemical reductions were performed at a slightly cathodic direction compared to the reduction peak maxima to ensure the complete reduction during chronocoulometry.The bulk electrolysis experiments of complexes 1 and 2 were performed in a four-neck glass vessel (volume 10 ​mL including the headspace) where three of the necks were occupied with a coiled Pt wire as counter electrode, Ag/AgCl (in saturated KCl) as a reference electrode, and a carbon rod (surface area 1.7 ​cm2) as working electrode, respectively. The last of the necks was connected with argon flow tubing. During the experiment, 10 ​mL of complexes (1.0 ​mmol/L catalyst, 10.0 ​mmol/L substrate) were added in the vessel, all the electrodes (along with a magnetic bead) were inserted along with a rubber septum cap (in a gastight manner), and the solution was purged with argon for 30 ​min. Then, the purging was stopped and the chronocoulometric experiment was started at corresponding catalytic potentials (−1.1 ​V vs. Ag/AgCl for complex 1 in the presence of BnBr). The solution was continuously stirred with a magnetic stirrer during the experiment.We acknowledge the generous financial support from King Abdullah University of Science and Technology (KAUST).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.gresc.2020.10.001.

We present the mechanistic understanding of an electrochemically-driven nickel-catalyzed coupling reaction. Computational analysis reveals that the spin density is mostly residing on the nickel (Ni) center when NiII is reduced to NiI. Ni-mediated halogen atom abstraction through outer-sphere electron-transfer pathway to yield coupling products under mild conditions is demonstrated. Importantly, we have elucidated the role of NiI and Ni0 for successive coupling of benzyl bromide and benzyl chloride derivatives, respectively, to corresponding bibenzyl products. The Ni-catalyst bearing a PN3P-ligand is an effective catalyst, producing a strong ligand effect on the reactivity and selectivity for the homocoupling reactions.

Furfural, obtained via acid-catalyzed dehydration of pentose such as xylose and arabinose or via fast pyrolysis of biomass, is one of the most promising biomass-derived platforms as a building block in the bio-refinery approach [1,2]. Selective hydrogenation of furfural has gained much attention for the production of furfuryl alcohol (FA), an important intermediate for the manufacturing of lysine, ascorbic acid (vitamin C), plasticizers, dispersing agent, and lubricants and is mainly used in the manufacture of resins, synthetic fibers, and agrochemicals [3]. In furan industrial process, copper chromite has traditionally been used as the catalyst for hydrogenation of furfural to FA [2,3]. However, concern of the environmental toxic impact due to the chromium presence in the catalysts led to the development of chromium free catalysts.Platinum-based catalysts are known as highly efficient catalysts used in the hydrogenation of furfural to FA because of their ability to hydrogenate the CO under mild conditions [3–5]. Pt nanoparticles in the range of 3–7 nm were found to favor the hydrogenation of furfural to FA whereas Pt particle size less than 3 nm promoted the decarbonylation of furfural to furan [3,6]. However, modification of Pt catalysts by alloying [7,8], adding promoters [7,9], and by using strong metal-support interaction to induce electronic effects [8,10] show significant impact on both activity and selectivity towards FA of the catalysts in furfural hydrogenation. For examples, The metal-support interaction on Pt/TiO2 induced hydrogen spillover, leading to the formation of furfuryl-oxy intermediate on the titania support [11,12]. Synergetic effect between Pt and Co on carbon obtained by co-impregnation was reported to be beneficial in furfural hydrogenation and product selectivity can be adjusted depending on the weight ratio of Pt and Co [13,14].In this study, monometallic Pt/TiO2 and bimetallic Pt-Co/TiO2 were prepared by flame spray pyrolysis (FSP) and studied in the liquid-phase selective hydrogenation of furfural to FA under mild conditions. The characteristics of Pt-Co/TiO2 formed by one-step FSP were quite different from those obtained by co-impregnation as illustrated by various characterization techniques such as CO pulse chemisorption, H2-temperature-programmed reduction (H2-TPR), transmission electron spectroscopy (TEM), and Fourier transform infrared (FTIR) spectroscopy of adsorbed CO. A structural-activity relationship of these FSP-PtCo/TiO2 catalysts was proposed.The Pt/TiO2 (0.7 wt% Pt) and Pt-Co/TiO2 (0.7 wt% Pt and 0–0.4 wt% Co) catalysts were prepared by FSP method according to the procedure described in Ref. [15] using platinum (II) acetylacetonate (Pt(C5H7O2)2, 99.99%, Aldrich), cobalt naphthenate (CoC22H14O4, 6 wt% in mineral spirits, Aldrich), and titanium (IV) butoxide (Ti(OCH2CH2CH2CH3)4, 97%, Aldrich) as Pt, Co, and TiO2 precursors, respectively. The catalysts prepared by FSP are denoted as (F) Pt/TiO2 and (F) Pt-Co/TiO2, respectively. The TiO2 supported Pt and PtCo catalysts were also prepared by incipient wetness impregnation using platinum (II) acetylacetonate (Pt(C5H7O2)2, 99.99%, Aldrich), cobalt naphthenate (CoC22H14O4, 6 wt% in mineral spirits, Aldrich) as Pt and Co precursors and P-25 TiO2 as the support. The catalysts were dried in an oven at 100 °C overnight, and calcined in air at 500 °C for 2 h. The impregnated catalysts are denoted as (I) Pt/TiO2 and (I) Pt-Co/TiO2, respectively.The XRD patterns were recorded using a Bruker D8 Advance X-ray diffractometer with Ni-filter CuKα radiation. The actual amount of Pt and Co loadings in the samples were analyzed by AAS. The N2 physisorption was conducted by using a Micrometrics ASAP 2020 instrument with the Brunauer–Emmett–Teller (BET) method. The TEM observations were performed with a JEOL JEM 2010 transmission electron microscope operated at 200 kV and equipped with LaB6 thermoionic electron-gun, an UHR pole piece (point resolution 0.196 nm) and a Pentafet Link-INCA EDX spectrometer (Oxford Instruments). The percentages of Pt dispersion were measured by CO pulse chemisorption using a Micromeritics ChemiSorb 2750 equipped with ChemiSoft TPx software. Prior to chemisorption, the catalyst was reduced with hydrogen (25 cm3/min) at 500 °C for 2 h and then cooled down to the room temperature under helium flow (25 cm3/min). The H2-TPR was carried out on a Micromeritics ChemiSorb 2750 with ChemiSoft TPx software. The sample was pretreated at 150 °C under N2 flow (25 cm3/min) for 1 h. The FT-IR spectra of adsorbed CO were collected using FTIR-620 spectrometer (JASCO) with a MCT detector at a wavenumber resolution of 2 cm−1. The sample was heated to 300 °C and reduced by H2 for 30 min.Prior to the reaction test, the catalyst was reduced with hydrogen (25 cm3/min) at 500 °C for 2 h. Approximately 0.05 g of catalyst was dispersed into the reactant mixture of 50 μL furfural (99%, Aldrich) and 10.0 mL methanol (98%, Aldrich) in 100 cm3 stainless steel autoclave reactor (JASCO, Tokyo, Japan). The selective hydrogenation reaction was carried out at temperature of 50 °C and H2 pressure of 2 MPa for 2 h. After that, the liquid product was collected and analyzed by a gas chromatography equipped with a flame ionization detector (FID) and Rtx-5 capillary column.Based on the XRD patterns of the Pt/TiO2 and Pt-Co/TiO2 catalysts (Fig. 1 ), the crystalline phases of TiO2 consisting of anatase phase at 2θ = 25° (major), 37°, 48°, 55°, 56°, 62°, 71°, and 75° and rutile phase at 2θ = 28° (major), 36°, 42°, and 57° [16] were observed. The characteristic peaks for Pt metal/Pt oxides and Co metal/Co oxides could not be detected due probably to the low amount of metal (Pt and Co) loading and/or high dispersion of these metals on the TiO2 support. As shown in Table 1 , the TiO2 anatase phase composition of (I) Pt/TiO2 (85.8%) was slightly lower than that of (F) PT/TiO2 (89.6%). Adding Co by different methods led to different effects on the TiO2 phase composition. When Co was co-impregnated with 0.7 wt% Pt on the TiO2 support, the anatase phase composition increased from 85.8 to 90.5% with increasing Co loading from 0.04 to 0.2 wt%. Because the ionic radius of Co2+ (0.075 nm) is not much different from Ti4+ (0.061 nm), it has been postulated that Co2+ ions could enter into TiO2 anatase structure and inhibit the phase transformation from anatase to rutile [17]. Upon substitutional dopants, the level of oxygen vacancies decreased, inhibiting the rutile phase transformation [18,19]. On the contrary, the TiO2 anatase phase composition of the FSP-derived bimetallic PtCo catalysts was significantly decreased to ~70% comparing to the monometallic (F) PT/TiO2. The ionic substitution of Ti4+ with metal cationic ions could occur during the particle formation step during FSP. The replacement of Ti4+ with metal cationic ions having lower valences such as Co2+ would generate more oxygen vacancies, promoting phase transformation from anatase to rutile along the flame. Simultaneous formation of TiO2 support and Co metal by using polymeric precursor [20] or by sol-gel technique [21] also showed similar effect of TiO2 phase transformation as in the FSP method. The average crystallite size of anatase phase TiO2 for the (F) PT/TiO2 remained unaltered at 12 nm upon Co loading by FSP method whereas those obtained by co-impregnation led to a slight increase of anatase phase TiO2 crystallite size from 12 to 16 nm.The BET surface areas of the Pt-based catalysts prepared by impregnation were in the range of 52–60 m2/g, and were slightly larger than those prepared by FSP (43–50 m2/g). Regardless of the preparation method used, Co addition as a second metal led to the increment of the BET surface area. For a similar amount of metal loading, Pt dispersion of the Pt-based catalysts prepared by impregnation was higher than that prepared by FSP. During FSP synthesis, suppression of CO chemisorption on surface Pt sites may be due to partial coverage of metal surface by the support matrix [22,23]. However, addition of Co as a second metal resulted in higher Pt dispersion regardless of the preparation method used. The interaction between Pt and Co was reported to compensate the electron deficiency of Pt sites by d-electron rehybridization, thus enhancing the adsorption ability of CO [24].The H2-TPR profiles of the catalysts are shown in Fig. 2 . All the synthesized Pt-based catalysts showed three major reduction peaks at 94 - 105 °C, 313–380 °C, and 509–687 °C, which were attributed to the reduction of PtOx particles to metallic Pt0 particles, the reduction of Pt species interacting with the TiO2 support in the form of Pt-TiOx interface sites, and the reduction of surface capping oxygen of TiO2 support, respectively [25]. It is obviously seen that the reduction peak of the Pt-TiOx interface sites for the (F) Pt/TiO2 catalyst shifted towards higher temperature and became broader compared to the (I) Pt/TiO2, indicating the stronger metal-support interaction induced by FSP method. For both cases, Co addition as a second metal in the Pt-based catalysts further shifted the reduction temperature of the Pt-TiOx species, due probably to the PtCo interaction and/or the migration of Co particles onto the Pt surface (decoration effect) [26]. Moreover, the reduction of surface capping oxygen of TiO2 drastically shifted from 545 °C for the monometallic (I) Pt/TiO2 to 615–687 °C for the bimetallic (I) Pt-Co/TiO2 while those of the FSP made catalysts remained unchanged at 510 °C. On the other hand, for bimetallic Pt-Co/TiO2, the Pt-TiOx reduction peak of (I) Pt-Co/TiO2 were located at higher temperatures compared to (F) Pt-Co/TiO2, which could be attributed to stronger metal-support interaction upon Co addition on the impregnated catalysts. As a consequence, the (I) Pt-0.2Co/TiO2 showed higher conversion and selectivity of FA compared to (F) Pt-0.2Co/TiO2, which was consistent with the metal-support interaction from H2-TPR. Nevertheless, it can be suggested that Co addition on FSP-made catalyst was not necessary for improvement of the interaction between metals and support in order to enhance the selectivity to FA.As observed from the TEM images (Fig. 3 ), the monometallic (I) Pt/TiO2 catalyst was significantly composed of very small and narrowly distributed nanoparticles (≤ 2 nm). Very rarely large (10–100 nm) Pt particles could be observed. (F) Pt/TiO2 catalyst showed three types of nanoparticle sizes. First, a collection of very small nanoparticle (≤ 2 nm) similar to the one found on (I) Pt/TiO2, then a collection of nanoparticles between 3 and 10 nm, and finally some very large spherical nanoparticles with sizes of several tens to several hundreds nm. It is suggested that (I) Pt/TiO2 essentially showed very small Pt nanoparticle sizes (≤ 2 nm) and the selectivity of FA was significant lower than the (F) Pt/TiO2. The selectivity of FA may depend on the Pt particle size. It has been report that Pt nanoparticles of sizes less than 2 nm supported on the silica are highly selective towards decarbonylation of furfural to main product of furan over vapor phase furfural hydrogenation [6].Concerning the Co addition as a second metal, (I) Pt-0.2Co/TiO2 was essentially consisted of very small nanoparticles as (I) Pt/TiO2. The average particle size of these nanoparticles slightly increased with the addition of Co (1.7 nm for (I) Pt-0.2Co/TiO2 and 1.3 nm for (I) Pt/TiO2) and the particle size distribution was substantially broader. There are also very rare of larger particles. The (F) Pt-0.2Co/TiO2 catalyst also showed characteristics similar to those of (F) Pt/TiO2. Two types of nanoparticles and some large spherical ones. The smaller nanoparticles also slightly increase in size of 1.9 nm for (F) Pt-0.2Co/TiO2 compared to 1.5 nm for (F) Pt/TiO2. The intermediate nanoparticles were not vary too much in size (6.8 nm for (F) Pt-0.2Co/TiO2 compared to 6.2 nm for (F) Pt/TiO2). The large spherical particles have several hundreds nm. The larger spherical particles that originated by FSP method were much larger than the size of the grains of titania and thus their interaction was poor, resulting to reduce the catalytic activity. The IR spectra of adsorbed CO on the different catalysts reduced at 500 °C are shown in Fig. 4 . The bands at 2188–2185 cm−1 represented to CO formed with Ti4+ cations on the surface [27]. The band at 2086–2078 cm−1 was assigned to CO linearly adsorbed on low - coordination Pt (Pt0-CO) or Co (Co0-CO) atoms on edge sites, while the band at 1850 cm−1 was assigned to CO bridged between two Pt metal atoms (Pt0-CO-Pt0). It has been reported that linear-type adsorbed CO dominated on small Pt particles while bridge-type adsorbed CO formed mainly on larger Pt particles [28,29]. The ratio of Pt on linearly CO adsorbed (Pt0-CO) to bridged CO adsorbed (Pt0-CO-Pt0) atom was high due to the presence of very small Pt particles size (≤ 2 nm), which was consistent with the average particle size observed from TEM and further increase selectivity of FA with the Co addition. The Pt0-CO band center of (F) Pt/TiO2 shifted to lower frequency with lower intensity compared to (I) Pt/TiO2. The lower shift and increase in intensity of the Pt0-CO band were also observed in bimetallic (F) Pt-0.2Co/TiO2 compared to (I) Pt-0.2Co/TiO2.In this study, FA was the only desired product being formed and 2-furaldehyde dimethyl acetal was the solvent product (SP) which can be generated upon furfural acetalization when using methanol as the solvent [3] in the presence of metal catalysts. The catalytic performances of all the Pt-based catalysts in the liquid-phase selective hydrogenation of furfural to FA are summarized in Table 2 . The furfural conversion and selectivity to FA over (I) Pt/TiO2 catalyst were 84.6% and 71.5%, respectively. Modification of (I) Pt/TiO2 by Co addition, furfural conversion and selectivity to FA increased with increasing Co loading up to 0.2 wt% in which furfural was fully converted within 2 h of reaction time and high selectivity to FA up to 97.5% was achieved. Improvement in furfural conversion upon Co addition by co-impregnation was correlated with high Pt dispersion and high TiO2 anatase phase composition. The subsurface oxygen vacancies on the reduced anatase TiO2-supported metal catalysts were found to be favorable adsorption sites for H atoms [30]. On the other hand, the hydrogen would weakly interact with the rutile TiO2-supported ones [31]. In addition, increase in the selectivity to FA could be attributed to the formation of Pt-TiOx interface sites, in which oxygen atom in the CO group of furfural could be coordinated via a lone pair of electrons, promoting the selective activation of carbonyl bonds in furfural [32]. In the perspective of electronic effect, the stronger interaction between Pt and TiO2 support indicated more electron transfer from support to metals, thus leading to the formation of electron-rich metal particles. The electronegativity of metal catalysts has been reported to affect the CO activation in furfural hydrogenation [32]. Stronger electronegativity of metal particles (or electron-rich metals) could activate the CO bond towards the furfural hydrogenation to FA. Similar results have been reported over Cu/TiO2-SiO2, Au/Al2O3, Fe-promoted NiB amorphous alloy catalyst, Au/ZrO2, and Au/TiO2.Furfural conversion of the monometallic (F) Pt/TiO2 was not much different from the (I) Pt/TiO2 catalyst but the selectivity to FA was found to be largely improved. This improvement was attributed to the stronger metal-support interaction upon FSP. However, Co addition upon FSP slightly improved the selectivity to FA but decreased furfural conversion. It is suggested that the presence of higher rutile TiO2 phase diminish the furfural conversion despite its high Pt dispersion based on CO chemisorption ability. Nevertheless, it could be said that Co addition into the Pt-based catalysts upon FSP was unnecessary for modification of the interaction between metals and support in order to enhance the selectivity to FA. Comparing the bimetallic Pt-0.2Co/TiO2 catalysts prepared by impregnation and flame spray pyrolysis, the (I) Pt-0.2Co/TiO2 catalyst showed higher furfural conversion and selectivity to FA than the (F) Pt-0.2Co/TiO2 catalyst because of higher Pt dispersion, TiO2 anatase phase composition, and the strong interaction between metals and support. Among all Pt-based catalysts in this study, the (I) Pt-0.2Co/TiO2 exhibited the highest FA yield at 97.5%. The activity of 1st cycle and 2nd cycle for (F) Pt-0.2Co/TiO2 catalyst were 71.1% and 67.4%, respectively. It can be observed that the activity decreased by 5.2% after 2nd reaction. Table 3 shows a comparison between the Pt-based catalysts in this study and those reported in the literature. It is notable that one of the best results with furfural conversion (100%) and FA selectivity (95.7%) could be obtained over 2 wt% Pt-1 wt% Re/TiO2-ZrO2 [10], however more severe conditions (130 °C, 5 MPa of H2) and long reaction time (8 h) were used in the presence of ethanol as a solvent. The catalysts in the present studies achieved similar conversion/selectivity under milder reaction conditions.While Co addition via co-impregnation on Pt/TiO2 catalysts improved both furfural conversion and selectivity to FA due to higher TiO2 anatase phase composition, high Pt dispersion, and stronger interaction between metals and support, formation of Pt-Co/TiO2 by one-step FSP resulted in a decreased furfural conversion despite its high Pt dispersion due to the acceleration of rutile phase transformation of the TiO2. However, a largely improvement of FA selectivity on the FSP-made catalysts was attributed to the stronger metal-support interaction upon FSP as revealed by the higher reduction temperature of the Pt-TiOx interface sites. Thus, modification of the Pt/TiO2 by Co addition was unnecessary for catalyst improvement when using the FSP Pt/TiO2 catalysts.Weerachon Tolek: Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection, Writing first draft.Kitima Khruechao: Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection.Boontida Pongthawornsakun: Data analyzing and discussion.Okorn Mekasuwandumrong: Data analyzing and discussion.Francisco José Cadete Santos Aires: Data analyzing and discussion.Patcharaporn Weerachawanasak: Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection.Joongjai Panpranot: Acquisition of the financial support for the project leading to this publication, Reviewing and 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 declare the following financial interests/personal relationships which may be considered as potential competing interests:The financial supports from the Rachadapisek Sompote Endowment Fund for the Postdoctoral Fellowship, Chulalongkorn University for W.T. and the Thailand Science Research and Innovation for the Basic Research grant BRG6180001, IRN62W0001, and CAT-REAC Industrial Project RDG6250033 are gratefully acknowledged. The authors also would like to thank the Research Team Promotion grant from the National Research Council of Thailand (NRCT).

Flame spray-synthesized Pt/TiO2 and PtCo/TiO2 catalysts with 0.7 wt% Pt and 0–0.4 wt% Co were studied in the hydrogenation of furfural to furfuryl alcohol (FA) at 50 °C and 2 MPa H2. Particle formation under high temperature flame facilitated high Pt dispersion and formation of Pt-TiOx interface sites, which were beneficial for furfural conversion to FA. Modifying with Co accelerated rutile phase TiO2 formation, which strongly diminished hydrogenation activity on the (FSP)-PtCo/TiO2. On the other hand, (I)-PtCo/TiO2 prepared by conventional impregnation, anatase phase TiO2 was preserved (> 85%) and both furfural conversion and FA selectivity increased upon increasing Co loading.

Hydrogen gas (H2) is eco-friendly, sustainable, efficient, and has renewable properties. Moreover, it is a clean, efficient energy supporter for fuel cells and an auspicious alternative to control air pollution, crises of future fossil fuel, and problems of renewable resources [1–3]. The electro-splitting of water is assumed as a sustainable technique of H2 generation among the diverse methods of hydrogen generation [4,5]. It is confirmed that the electrolysis of alkaline water links to the small efficiency and high exhaustion of energy. While the utilization of acid solutions supplies a probable alternative to that matter [6]. Where HER in acidic aqueous solution necessitates lesser applied overpotential, which is more effective and economic than the reaction in alkaline aqueous electrolytes. Although, alkaline media is still utilized because of its possibility for prevailing the overall water electrochemical splitting reaction accompanied by the production of pure hydrogen gas accumulated at the negatively polarized cathode and oxygen gas at the positively polarized anode department concurrently [7,8].Noble metals including platinum are particularized as the most appropriate electrochemical catalysts for HER. This is owing to their low electro-reduction overpotentials and fast electron transfer kinetics [9–11]. On the contrary, their application in the industry of large-scale production of hydrogen is obstructed by the high-cost and rarity [12–14]. Subsequently, finding earth-abundant metals as alternatives to the rare and costly noble metal electrocatalysts is an indispensable request to consolidate the growth of the hydrogen production economy [15–18].Among electrocatalysts containing non-precious metals, nickel and nickel-based materials are strongly recommended as premium candidates to substitute Pt-based materials as electrocatalysts for HER in acidic aqueous media. This is attributed to their high stability and electrocatalytic activity [6,19,20]. The incorporation of Ni with different elements to yield alloys such as NiFe [21], NiCo [22], NiSe [23], Ni2P [24], NiCu [25], and NiMo [26] is a noteworthy protocol to fabricate electrochemical catalysts with enhanced performance of HER. Particularly, Ni–Co systems were previously examined and characterized to behave as efficacious electrocatalysts for HER. This is due to their intrinsic electrocatalytic validation and increases the corrosion resistance in comparison to the pure Ni in acidic media [27–29]. Nevertheless, Ni–Co system up to this time display considerably application in HER. This is due to their stability compared to noble metals in acidic media. For obtaining Ni–Co alloy of highly catalytic performance for the HER, two general strategies should be adopted. One of them is preparing catalysts of high surface area to dramatically raise the number of active centers, and the second is preparing catalysts of porous structures. This leads to enhance the electrical conductivity of the catalyst and increase the mass transport process for the species under the influence of the polarization of the electrode [30–33].Ni–Co alloys have been produced by various techniques including hydrothermal/solvothermal [34–37], polyol method [38,39], double composite template approach [40], chemical vapor deposition [41], mechanical alloying [42], non-aqueous ethylene glycol refluxing [43] and electrodeposition [44–46]. The electrodeposition technique has numerous benefits including low cost, facile, reproducible, and highly efficient. For that, it can extensively be used for fabricating thin-film coating [47,48]. Moreover, the electrochemical deposition technique does not demand hard conditions. Furthermore, the particle sizes and the structure can be readily controlled by simply adjusting the parameters of electrodeposition, such as time, concentration, temperature, or applied potential [48]. In this work, NiCo alloy flower-like structure loaded on C-steel, have been successfully fabricated via electrodeposition methods. The as-synthesized NiCo/steel inherit significant HER catalytic properties in HCl solution for the first time in that system. The constructed NiCo/steel catalyst possesses excellent electrocatalytic performance for HER. Furthermore, the obtained catalyst has higher anti-corrosion and excellent long-term stability in the aggressive acidic medium of HCl.Cobalt chloride hexahydrate (CoCl2·6H2O), nickel chloride hexahydrate (NiCl2·6H2O), and potassium chloride (KCl) were purchased from Merck. All electrolytes were prepared utilizing bi-distilled water. The bath of electrochemical deposition contains 0.02M CoCl2·6H2O, 0.05M NiCl2·6H2O, and 0.1M KCl. All the electrochemical experiments were implemented using a 3-electrode cell contains a Pt sheet and Ag/AgCl as a counter and reference electrodes, respectively. Cu sheet (1cm2) and carbon steel rode (4.55cm2) were used as working electrodes. All electrochemical experiments were accomplished by using a computerized Potentiostat/Galvanostat of VersaSTAT4. Potentiostatic electrodeposition method of Ni–Co alloys on carbon steel rods was applied at various potentials of −0.8, −0.9, −1, −1.1, and −1.2V for 1h.The chemical composition and morphology of the electrodeposited Ni–Co alloys were investigated using the scanning electron microscope of JEOL JSM-5500 LV, from JEOL company in Japan, supplemented with Microanalyzer of Energy Dispersive X-ray (EDX) driven via a system with Link Isis @ Software of model 6587 from Oxford, England. The electrodeposited films of Ni–Co alloy have been introduced for the measurements of XRD. The formed phases of the Ni–Co alloy deposits were detected via a phaser of Bruker Dron-2 table-top model using Cu Kα radiation source of wavelength 1.5418Å.Experiments of the electrochemical corrosion were achieved at 30°C in 1.0M HCl. The PDP for the bare and surface covered carbon steel rods were performed after 1h of dipping in the test electrolyte, by applying a potential ranged from −0.25 to +0.25V versus the open circuit potential (OPC, E corr) at a potential scan rate of 0.001Vs−1. The experiments of EIS were accomplished using a DC potential amplitude of 0.01V by applying a frequency ranged from 100,000 to 0.5Hz. EIS data has been fitted utilizing Z-view by selecting the most appropriate equivalent circuit.The HER effectiveness of the as-prepared catalysts was examined with linear sweep voltammetry (LSV) at a potential sweep rate of 0.005V/s in the test solution of 0.1M HCl.The deposition potential and other optimized conditions for the electrodeposition of NiCo alloys on the steel rod surface have been accurately determined via cyclic voltammetry, and via the variation the concentration of the electrodeposition bath components. The potentiostatic method (chronoamperometric) has been used for the achievement of the electrochemical deposition process of the alloys in demand. At ambient temperature and pH=6.5, suitable deposition potentials (chosen with the aid of cyclic voltammetry, near from the peak potential of reduction of Ni and Co ions) of −0.8, −0.9, −1.0, −1.1, −1.2V were applied for detecting the best potential at which the alloy is deposited as well, shown in Fig. 1 .It seems from Fig. 1 that the onset of the curves of current–time is distinguished by an abrupt convert up or down, depending on the applied potential. This may be attributed to the presence of a double layer between the negatively charged electrode surface and the cations of the electrolyte under consideration. Accordingly, instantaneous nucleation occurred for all investigated alloys particles as exhibited in Fig. 1. Then, the current is accompanied by a slight increase. This means that the electrodeposited layer still grows at this current which should lead to an increase in the electroactive surface area, during the deposition process. But there is a sudden increase of current (more negative) value in the case of the sample (e) at −1.2V. This can be explained as at the higher potential, the rate of electrodeposition increases and hence the nucleation also increases. This led to an accumulation of the loaded electrodeposited particles which cause a limitation of the electro-reduction reaction progress. Fig. 2 exhibits the XRD analysis of the electrochemical deposited Ni–Co alloys via the potentiostatic technique. The alloys have been electrodeposited at various deposition potentials as follows: a) −800, b) −900, c) −1000, d) −1100, and e) −1200mV versus Ag/AgCl, at pH=6.5, and room temperature on Cu substrate. The data of XRD demonstrate that crystalline Ni–Co alloy has been obtained as illustrated in Fig. 2.From the figure, it can be noticed that all samples contain four lines of Ni–Co appeared at 2θ =75.85, 51.14, 41.46, and 43.80 corresponding to Ni–Co FCC (220), Ni–Co FCC (200), Ni–Co FCC (111), and Ni–Co HCP (100), respectively. Also, it can be noticed that all samples contain nickel metal except sample (a). Where, at sample (a) the deposition potential is −0.8V. This potential is not sufficient for the deposition of nickel metal at these conditions. Furthermore, it can be observed that sample (c) has the highest percent of the phases Ni–Co FCC (200) and Ni–Co FCC (220) compared with the other samples. This will cause a difference in the measurable physical properties of this sample compared with the other samples as shown in the next sections.The surface morphology of the electrochemically deposited Ni–Co alloys at the optimized conditions was investigated using SEM, as shown in Fig. 3 . Image (a) shows the bare copper substrate which is chemically treated and etched (cleaned) with the Piranha solution before the electrodeposition process. As shown in Fig. 3, the surface morphology of the electrodeposited alloys heavily is controlled by the potential of the electrochemical deposition. It is noticed that the etched area of the bare Cu substrate has been covered with the electrodeposits of Ni–Co alloy, as shown in image (b). Also, it contains some cracked parts and large crystals. The shape of the electrodeposited Ni–Co has been changed by increasing the electrodeposition potential E d to −0.9V to be spongy like structure, as shown in image (c). A very interesting flowered like structure has been obtained at E d −1.0V which of course has a high surface area compared with other samples. When the E d was increased to −1.1V, smooth carrying small spherical particles and some large particles of structure like sticks were observed (image (e)). At the applied potential of −1.2V, the obtained structure looks like that obtained in the case of E d of −1.1V, but with larger spherical particles and sticks. These differences in morphology, of course, affect the physical properties of the obtained samples.Furthermore, the value of electrochemical deposition potential has a significant impact on the composition and stoichiometry of electrodeposited Ni–Co alloy, as shown in Fig. 4 . It can be observed that by increasing the applied deposition potential the percent of Ni in the electrodeposited alloy increases from 61.63% to 75.46% at the applied range of E d (from −0.8 to −1.2V). In contrast, by increasing the applied deposition potential the percent of Co in the electrodeposited alloy decreases from 38.37% to 24.54%. This may be interpreted that Ni ions are firstly electro-reduced due to that the charge on Ni ions is larger than that of Co ions and hence the Ni ions can be attracted towards the opposite polarized electrode faster than that of Co ions. Besides, the hydrogen evolution at high negative potentials can affect the completion of the electrodeposition of Co more than Ni.The data of potentiodynamic polarization for the uncoated and coated mild steel with the electrodeposited Ni–Co alloys have been examined (Fig. 5 ). Some important data including the current density of corrosion (I corr), potential of corrosion (E corr), and slopes of Tafel's plots (β c, β a) have been determined using curves of the polarization (Table 1 ). The efficiency of protection (η%) was intended via the values of current density of corrosion (I corr) using the following equation [49]: (1) η % = I corr ( B ) − I corr ( D ) I corr ( B ) × 100 where I corr(B) and I corr(D) are the current densities of corrosion for the uncoated and the coated steel samples, respectively. Based on the data in Table 1, the C-steel sample had a corrosion current density (I corr) of 3.9mAcm−2 attributed to the highly corroded of metal in the corrosive medium. The I corr of Ni–Co coated alloys decreased to starts from 0.490 and reaches to 0.026mAcm−2 for the electrodeposited Ni–Co alloys deposited at −0.8 and −1.0V, respectively. Then, I corr was increased to 0.065mAcm−2 for the Ni–Co alloy electrodeposited at E d =−1.2V. The decrease of I corr in case of coated steel samples indicating a protection impact for these coatings. This was ascribed to the existence of the deposited Ni–Co layer on the surface of C-steel, which acts as a protective layer (barrier) and reduces the contact between HCl and metal surface. Ni–Co alloy, deposited at E d =−1.0V, exhibits the highest protection among the studied alloy coatings. This due to the full covered flowered structure of the deposited film for that alloy, which prevents the reach of Cl− ions to the metal surface more than the other alloys and also the percent ratio of Ni and Co in the deposited alloy.It is observed from Fig. 5 and Table 1 that the corrosion potential (E corr) for all coated samples is more positive than the uncoated mild steel, which implied that samples coated with Ni–Co alloys enhance the barrier effect and protect the samples from corrosion. According to Table 1, β a and β c values for coated samples are slightly changed indicating the coating barriers are blocked the anodic sites at the same corrosion pathway [50]. It is noteworthy to emphasize that the good corrosion resistance of the prepared coated films can be partly related to their structure and denseness (i.e., lack of porosity). The porosity (P o) of the coating films of the electrodeposited Ni–Co alloys can be determined from the PDP measurements via the following Eq. [51]: (2) P o = R s R d × 10 − ( Δ E corr / β a ) where R d and R s are the polarization resistances of the deposited film and substrate, respectively, ΔE corr the potential difference between the coated substrate and the uncoated mild steel, and β a is the anodic slope of Tafel's plot for the substrate. The obtained P o values are in the range between 1.4% and 4.6% (Table 1). The higher porosity value 4.6% was obtained for the sample coated with Ni–Co (deposited at E d =−0.8V), which would facilitate Cl− anions diffusion towards the layer/bare interface.As mentioned above, that all the corrosion currents values for the Ni–Co coatings were smaller than the C-steel substrate. This displays that the critical coating thickness for the steel substrate protection was accomplished. On the other hand, the lowest corrosion current and highest protection capacity (η%) were obtained for the Ni–Co (3) sample at deposition potential −1.0V (Table 1). For the coatings at deposition potential −1.1 and −1.2V, the corrosion currents were found to be increased and η% decrease. The efficiency of corrosion protection of electrodeposited Ni–Co alloys can be returned to many important factors such as roughness and surface morphology, defects (porosity), phase composition, chemical composition, and grain size of the coating layer [52].To confirm the obtained results from the PDP experiments, the protection effect of electrodeposited Ni–Co films in the solution of 1.0M HCl was examined using EIS. The obtained data by EIS are shown in Fig. 6 (Nyquist plot) and Fig. 7 (Bode and Phase angle plots). Fig. 6 exhibits impedances spectroscopic behavior of the electrodeposited Ni–Co alloys that are uncommonly greater than that of the uncoated steel. Ni–Co films exhibit significant protective properties with increasing deposition potential until −1.0V after that the protective properties were decreased.The comparison between EIS experimental results and the fitted data for coated and uncoated C-steel substrates immersed in HCl was presented in Fig. 6b and c. A remarked fitting was obtained for the proposed model that utilized for all resulted data. In the case of the uncoated sample, a one-time constant only is observed; the proposed model of EIS is exhibited as in the inset of Fig. 6b. The proposed model can be identified as a simple equivalent circuit of Randel [53,54], comprises of parameters namely R s is the solution resistance which connected in series with the double layer capacitance (C dl) and series with the charge transfer resistance (R ct). Equivalent circuit characterizing the behavior of C-steel coated by electrodeposited Ni–Co samples is frequently suggested for the modified substrate surface by the electrodeposited alloys (inset of Fig. 6c). As in the exhibited circuit, C coat represents pseudo-capacitance of coating and R pore is the electrical resistance of pore against the ionic current that directed to the pores. Similar equivalent circuits have been used by many researchers to simulate EIS data for coated systems, for example, Ullal and Hegde in case of electrodeposited multilayer nanocomposite of Zn–Ni–SiO2 on mild steel [55], and by Cheraghi et al. [49] in the case of AISI 316L stainless steel covered by TiO2–NiO nanocomposite thin films in a solution of 3.5% NaCl. Fig. 7a and b exhibits the Bode and phase plots for C-steel substrate and Ni–Co deposits on C-steel substrate in the test solution of HCl after immersion for 30min at the corrosion potential E corr and temperature of 30°C. From Fig. 7a C-steel displays one-time constancy, it was accomplished that only one mechanism was proper for the corrosion process. However, after coating with electrodeposited Ni–Co, the curve shows two-time constancy. The first at higher frequencies (HF) can be connected with the coating's response; the second at low frequencies (LF) can be attributed to the process of corrosion. It could be elucidated that the electrochemical corrosion of samples in HCl is predominantly depending on the process of charge transfer. In general, the greater Z modulus LF manifests the best resistance against corrosion [56]. The obtained data in this study exhibited the highest Z modulus for Ni–Co coatings as shown in Fig. 7a. The bode phase plots of all the samples in Fig. 7b shows only one-time constancy in the case of the uncoated C-steel substrate, while in the case of coated samples showed two-time constancy. The electrochemical parameters related to EIS measurements were measured from curve fitting method and these are exhibited as in Table 2 . The obtained R s values are very low around 0.22–0.48 due to the higher conductivity of HCl solution (Table 2). By comparing the R ct between the coating samples which deposited at different deposition potentials, it is observed that the sample 3 (at deposition potential −1000mV) has the highest R ct and thus corrosion protection. The reason for this corrosion protection could be the higher amount of Ni in the film, which promotes corrosion resistance.Indeed, from Table 1, it is revealed that the R ct increased and the protection efficiencies improved by increasing the potential of the electrochemical deposition until −1000mV, then decreased. The highest P% of 88.72% was obtained for the films that electrochemically deposited at −1000mV.The high R ct values of deposited coatings in comparison to C-steel bare indicated that the Co–Ni deposited films can efficiently suppress the diffusion of corrosive electrolyte into the metal surface by forming a strong hurdle between the interface of the metal and the media of corrosion (HCl solution). Coating resistance (R c) is an apparent index that reflects the protective effect of the coating film, the higher R c value can lead to better corrosion protection. It is worth noting that EIS measurements and Tafel plots are consistent with each other. As a result, the coated substrates have better corrosion protection compared to the uncoated specimens. However, samples coated with Co–Ni deposited at −1000mV manifested the highest protection from corrosion in a test solution of HCl.The electrocatalytic activities for HER of the electrochemically deposited NiCo on C-steel substrate were performed in a solution of 0.1M HCl. Fig. 8a exhibits the LSV curves for the electrodeposited alloys of NiCo, by the side of those the blank C-steel electrode has been used for comparison. As observed, all the NiCo/steel electrodes exhibited better electrocatalytic activity than the blank C-steel indicating that NiCo/steel can provide as a high-performance cathode for electrocatalysis of HER. The high-performance of the electrocatalytic HER for NiCo/steel may be related to the synergistic catalytic activity of the alloyed Ni with Co [57,58]. One can be elucidated that aside from noble metals, the Ni and Co metal system manifested a significant HER activity among transition elements [59]. Moreover, it has proved that the NiCo alloy can elevate the splitting of the attached water on the electrode surface by gradually destroying the O–H bond. While the produced hydrogen as intermediates maybe get adsorbed on the surface of NiCo alloy [60]. Accordingly, it provides extra active sites on the surface and then accelerates the rate of HER.Among the NiCo catalysts, the Ni–Co (3) alloy exhibits considerable HER activity at the smallest start of overpotential, and an instantaneous increase of cathodic current under more negative potentials occurred. Moreover, the appropriated overpotential to generate a current density of −10mAcm−2 for the Ni–Co (3) alloy is solely −535mV. This value is lower than that of other NiCo alloys and the blank C-steel electrocatalysts as seen in Fig. 8b.The reaction mechanism of HER, generally, can be deduced from the plot of Tafel (ƞ vs log I). Moreover, the rate-determining step for the reaction of hydrogen evolution can be recognized. The linear cathodic part of the Tafels’ curves (Fig. 5) has been fitted according to the Tafel's equation (ƞ = a + β c log I; where “β c” is the Tafel slope of the cathodic line and “a” logarithm of the exchange current) giving cathodic values of Tafel slope for all NiCo/steel and blank C-steel catalysts as seen in Table 1. During the reaction of the evolution of hydrogen in the acidic medium, three possible intermediate steps can occur [61–63]. Form the Tafel slope, it can be elucidated that the first comprises proton-coupled electron transfer at the surface of catalyst producing adsorbed hydrogen, this step is a discharge one, recognized as the Volmer reaction: (3) H3O+ +e− =Hads +H2O (Volmer reaction, β c =120mV/dec). The following process is desorption which occurs by either mechanism of Heyrovsky, at which the adsorbed hydrogen combines with a hydrogen ion (proton) from the working solution, (4) Hads +H3O+ +e− =H2O+H2 (reaction of Heyrovsky, β c =40mV/dec). or by combining two hydrogen atoms that are adsorbed on the electrode surface, this known as the Tafel reaction: (5) Hads +Hads =H2 (reaction of Tafel, β c =30mV/dec). Despite that, it is a complicated matter to visualize a precise mechanism to diverse performance of electrocatalysts towards HER, the slope of Tafel is considered to be a perfect indication of the rate-determining reaction step. Regarding this work, the slopes of Tafel for the electrochemically deposited Ni–Co alloys are within the range of 40–120mV/dec. Accordingly, the mechanism of the hydrogen evolution for these modified electrodes can be ascribed to the Volmer–Heyrovsky mechanism, in which the adsorption of the hydrogen atom on the surface is the slow step that determines the reaction (r.d.s.). Where little slopes of the Tafel plot means a more catalyzed rate for HER can take place. One could be noted that the slope of Tafel for Ni–Co (3) alloy is found to be 81mV/dec, which is the smallest value compared with all values of the alloys under investigation, as listed in Table 1. This reveals that the HER at Ni–Co (3) catalyst is kinetically fast. It should be noted that Ni–Co (3) catalyst is the best HER catalyst among all investigated catalysts. Aside from the activity towards HER, the durability of the electrocatalyst is a significant property for the investigation of actual electrocatalytic efficiency.To investigate the stability for the long-term of all the NiCo/steel electrocatalysts and blank C-steel, chronoamperometry at fixed potential of −500mV versus Ag/AgCl for an hour in 0.1M HCl solution at room temperature should be carried out, as exhibited in Fig. 8c. Fig. 8c exhibits that, a higher current density of Ni–Co (3) than that of all other catalysts at the same overpotential (−500mV) which proves the high electrocatalytic activity of Ni–Co (3) catalyst for HER. An apparent change of current density can be also noticed after 1h at all investigated catalysts except for Ni–Co (3). Also, a long-term chronoamperometric test was performed for the most stable alloy (Ni–Co (3)) which exhibits and confirms the stability of the alloy form a long time (20h). No obvious current density change with time proves the excellent electrochemical stability of Ni–Co (3) compared to all of the electrochemically deposited Ni–Co catalysts, and blank C-steel in acidic media. Furthermore, it seems to be there is no change in the SEM and XRD data for all samples after chronoamperometry and HER experiments. Its expected results because they exhibited high corrosion resistance in this media and gave a considerable HER. Besides, they exhibited a reproducibility and repeatability.Potentiostatic route was applied for the synthesis of Ni–Co alloys at various deposition potentials, neutral media, and room temperature and without any additives. SEM images show different structures by varying the deposition potential and flowered like structure was obtained at a constant potential of −1000mV. This alloy gives a chemical composition of Ni70Co30 under EDX analysis. The increase of deposition potentials increases Ni content and decreases Co content in the electrodeposited alloys. The results of XRD confirm a good agreement with the crystalline phases (fcc) and (hcp) for the electrodeposited Ni–Co alloys. According to potentiodynamic polarization and EIS measurements, it has been concluded that all the rate values of the electrochemical corrosion for the Ni–Co coatings were smaller than that of the carbon steel substrate. Furthermore, the lowest corrosion current and highest protection efficiency were obtained for Ni–Co (3) sample at deposition potential −1.0V. This could be correlated to several substantial factors including phase composition, chemical composition, porosity, and surface morphology. It can be concluded that this alloy of unique properties is qualified and promising to be a stable and efficient electrocatalyst material for HER in HCl solution.None declared.

Ni–Co alloy of flowered like structure is successfully electrodeposited from neutral aqueous solutions of NiCl2 and CoCl2 salts as precursors at room temperature. The prepared alloy at our optimized conditions has a high protection effect from the electrochemical corrosion of steel in the hydrochloric acid solution. The electrochemically deposited alloys are examined via X-ray diffraction (XRD) giving two main phases of (fcc) and (hcp). The morphology is investigated via scanning electron microscope (SEM) showing flowered like structure for the alloy that is electrodeposited at −1.0V only. The obtained data of the potentiodynamic polarization (PDP) and spectroscopy of electrochemical impedance (EIS) demonstrates that Ni–Co coatings have high stability in the highly corrosive media, especially the alloy of the flowered like structure. This alloy of unique properties is subjected to be an efficacious electrocatalyst for the reaction of hydrogen evolution (HER) in HCl solution.

Multi-metallic formulations have raised a lot of interest in catalysis and electrochemistry because of the synergies that may be achieved by combining different elements [1–4]. The possibility to dilute expensive noble metals into base metals, while maintaining acceptable or even improving catalytic properties, is also clearly valuable. The catalytic activity in terms of reactant conversion rates and selectivity to various reaction pathways will directly depend on the surface composition and local geometric arrangements of atoms [5]. The bulk or surface composition or structure of nanoparticles will often be modified under reaction conditions [6,7], making the use of in situ and operando methods a requirement to obtain meaningful structure-activity relationships. in situ and operando IR spectroscopies applied to catalysis have been primarily used to determine the nature and reactivity of adsorbates present under reaction conditions [8–12]. This is because many vibrational modes associated with adsorbates bonds are highly IR sensitive and lead to sharp bands enabling detection at the nanomole levels using only milligrams of catalysts [13]. In contrast, the vibrational bands of catalysts are difficult to investigate by IR because the corresponding lattice vibrations lead to very broad absorption bands typically located in the far-IR region.IR can yet lead to indirect information about the structure of nanoparticles under reaction conditions if an adsorbed intermediate leads to a measurable signal. This can be the case for reactions involving strongly IR-sensitive molecules such as carbon oxides and unsaturated hydrocarbons. The examples dealt within the present review focus on the interpretation of IR spectra of CO adsorbed on various bimetallic catalysts. A particular interest is given to Au, Pt, Pd and Sn-containing bimetallics, as these elements commonly appear in catalytic formulations.IR spectra analysis is yet not straightforward, as exemplified by recent misinterpretations [14–17]. Even the case of monometallic samples is complex as carbonyl band position will depend on several parameters such as surface coverage, CO coordination mode, coordination number and the oxidation state of the metallic site. The case of monometallic Pt-based catalysts was recently reviewed in details and is therefore not recalled here [18,19]. The interpretation of spectra obtained over Au monometallic materials is examined in the first section and its impact on the interpretation of spectra obtained on some bimetallic materials will then be explored. This contribution discusses data obtained from transmission IR, diffuse reflectance FT-IR (DRIFTS) and reflection absorption IR (RAIRS) spectroscopies.Gold can lead to highly active CO oxidation catalysts at low temperatures [20] and has been used for many other reactions [21]. There have been numerous characterisations of CO adsorption on Au-based samples. Bands in the range 2120−2080 cm−1 have typically been assigned to CO adsorbed at steps or defects on metallic gold [22,23], while those located within 2080−2000 cm−1 have been assigned to CO adsorbed on negatively charged Au clusters [24–27]. Bands located at 2140−2120 and above 2140 cm−1 have been assigned to CO adsorbed on positively charged Auδ + with low (δ < 1) and high (1 or 3) charges, respectively. [28–31]The assignment of Au−CO bands is in fact not straightforward, especially below 2080 cm−1. Some of us reported on the growth of a band at 2070−1950 cm-1 derived from neutral Au° (2098 cm-1) and Auδ + (2125 cm-1) species during an operando investigation of the water-gas shift reaction over CeO2-supported Au catalysts (Fig. 1 .A) [29]. The two latter species were actually quantitatively interconverted into the former, as an isosbestic point was observed (red circle in Fig. 1.A). It was suggested that the 2070−1950 cm-1 band could possibly be associated with negatively charged gold particles that was gradually building on CeO2-x oxygen vacancies.Behm and co-workers had reported a similar interconversion of a band at 2110 cm−1 into one at 2060 cm−1 in the case of TiO2-supported Au32. These authors stressed that a CO pressure higher than 1 kPa was necessary for the interconversion to occur and that a spreading of the Au also took place, deduced from Au(4f)/Ti(2p) XPS signal ratios. The 2060 cm−1 band was not formed in the presence of O2 and was removed when O2 was introduced after CO. Behm et al. concluded that the 2060 cm−1 band was due to the formation of negatively charged Au clusters located at TiO2-x oxygen vacancies.Bianchi and co-workers have observed a similar interconversion over Au supported on alumina, which importantly is a non-reducible support. A band at 2095 cm−1 (Au°) was transformed into a broader one between 2070 cm-1 (Fig. 1.B) [33]. The 2070 cm−1 band was assigned to an Au phase restructured by CO. The restructuring happened faster with increasing CO pressures. The heat of CO adsorption at zero coverage on the restructured 2070 cm-1 species (100 kJ mol−1) was significantly higher than that on the species at 2095 cm−1 (62 kJ mol−1), thereby providing the driving force for the restructuration.The presence of an Au−CO band at a wavenumber as low as 2060 cm−1 has been reported by Lee and Schwank on Au supported on SiO2, another non-reducible support [34]. We were able to observe the interconversion at 50 °C of a band at 2113 cm−1 (Au°) into one located at 2076 cm−1 over an Au/SiO2 pre-reduced with H2 (Fig. 2 .A). Isosbestic points were observed in most cases (red circles in Figs. 1.A, 1.B and 2.A), signaling the quantitative interconversion of one species or site into the other one.The carbonyl bands located between 2080−2060 cm−1 formed upon exposure to CO of Au supported on non-reducible supports such as alumina33 and silica34 are unlikely to be related to the formation of negatively charged Au species, since no oxygen vacancies can be created under the corresponding conditions on these supports. Therefore, another explanation has to be put forward to account for the evolution of the spectra and the origin of these low frequency bands.The propensity of Au surfaces to reconstruct has long been reported. Theoretical work has even proposed that the shape of the whole nanoparticle may vary upon CO adsorption [36]. The tendency for steps on surfaces vicinal to Au(111) to exhibit a much higher kink density when exposed to CO than those on other metals has been stressed [37,38]. Piccolo et al. [39] observed a roughening of the step edges near Au(111) terraces by scanning tunnel microscopy (STM) at room temperature and CO pressures as low as 1 torr. A plethora of rough islands was then noted at 100 torr. The corresponding reflection absorption IR spectroscopy (RAIRS) signal of adsorbed CO was too low to be observed on the initial surface at low CO pressures, related to the weak adsorption of CO on Au(111) terraces [40,41]. Yet, the reconstructed surface exhibited a clearly visible IR band at 2060 cm−1. DFT calculations indicated that such wavenumber was consistent with on-top CO adsorption on kinks such as those found on Au(874) surfaces [39].RAIRS experiments over Au(110) [22] and Au(332) [42], which are stepped surfaces that do not exhibit kinks in the absence of reconstruction, revealed bands at 2115 and 2125−2110 cm−1, respectively. These bands can be primarily assigned to CO adsorbed on step edges rather terraces, since CO adsorption on the latter is much weaker. Nakamura et al. [43] reported similar experiments over Au(311), Au(100) and Au(111). The low index surfaces Au(100) and Au(111) yielded IR bands at 2076 and 2081 cm−1, respectively. These bands were only measurable above 0.1 torr of CO and, though not mentioned by Nakamura et al., probably corresponded to kinks of a reconstructed surface (i.e. no longer Au(100) and Au(111)) as discussed above.The case of the Au(311) surface reported by Nakamura et al. is puzzling43. A band at 2117 cm−1 was initially observed, which was then gradually replaced at higher pressure (and incidentally time) with one at 2071 cm−1. The authors concluded that the 2117 cm−1 was due to CO adsorbed on step edges, which is agreement with the above-discussion. Surprisingly, these authors proposed that the 2071 cm−1 was due to atop adsorption on terraces at 273 K. This assignment is unlikely to be correct in view of (i) the negligible adsorption of CO on Au(111) terrace sites at this temperature [40–42], (ii) recent surface science work [41] carried out at 30 K showed that the band of CO on Au(111) terrace was located at 2130 cm−1 and (iii) the simultaneous unexplained disappearance of the 2117 cm−1 steps. An alternative interpretation of the data of Nakamura et al. was that the Au(311) surface gradually restructured to a kink-rich surface (band at 2071 cm−1), leading to a disappearance of the regular step edges and the corresponding band at 2117 cm−1.In conclusions, Au terrace sites lead to bands at around 2130 cm−1, but are not occupied above 70 K [41]. At room temperature, non-restructured edges and corners of nanoparticles lead to bands in the range 2115−2095 cm−1] (Fig. 2.B) [22,28,33,42]. Yet, Au° surfaces readily evolve (within minutes) under CO exposure, already at low temperatures. The band in the region 2080−2060 cm−1 can be assigned to CO adsorbed on Au° at low coordination sites such as kinks and roughened edges [39] (Fig. 2.B). Bands in this region can also potentially be assigned to atop CO on negatively charged Auδ- species, whenever Au is supported on a reducible oxide [24,28,32]. These conclusions stress the challenges existing in band assignment on Au-based materials, which will of course be even greater when bimetallics are considered.Au-Pt alloys have been investigated as catalysts for naphta reforming [44] and electrocatalysts [45,46]. The characterization of Pt-Au bimetallic nanoparticles (size ca. 4 nm) supported on silica by IR of CO adsorption at room temperature has been reported by Mott et al. (Fig. 3 ) [47]. The monometallic Au/SiO2 exhibited a single band located at 2115 cm−1, typical of CO adsorbed on edges of Au° nanoparticles [22,42]. No other band was observed, particularly between 2080−2060 cm−1 region (delimited by the red dotted box), indicating that Au° reconstruction did not occur over the duration of the experiment for this sample. The monometallic Pt/SiO2exhibited a single band located at 2096 cm−1, typical of CO adsorbed on Pt° nanoparticles [18,19].The bimetallic formulations exhibited markedly different IR spectra, particularly in the 96−65 Au at.% range (Fig. 3, curve b–e). New large bands between 2080 and 2050 cm−1 were observed. The authors appeared to assign this band only to atop CO on Pt in an alloyed phase. This is partly sensible, since diluted Pt should lead to a lower CO wavenumber as compared to plain Pt because of reduced dipole coupling [48]. An electronic transfer from Au to Pt could also explain this shift, as proposed by Thomas and co-workers [49].However, Au has a significantly lower surface tension than Pt (i.e. 1333 and 2203 mJ m−2, respectively [50]) and should therefore preferentially occupy surface sites with low coordination numbers such as corners, equivalent to kinks on a stepped surface. Such sites were shown to exhibit bands in the 2080−2060 cm-1 range in the previous section. Surface reconstruction of Au-rich nanoparticles could potentially occur at a faster rate as compared to the case of plain Au. Therefore, it is possible that part of the signal in the 2080−2060 cm-1 region could be assigned to Au−CO species.A signal of Au−CO in the region 2080−2060 cm−1 would also explain the surprisingly low signal observed above 2100 cm-1 (typical of monometallic Au edges, Fig. 3, spectrum a) for the samples still exhibiting high Au contents (Fig. 3, spectra d–h). The lack of Au at the surface could yet be explained by surface segregation, CO preferentially pulling Pt towards the surface because of the stronger bond Pt−CO strength49, though this possibility was not discussed by the authors. Surface segregation effects induced by CO adsorption can possibly be observed with time resolved analysis (vide infra in the case of Au-Pd and Pt-Sn) and were reported by Thomas and co-workers on Pt-Au materials [49].The desorption of CO from Au-Pt bimetallics supported on silica [51] and zirconia [52] has been investigated. A fast desorption of the bands associated with Au (above 2100 cm−1) could be observed at room temperature, while those associated with Pt (below 2100 cm−1) were stable. None of these groups could yet provide a clear evidence of the presence of alloyed nanoparticles, which would probably exhibit an intermediate behavior, since Au could serve as a porthole for CO desorption after surface diffusing from Pt sites.In conclusion, the assignment of band in the region 2080−2050 cm−1 solely to Pt−CO species is not a priori warranted in the case of Au-Pt bimetallics and more attention should be given to surface segregation dynamics induced by the addition/removal of CO. To our view, no conclusive evidence for the formation of an alloyed phase has yet emerged solely on the basis of IR analyses of CO adsorption on Au-Pt bimetallics.Au–Pd bimetallic systems are more active and selective than monometallic Pd catalysts in the direct synthesis of H2O2 and have therefore raised of lot of interest [53]. Guesmi and co-workers reported DRIFTS of CO adsorption on alumina-supported Au, Pd and Au-Pd nanoparticles between 2 and 3 nm in size (Fig. 4 .A) [54]. In contrast to the case of Pt, Pd usually lead to a large signal of carbonyl bands below 2000 cm−1 associated with bridged and multi-bonded CO (Fig. 4.A, spectrum b) [55–57]. These bands are crucial to ascertain the presence of Pd at the surface of alloy particles in the case of Au-Pd bimetallics, like those at 1950 and 1927 cm−1 in spectrum c, since the other bands around 2077 cm−1 could in theory also be assigned to Au−CO species.Interestingly, the spectrum of the bimetallic sample evolved with time under CO (Fig. 4.B). The intensity of the Au−CO band at 2109 cm−1 declined, while those at 2077, 1950 and 1927 cm−1 increased in concert. This observation indicates that the latter bands were associated in most part to Pd and that the surface was getting richer in Pd with time under CO. DFT calculations have quantified the various energy gains of various nanoalloy configurations associated with pulling Pd to the surface due to the stronger bonding between CO and Pd in comparison to that of Au and CO [58].A similar surface segregation of Pd was observed over alumina-supported Au-Pd particles (size between 2 and 6 nm) by combined X-ray absorption and DRIFTS spectroscopies during CO oxidation [59]. The authors proposed that no Au remained at the nanoparticle surface after cooling the catalyst back to room temperature, based on the absence of the 2110 cm−1 band typical of Au−CO. The latter statement should yet be taken with caution, as the previous section showed that restructured Au can lead to bands in the 2080−2060 cm−1 region (actually similar to that of Pd) instead of that around 2110 cm−1. In addition, Au atoms on terrace sites would not adsorb CO at room temperature [41].Titania-supported Au-Pd and Au-Rh nanoparticles were investigated by Piccolo and co-Workers [60]. Again, the authors suggested the absence of Au at the surface of a reduced Au-Rh sample based on the absence of a band at 2100 cm−1. This somewhat surprising conclusion was explained by a CO-induced surface segregation, which overcame the effect of the higher surface tension of Rh (2325 mJ m-2) [50] in comparison to that of Au (1333 mJ m-2). The same caution as noted above applies here, since restructured Au does not lead to Au−CO bands above 2080 cm−1 and thus the absence of bands above 2080 cm−1 cannot be taken as a proof of the absence of Au at the surface. The same observation can be made again for the DRIFTS data reported on SiO2-supported Au-Pd by Mou and co-workers [61].In conclusion, the assignment of band in the region 2080−2050 cm−1 solely to Pd−CO species is not a priori warranted in the case of Au-Pd bimetallics, similarly to the case of Au-Pt sample. Yet, surface segregation dynamics induced by the addition/removal of CO and the presence of strong bands below 2000 cm−1 on Pd can provide clear and direct evidence of the presence of alloyed phases, as convincingly reported by Guesmi and co-workers [54].Some of us recently reported an operando DRIFTS investigation of CO oxidation over Au-Ag embedded on silicalite-1 (a zeotype with MFI structure) [27]. The formation of an alloyed phase was conclusively demonstrated from (i) the strongly shifted bands of Agδ +−CO in the case of the alloy (ca. 2160 cm−1) as compared to the case of the plain Ag sample (2189 + 2172 cm−1) and (ii) the reduction by half of the Auδ +−CO signal at 2134 cm−1, while Au loading and particle size remained the same, due to the preferential presence of Ag on low coordination surface sites (Fig. 5 .A). Ag is in fact one of the few catalytic metals that presents a surface tension lower than that of Au (1086 mJ m−2 for Ag against 1333 mJ m−2 for Au) [50] and is thus expected to be preferentially segregated at the surface of the nanoparticles on low coordination sites such as edges and corners (Fig. 5.B).This section discusses bimetallics based on Sn and Zn. These two metals commonly form intermetallic compounds (i.e. ordered alloys) with catalytic metals and present the particular feature of not adsorbing CO under standard conditions [62]. The latter aspect facilitates markedly spectral analysis, since no direct contribution is expected from Sn and Zn.CO adsorption on Pd was noted above (Fig. 4) as usually exhibiting strong bands below 2000 cm−1 due to the presence of bridged and multi-bonded carbonyls. Another example of spectrum obtained over an alumina-supported Pd sample is shown in Fig. 6 .A, in which linearly adsorbed (atop) CO is located at ca. 2090 cm−1 and two bridged species are present at 1987 and 1954 cm−1. A bimetallic Pd-Sn sample prepared with the same wt.% of Sn and Pd exhibited only the linear species at ca. 2075 cm-1 (Fig. 6.B). This observation indicates that there was no Pd-Pd pairs available at the surface of the bimetallic sample and therefore the IR analysis proved the formation of an intermetallic compound, at least superficially.The IR-based technique (so-called Adsorption Equilibrium IR, AEIR) was used to determine the heat of adsorption at low coverages of the various CO species observed. The linear CO displayed an adsorption heat of around 90 kJ mol−1 both in the Pd and Pd-Sn samples. These values are much lower than that of bridged species measured over the Pd sample, which was higher than 165 kJ mol−1. Overall, the modifications induced by Sn were found to be essentially geometric and not electronic.Pd-Zn bimetallics are highly active and selective for the selective hydrogenation of acetylene into ethylene [57,63]. The Pd-Zn bimetallics discussed here were prepared by reducing Pd and Zn nitrates impregnated over CeO2[57]. The effect of Zn over Pd in terms of CO adsorption was found to be rather similar to that of Sn described in the previous section. The bridged species (and hence Pd-Pd pairs) were completely eliminated in the case of Zn-containing formulations as compared to the corresponding Zn-free material (Fig. 7 .A). The heat of adsorption of the linear CO at low coverage determined by the AEIR method was ca. 115 kJ mol−1 over Pd-Zn/CeO2 (Fig. 7.B), somewhat higher than that measured over the Pd-Sn sample (90 kJ mol−1), suggesting a more significant electronic effect of Zn, besides the obvious geometric effect.It must be noted that the absence of Pd-Pd pairs was not sufficient to achieve the highest selectivity. The samples reduced at 673 K and 773 K both showed no evidence of bridged CO, but the former was significantly less selective than the latter [57]. The linear CO signal was actually different, with maxima at 2054 and 2039 cm−1 in the case of the lowest and highest reduction temperatures, respectively (Fig. 7.A). This shift and difference in band shape probably corresponded to differences in the nature of the terminal crystallographic planes exposed by the Pd-Zn nanoparticles, which may be of crucial importance to selectivity.Cobalt-based catalysts typically produce linear alkanes and short alkenes during CO hydrogenation with H2 (Fischer-Tropsch synthesis) [64–66]. The IR spectra obtained under reaction conditions are quite complex (Fig. 8 .A). The CO(ads) IR signal between 2100 and 1950 cm−1 can be decomposed into several bands that can be assigned to various linear CO, mostly relating to (111) and (100) planes (Fig. 8.A and B) [64,67–72]. A broader band is also observed at ca. 1850 cm−1, typical of multi-bonded CO (noted Hollow−CO). The latter band appeared to correlate with the sample activity during an investigation of catalyst resistance to chlorine [70].Co and Sn form various intermetallic compounds (Co3Sn2, CoSn, CoSn2), but in view of the lower surface tension of Sn (675 mJ m−2) as compared to that of cobalt (2550 mJ m−2), this element is expected to segregate at the surface at low loadings. Small proportions of Sn was added to alumina-supported cobalt (molar ratios Sn:Co = 1:120 up to 1:30) and appeared to selectively titrate the Hollow−CO sites associated with multi-bonded CO (Fig. 9 .A) [11].Methane and propene were the main reaction products under these reaction conditions (atmospheric pressure). Interestingly, the rates of methane and propene formation appeared to be proportional to the fraction of Hollow−CO (Fig. 9.B). This observation suggested that these Hollow−CO were potential reaction intermediates in the formation of hydrocarbons.A full quantitative analysis of adsorbate concentration was carried out to estimate specific decomposition rates of the Hollow−CO and relate it to product formation rates. The specific rates of decomposition of these Hollow−CO appeared to be about twice as high as the rates of production of methane, the main product formed (Fig. 10 ). These Hollow−CO were thus proposed as being part of the main reaction pathway into hydrocarbon formation, taking into account that methane selectivity is usually around 40–50% under these conditions [11]. In contrast, the linear CO(ads) species were shown to be unimportant for the reaction.Catalysts based on bimetallic Pt-Sn have been investigated for many reactions such as the selective hydrogenation of unsaturated compounds [74,75], the dehydrogenation of light alkanes to alkenes [76,77],and the low temperature oxidation of CO without [78] or with an excess of H2 (preferential oxidation of CO, PROX), [7,79].Pt and Sn are fully miscible and form numerous intermetallic compounds. Sn does not bind CO under standard conditions. In a recent combined DRIFTS - DFT study [80], a gradual decay of the linear Pt−CO wavenumber from plain Pt (2075 cm−1), through a Sn-poor Pt-Sn (2054 cm−1) and to a Sn-rich Pt-Sn (2039 cm−1) were reported. The heat of CO adsorption also decreased in the same order [80,81], explaining the improved resistance of Pt-Sn electrodes against CO poisoning as compared to plain Pt [82]. The heat of adsorption of CO over plain Pt [81]was about 185 kJ mol−1, while that over a Sn-rich alloy [80] was only 85 kJ mol−1.The difference of affinity for CO between Sn and Pt explained the gradual increase of the Pt−CO band around 2030 cm−1 corresponding to the gradual Pt-enrichment at the surface of the nanoparticles in a Sn-rich sample following the introduction of CO (Fig. 11 .A) [80]. DFT calculations confirmed the relevance of this surface segregation, as the segregation energy for Pt in a Sn(100) surface was dramatically lowered in the presence of CO (Fig. 11.B).A different type of segregation was observed in the presence of CO at lower temperature. A phase segregation occurred in which metallic Pt-Sn nanoparticles were converted into Pt + SnOx following CO dissociation83. This phase segregation was apparent through the gradual shift of the carbonyl band from 2046 cm−1 (corresponding to a Pt-Sn alloy) to higher wavenumber towards plain Pt phases (> 2060 cm−1) following H2 removal (Fig. 12 .A). This phenomenon could be reversed upon heating the sample in a CO + H2 feed up to 300 °C (Fig. 12.B). The presence of both Pt and Pt-Sn phases was obvious as the Pt−CO and (Pt-Sn)−CO bands could be resolved during the conversion of the Pt phase into Pt-Sn.Somorjai and co-workers have long ago evidenced CO dissociation and the related carbon deposition on Pt steps at room temperature [84] and also on Pt(100) at 498 K [85]. Thus, observing CO dissociation over Pt-Sn nanoparticles is not totally unexpected. The formation of graphitic carbon deposits derived from CO dissociation over the Pt-Sn nanoalloys was confirmed by an in situ XPS analysis [83]. One driving force of the process was the formation of a stable SnOx phase. Thermodynamic calculations (realised on bulk phases) showed that the reaction between CO and Sn to give graphitic carbon + SnO was as favourable as the well-known Boudouard reaction (i.e. 2 CO ⇒ C + CO2) (Fig. 13 ). The formation of graphitic carbon + SnO2 was even more favourable.In the case of CO oxidation with O2, Sn oxidation and the formation of metallic Pt intimately mixed with SnOx domains has been observed whether H2 is present [7] or not [78]. The segregation of Pt from the Pt-Sn alloy was clearly observed by operando DRIFTS at 225 °C (Fig. 14 .A) [7]. The alloy was present before O2 admission, exhibiting a band at 2050 cm−1. The band shifted to higher wavenumbers typical of plain Pt (2067 cm−1) as soon as O2 was introduced. The higher activity over the Sn-containing catalysts was therefore not due to the presence of an alloyed Pt-Sn phase, but instead to the intimate dispersion of Pt into a SnOx matrix that participated into the oxidation of CO by supplying O to CO adsorbed on Pt (Fig. 14.B). This was reflected in a lower reaction order in O2 (i.e. + 0.2) over the Pt-SnOx sample, as compared to an order of +1 for the plain Pt.An ideal IR probe molecule should not react or degrade the sample under investigation during the IR analysis. Ni forms highly volatile Ni(CO)4 at temperatures as low as 85 K and the erosion of Ni-based catalysts under CO has been documented [86]. The effect of Ni(CO)4, formed in situ or over equipment not suited for use with CO, is often unaccounted for and can lead to gross data misterpretation [14]. Thus, Ni-based catalysts should not be investigated with CO below 200 °C, to prevent the formation of Ni(CO)4 and the resulting erosion and dispersion of Ni in the apparatus. CO reactivity at low temperatures has also been reported over cobalt-based samples with the formation of surface carbides leading to bands in the 2060−2050 cm−1 region, instead of 2030−1990 cm−1 measured over metallic surface [68,69]. Cobalt is highly active for CO dissociation and is thus commonly used for CO hydrogenation in Fischer-Tropsch reactions.CO dissociation at room temperature has even been reported on elements less known for their CO hydrogenation abilities, e.g. Pt steps [84] and, even more surprisingly, on Au(110) surfaces [38]. Marie and co-workers [87] also reported that CO was able to oxidize Ag nanoparticles at room temperature, with the concomitant formation of elemental carbon through Boudouard reaction. These observations stress that the occurrence of CO reactivity as observed on our Pt-Sn (Fig. 12) [81,83] is more common than usually thought and probably was overlooked in many instances.A strategy to limit or avoid carbon deposition is to carry out the adsorption of CO in the presence of H2, though the level of CO conversion (e.g. to CH4) should thus be checked as to not modify significantly the partial pressure of CO used [81,57]. The same “cleaning” effect of H2 was observed in the case of the Pt-Sn materials, where hydrogen helped removing adsorbed O that would otherwise oxidize Sn and lead to phase segregation (Fig. 12) [83].The use of IR spectroscopy of CO adsorption to investigate the surface of bimetallics or alloys is challenging and not conclusive in many instances. Firstly, this results from the complexity of the spectra exhibited by individual metals that depends on many structural parameters. Secondly, CO adsorption often results in structural modification of the surface composition of nanoalloys via reconstruction (as in the case of monometallic surfaces) or surface segregation when one metal displays a higher affinity for CO. The dynamics of these modifications may actually be used to ascertain the presence (or absence) of alloyed surfaces.The addition of catalytically inactive elements such as Sn and Zn can be used to improve selectivity (via a dilution geometrical effect) of otherwise highly active but unselective metals, as in the case of Pd used for acetylene hydrogenation. Conveniently, these inactive elements (e.g. Sn and Zn) do not chemisorb CO under standards conditions and can be used to poison selectively surface sites and facilitate the understanding of surface structure and reactivity. Sn added to cobalt nanoparticles was shown to specifically poison the surface active sites associated with the hydrogenation of multi-bonded CO into hydrocarbons.CO was also shown to dissociate on Pt-Sn surfaces, albeit at a slow rate over timescale of minutes around room temperature. CO dissociation has also been reported for Pt, Ag and Au-based surfaces under similar conditions and should be a reminder that CO is not always inert under these conditions, even on metals not efficient in CO methanation-like reactions.All the authors participated to the writing of the review and approved its final version.The authors declare no competing interests.R.A. acknowledges a PhD scholarship from the Ministry of higher Education and Research of France at the University of Lyon. T.E. acknowledges a PhD scholarship from the ANR, project DECOMPNOx (ANR-18-CE07-0002-01).

This contribution reviews some of the structural features of supported bimetallic catalysts that could be unravelled using IR spectroscopy of CO adsorption. The few examples presented are focussed on active metals such as Au, Pd, Pt and Co and modifications by inert elements such as Sn or Zn. The difficulty in interpreting the IR spectra during CO adsorption over Au-Pt and Au-Pd is underlined, because of Au restructuring that leads to bands typically in the range associated with Pt and Pd carbonyls. Crucial aspects of metal dispersion and surface segregation can yet be obtained, such as active metal site isolation by the disappearance of bridged or multi-bonded CO. The use of IR spectroscopy of CO adsorption to investigate the surface of bimetallics or alloys remains yet challenging and not conclusive in many instances. Firstly, this results from the complexity of the spectra obtained (even over single metals) that depends on many parameters. Secondly, CO adsorption often results in modifications of both surface structure and composition of nanoalloys via reconstruction, as in the case of monometallic surfaces, or surface segregation when one of the metals exhibits a greater affinity for CO. CO dissociation near room temperature has also been documented on many metals, including cobalt and more surprisingly Pt, Ag and Au-based materials.

In so many aspects, the irruption of Metal-Organic Framework (MOF) materials revolutionized the applications of the porous materials [1–3]. A good example could be found with Fe-based MOFs, taking advantage of some favorable properties of iron amongst the possible metals forming MOFs, such as its low price, high abundance, low toxicity, redox behavior, etc. As a consequence, Fe-carboxylate MOFs are nowadays very promising materials in applications as diverse as biomedicine [4], redox and/or acid catalysis [5–11], supports of enzymes [12–15] or photocatalysis [13,16], amongst others.On the other hand, like other trivalent metals such as Cr [17,18], Sc [19,20] or V [21,22], iron is able to form the pair of closely related MOFs MIL-100(Fe) [23] (iron trimesate) and MIL-101(Fe) [24] (iron terephthalate) which were (in their chromium form) the first reported MOFs having mesocavities. Moreover, unlike the above mentioned trivalent metals, iron can also form the semicrystalline Fe-BTC material, which is structurally close to MIL-100(Fe) [7,25]. In spite of its semiamorphous nature, it sometimes surpasses the catalytic activity of MIL-100(Fe), especially when Lewis acid sites are demanded by the reaction [26]. The interest in Fe-BTC material is such that it was one of the first MOFs to be commercialized, specifically as so-called Basolite F300 by BASF/Aldrich. As a consequence, it was used in different applications [27,28], particularly in catalysis [5,26], before its indirect [29] or direct [7,30] syntheses were reported. Since then, it continues being applied in some other different fields [31–45], again with special emphasis on heterogeneous catalysis [12,13,16,46–52], although its structure continues being unknown in spite of an important recent progress [53].The relationship between MIL-100(Fe) and Fe-BTC is very singular in so many aspects. Firstly, although it has been shown that both materials possess many similarities (composition [25], thermal stability [54], nature/structure of metal clusters [55], a common mesocavity [7], the possibility of being prepared at room temperature and in water [7,54] or similar catalytic applications [26]), they do not form a nanocrystalline-microcrystalline pair. They are rather a semicrystalline-full crystalline pair of materials. To the best of our knowledge such relationship is unique in the huge family of MOFs. In this sense, it is quite remarkable the difference in their preparation at room temperature: starting from Fe(III) source, Fe-BTC is formed instantaneously [7], whereas if Fe(II) source is used in otherwise equal synthesis, crystalline MIL-100(Fe) is formed more slowly, in parallel with the oxidation of Fe(II) to the more stable Fe(III) in aqueous solution [54]. Secondly, as it was mentioned above, the semiamorphous Me-BTC has been only described when the metal Me is Fe but not for other trivalent counterparts like Cr, Al, V or Sc, which suggests that Fe possesses singular trend to form this kind of semiamorphous MOFs.Furthermore, apart from Fe-BTC, MIL-100(Fe) has another highly related MOF material: MIL-101(Fe). Although both MIL materials are prepared with different linkers (trimesate and terephthalate, respectively), they have: (i) the same metal clusters [Fe3O(X)(H2O)2]6+ (X = F- or OH-); (ii) the same topology MTN [56]; (iii) the possibility of being easily extended or functionalized [56], and (iii) two different but highly-related mesocavities (with diameters of 25 and 29 Å, and 29 and 34 Å, respectively), all having microporous entrances [17,18]. Moreover, MIL-101(Fe) has in principle two key advantages over MIL-100(Fe): the lower price of its organic linker and its much higher textural properties (surface area, pore volume, diameter of its cavities and the entrance to these cavities, etc.). With this background, one might wonder if MIL-101(Fe) could also have a semiamorphous counterpart yet to be discovered: Fe-BDC, the missing link amongst the Fe3O-clustered carboxylate-based MOFs.This work describes systematic approaches to afford the semiamorphous Fe-BDC at room temperature. Although it was not achieved in water as solvent, the use of iron(II) acetate as Fe source in either N,N-dimethylformamide or ethanol as solvents led to a relatively porous semiamorphous Fe-BDC material, with evident similarities to both the crystalline Fe-BDC MOF MIL-101(Fe) and the semiamorphous Fe-BTC. The catalytic performance of this Fe-BDC was tested in the aerobic oxidation of cyclohexene. It at least equals the catalytic activity of the commercial Fe-BTC Basolite F300.A typical synthesis of the samples FeBDC starts with the preparation of two solutions/suspensions. Solution 1 is formed by dissolving 1.5 mmol of iron(II) acetate (Fe(OAc)2, supplied by Sigma-Aldrich) in 5 mL of one of the following solvents: DMF, ethanol (EtOH), methanol (MeOH) or H2O. Similarly, solution/suspension 2 is prepared by adding 1.5 mmol of linker H2BDC in 5 mL of the same solvent; actually it only becomes a solution when the linker is added to DMF, as such amount of linker is not soluble in any of the other three solvents. Next, solution 1 was added dropwise (for 5 min) over the solution/suspension 2 under stirring, what provokes the immediate appearance of a brownish orange precipitate. The resultant mixture was kept at room temperature for 18 h under stirring, and then the suspension was filtered, and the solid were washed 2 times with the synthesis solvent and 3 more times with ethanol. Then, the material was immersed in ethanol for 6 days changing the washing solvent by fresh ethanol every 2 days. Finally, the samples were dried at 80 ºC for 2 h, and were labeled as Fe-BDC-DMF, Fe-BDC-EtOH, Fe-BDC-MeOH and Fe-BDC-H2O according to the solvent used in their synthesis.The same methodology was used for the preparation of the samples denoted as Fe-BDC-DMF-75 and Fe-BDC-DMF-100, except the temperature at which the mixture was treated during the 18 h after the addition of solution 1 over solution/suspension 2, which was 75 and 100 ºC, respectively, instead of the room temperature (and stirring) used for the preparation of the sample Fe-BDC-DMF.For comparative purposes, a commercial Fe-BTC (Basolite F300) was purchased from Sigma-Aldrich.X-ray diffraction (XRD) patterns were collected with a Philips X’PERT diffractometer having a X’Celerator detector and using Cu Kα radiation. Nitrogen adsorption/desorption isotherms were measured at − 196 °C in a Micromeritics ASAP 2420 device; the samples were previously degassed at 150 °C under high vacuum for at least 16 h; Specific surface areas were estimated by BET method, external/micropore surface area by t-plot method, and pore size distribution (PSD) by BJH method. Thermogravimetric analysis (TGA) was performed using a Perkin-Elmer TGA7 instrument, with a heating rate of 20 °C/min under air flow. Scanning electron microscopy (SEM) studies were carried out in an ultrahigh resolution FEI-NOVA NanoSEM 230 FESEM instrument. Attenuated total reflection-Fourier transform infrared (ATR-FTIR) spectra of samples were recorded using a Thermo Nicolet Nexus 670 FTIR spectrometer equipped with a SensIR Technologies DurasamplIR horizontal ATR accessory and a liquid nitrogen-cooled MCT detector. Fe content of the samples was analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES) in a PlasmaQuant PQ 9000 after the samples were dissolved by acidic treatment in a microwave-assisted oven.In a typical catalytic reaction, 50 mg of the catalyst were activated at 100 °C overnight under N2 flow (20 mLmin-1) within a round bottom flash of 25 mL. 8 mL (80 mmol) of cyclohexene and 0.85 mL (5.23 mmol) of octane (chromatographic internal standard), were added under a flow of O2 (20 mLmin-1). The flask containing the reaction mixture was submerged within a silicon bath and connected to a refrigerant where water at 5 ºC (to avoid evaporation of the volatiles reactants and products) recirculated. Once the reaction temperature is reached (60 ºC inside the flask), oxygen was slowly bubbled into the reaction mixture. Aliquots of 0.1 mL diluted in 1 mL of acetonitrile were taken for quantification in GC Varian 430 with a 15 m x 0.25 mm diameter column and a FID detector.As explained in the introduction, the semiamorphous Fe-BTC and the crystalline MIL-100(Fe) possess both structural similarities and structural differences. Such similarities/differences are manifested in the XRD characterization ( Fig. 1A) in such a way that the broad XRD reflections characteristic of Fe-BTC are reproduced in corresponding groups of sharp XRD peaks in the pattern of the MIL-100(Fe) [7,25] (pointed by arrows in Fig. 1A). The XRD patterns of the samples Fe-BDC prepared at room temperature in different solvents are plotted in Fig. 1B. Just like it occurs in Fig. 1A, the broad XRD reflections of the pattern of the samples Fe-BDC prepared in either ethanol or DMF media, match very well with different set of reflections found in the simulated pattern of MIL-101(Cr). (The simulated pattern of MIL-101(Fe) is not plotted instead of that of MIL-101(Cr) because, to the best of our knowledge, the cif file of this material is not available in the literature and/or crystallographic databases. Nevertheless, the patterns of MIL-101(Fe) and MIL-101(Cr) should be practically equal, just like the patterns of the also isostructural materials MIL-100(Fe) and MIL-100(Cr) are [23,57]). It strongly suggests that these two samples are related to MIL-101(Fe) in a way very similar to how MIL-100(Fe) and the semi amophous Fe-BTC are related. In other words, it seems that the Fe-BDC and MIL-101(Fe) materials would form a semiamorphous / crystallized pair of MOFs, just like Fe-BTC and MIL-100(Fe) materials do.On the other hand, the number of arrows indicating the structural similarities between XRD patterns of Fe-BDC and MIL-101(Cr) (Fig. 1B) is lower than that indicating the same for Fe-BTC and MIL-100(Fe) (Fig. 1A). It is particularly remarkable the absence of any reflection bands in the 2θ region of 4–6.5º in the XRD patterns of both samples Fe-BDC (that is normally associated to relatively large microporosity within the material), whereas the XRD pattern of the MIL-101 materials has a group of quite intense bands precisely in such 2θ region ( Figs. 1B and 2). The equivalent reflections in the simulated XRD pattern of the material MIL-100(Fe) (in this case slightly shifted towards the 2θ range of 5.5–8º) found their corresponding XRD band in the XRD pattern of Fe-BTC (Fig. 1A). All this suggests that Fe-BTC somehow have higher construction degree than the Fe-BDC prepared at room temperature, that is, Fe-BTC is a bit closer of becoming a MIL-100 material than the here-presented Fe-BDC of becoming a MIL-101 material.However, the XRD patterns of the samples Fe-BDC prepared in either water or methanol at room temperature are clearly different to these of the samples Fe-BDC prepared in ethanol or DMF (Fig. 1B). The former are dominated by three intense and sharp reflections at ca. 17.5, 25.4 and 28.1º (marked by asterisks in Fig. 1B) which are typical of the XRD pattern of the unreacted protonated linker H2BDC [58]. Some other relatively sharp XRD reflections, which does not match with the simulated pattern of MIL-101, were detected at 2θ below 13º (and then potentially related to the existence of certain microporous features) in the diffractogram of the sample prepared in methanol but not in the one prepared in water. It must be noted that the solubility of terephthalic acid, which is essential to get a BDC-based MOF under the studied conditions, increases as a function of the solvent nature in the following order: water (practically insoluble) < methanol < ethanol < DMF (very soluble). Obviously, unlike the samples Fe-BDC-DMF and Fe-BDC-EtOH, the samples Fe-BDC-MeOH and Fe-BDC-H2O cannot be considered a semiamorphous Fe-BDC.Trying to certify that our samples Fe-BDC-DMF and Fe-BDC-EtOH prepared at room temperature really possess strong structural likenesses with MIL-101(Fe), the Fe-BDC synthesis mixture prepared in DMF was heated up different temperatures for 18 h, since, to the best of our knowledge, the synthesis of MIL-101(Fe) has not been described at room temperature and then heating could be an essential stimulus to the formation of this crystalline phase. The low angle XRD patterns of the corresponding samples compared to the simulated pattern of MIL-101(Cr), are shown in Fig. 2 whereas the high-angle XRDs are plotted in Fig. S1. The low-angle diffractogram of the sample Fe-BDC-DMF prepared at room temperature does not practically show any indication of reflection at 2θ below ca. 7° (Fig. 2), suggesting that this sample lacks spatially ordered mesocavities. Contrasting with this fact, Fe-BTC possesses the smallest ordered mesocavity of the two found in MIL-100(Fe) [7]. As the synthesis temperature rises, some low angle reflections start to appear. These reflections, which are very broad after a thermal treatment at 75 ºC for 18 h, become unequivocally demarcated when the sample results from treating the synthesis mixture at 100 ºC. These peaks match very well with the XRD pattern of the MIL-101(Cr) at low angle. It indicates that the structural units responsible for the reflections in the XRD pattern of the sample Fe-BDC-DMF prepared at room temperature are capable of recombining to give rise to MIL-101(Fe)-like material with the unique stimulus of temperature. Therefore, one would expect that at least some of the structural building units forming the MIL-101(Fe) structure are already present in the sample Fe-BDC-DMF. It is important to remark that the transformation of such structural units into MIL-101(Fe) is not a crystal growth phenomenon (which would transform a nanocrystalline material into a microcrystalline one) but the construction of an ordered structure from its already existing ‘bricks’. The absence of low angle XRD reflections in the pattern of the room-temperature prepared sample Fe-BDC-DMF as well as the generation of new reflections (rather than a simple narrowing of the existing ones) as a consequence of the increase of synthesis temperature, support this hypothesis.Unlike the long-range information given by XRD, IR spectroscopy region of ca. 650−1800 cm−1 ( Fig. 3) is rather sensitive to the conformational and/or local environment of organic molecules (short-range information). That IR region can be taken as a fingerprint of the materials nature, especially when they contain organic entities like MOFs [59,60]. Therefore, this technique can somehow complement the limited structural long-range information provided by XRD on the semiamorphous materials. It is remarkable the substantial difference of this spectra in comparison with that of the protonated linker H2BDC (Fig. 3), in spite of the FTIR features of this spectra region are due to almost the same chemical species: terephthalate linked to either iron metal ions or protons. The spectra of the four sample Fe-BDC are practically identical. Only the broad bands at 1100 and 1651 cm-1 in the spectra of all three samples Fe-BDC-DMF, attributed to CO bond strength of DMF, differentiate the FTIR spectra of the samples Fe-BDC. The most intense bands were assigned as follows: (i) 746 cm-1, out-of-plane bending vibration of C–H in aromatic rings (750 cm-1 [61] or 749 cm-1 [62] reported for MIL-101(Fe)); (ii) 1381–1384 cm-1, symmetric stretching of carboxyl groups of the aromatic carbon C–C vibrational mode in terephthalate (~ 1400 cm-1 [61] and 1394 cm-1 [63] reported for MIL-101(Fe); (iii) 1504 cm-1 asymmetric stretching of CO bonding carboxyl groups in terephthalate (~ 1500 cm-1 reported for MIL-101(Fe) [61]); (iv) 1567–1571 cm-1, CO bonding in the carboxylates (1584 cm-1 reported for MIL-101(Fe) [63]). (It must be noted that the FTIR-ATR spectra presented in this study has not been corrected, so the frequency of the bands should be shifted to a few higher wavenumbers [64] to make them comparable to reported transmission/absorbance FTIR values, which would actually lead to a better agreement between the frequencies found in this work and these of the reported ones). In summary, all the main bands found in the FTIR spectra of the semiamorphous samples Fe-BDC were also found in the FTIR spectra of MIL-101(Fe). It suggests that terephthalate anions within Fe-BDC materials adopt practically equal conformation/environment to that given within MIL-101(Fe). Fig. 4 plots the TGA profiles of the samples Fe-BDC prepared in different solvents and at different temperatures. The temperature at which linker decomposes (event normally related to the thermal stability of MOFs), is practically the same in all Fe-BDC samples, at ca. 360 ºC, under the studied analysis conditions. The weight loss of linker is taken place in one relatively quick step. Terephthalate in MIL-101 (Fe) has been reported to begin to decompose at 350 °C [65] (in two steps) and at 345 ºC [57] (in practically one step). Although the analysis conditions could slightly alter the temperature of the weight losses, the similar values in linker decomposition suggest that terephthalate is in very similar coordination to iron metal clusters in MIL-101(Fe) and in our semiamorphous Fe-BDC samples. Moreover, the linker decomposition temperature is slightly higher than the linker decomposition temperature found in Fe-BTC materials (341–343 ºC either commercial one or prepared in the laboratory) under the same TGA conditions [7]. It has to be noted that both materials, Fe-BDC and Fe-BTC, are based on different organic linkers, so differences in thermal stability of such order were expected.The textural properties of the different Fe-BDC samples prepared in DMF were evaluated by means of N2 sorption isotherms at − 196 ºC ( Fig. 5 and S2 and Table 1). All the samples Fe-BDC are quite porous, having BET specific surface areas above 500 m2g-1, in particular, 536, 675 and 926 m2g-1 for the samples prepared at room temperature, at 75 ºC and at 100 ºC, respectively. As expected according to the XRD discussion, the samples Fe-BDC prepared at temperatures above room temperature (Fe-BDC-DMF-75 and -100), which possesses some XRD features of the material MIL-101(Fe), widely surpass the textural properties of the sample Fe-BDC-DMF prepared at room temperature. Fig. 5A makes clear that the difference in the textural properties of the three samples is specially given at small relative pressures p/p0 (in the 0.01–0.3 range). It means that the porosity of these samples are mainly differentiated in relatively small mesoporosity, precisely in the isotherm region at which the mesocavities of the MIL-101(Fe) material, which are absent in the sample Fe-BDC-DMF but present in the samples Fe-BDC-DMF-75 and -100, should manifest. This fact is even more evident in the PSD curves presented in Fig. 5B, which does not show any small mesoporosity for the sample Fe-BDC-DMF, whereas the other two samples unequivocally present relatively sharp PSD maxima at ca. 25.1 and 30.7 Å, which are more abundant and well-defined at higher synthesis temperature, in good agreement with the intensity of the low-angle XRD peaks (Fig. 2). We ascribed these PSD maxima to the mesocavities of the MIL-101(Fe) material, having diameters of 29 and 34 Å (for MIL-101(Cr) [17]. The slight deviation of the here-estimated diameter cage (of ca. 3.3–3.9 Å) must be due to the well-known underestimation in pore size of mesoporous by the BJH method [7,54,66].It must be noted that the BET surface areas of these samples Fe-BDC (in the range 536–926 m2g-1) are significant even when compared with reported textural properties of MIL-101(Fe). In principle, one could expect that MIL-101(Fe) would reach similar BET specific surface areas close to that of 4100 m2g-1 reported for MIL-101(Cr). However, the reported surface areas of MIL-101(Fe) are quite lower than this expected value: 101 [67], 125 [68], 560 [69], 1018 [57], 1312 [70] or 1642 [71] m2g-1. According to a recent publication, the reason behind so poor textural properties is the co-crystallization of MOF-235 and MIL-101(Fe) in different proportions as a function of synthesis conditions [72].In addition, the isotherms of the three samples Fe-BDC-DMF show high N2 adsorption at high relative pressure p/p0 (Fig. 5A), which should be ascribed to large interparticles mesopores (above 30 nm of average diameter, Fig. S2), probably as a consequence of the agglomeration/aggregation of the particles formed by an instantaneous and massive precipitation. To shed light on the morphology of the samples, Fig. 6 shows some representative FE-SEM images of the sample Fe-BDC-DMF, which would be the ‘real’ semiamorphous Fe-BDC. As forecasted by the isotherms from Fig. 5A, the sample is indeed formed by large particles that are formed by a huge number of much smaller aggregated nanoparticles, leaving some meso-/macro-porosity in their aggregation. It could explain the high adsorption of N2 at high relative pressure p/p0 (above 0.8). The morphology of this sample resembles that of the Basolite F300 and the lab-made Fe-BTC [7], the latter also obtained by immediate precipitation as soon as iron and linker sources made contact.The catalytic potential of some of these Fe-BDC materials was tested in the solvent-free aerobic oxidation of cyclohexene. Cyclohexene can be oxidized by two different mechanisms, either through epoxidation or allylic oxidation, leading to different products (Scheme S1). Fig. 7 shows the cyclohexene conversion using the following catalyst: (i) the most semiamorphous sample Fe-BDC-DMF (prepared at room temperature), (ii) the sample Fe-BDC structurally closest to MIL-101(Fe) (prepared at 100 ºC), (iii) the commercial Fe-BTC Basolite F300, (iv) a Zn-MOF-74 prepared according to literature [60] as an example of MOF-based catalyst having open metal sites but free of any redox-active metal center, and (v) no catalyst (the blank experiment). The first remarkable aspect from Fig. 7 is that a MOF that lacks redox metals like Zn-MOF-74 is not active in the reaction, even becoming an inhibitor as the reaction goes at some extent in the absence of any catalyst (blank experiment) but not in the presence of Zn-MOF-74. On the contrary, all Fe-MOF-based catalysts gave higher conversion than the blank. Therefore, the semiamorphous Fe-BDC is an active catalyst in oxidation reactions.The catalytic performance of the semiamorphous Fe-BDC prepared at room temperature is below that given by the commercial Fe-BTC Basolite F300 after short reaction times, due to a longer induction time by the former (it does not show any activity until 4 h). It could be related to a limited accessibility of the reactants to the metal centers, as the samples Fe-BDC-DMF does not have any mesocavities (Figs. 1B and 5B), whereas Fe-BTC does[7]. Supporting this interpretation about the catalytic activity delay of Fe-BDC-DMF, the induction time is also shorter when the catalyst is Fe-BDC-DMF-100, which is also a semiamorphous Fe-BDC but having further structural similarities with MIL-101(Fe), including certain amount of mesocavities. It must be noted that, once the induction period has passed, the kinetics of the cyclohexene conversion follows the same slope for both samples Fe-BDC-DMF and Fe-BDC-DMF-100, indicating the same intrinsic catalytic activity of their metal centers. Moreover, at longer reaction times, the catalytic performances of the samples Fe-BDC-DMF equals or even slightly exceeds that of the Basolite F300, which has been previously proved to be a good heterogeneous catalyst in the cyclohexene oxidation [7]. For strict comparison, the concentration of the active Fe centers should be also taken into account. In good agreement with what was expected from the supposed composition, Fe-BTC contains slightly lower iron concentration (23.0 wt% Fe in the dry sample, according to ICP-OES) than Fe-BDC-DMF (26.6 wt% Fe) and Fe-BDC-DMF-100 (24.5 wt%). Fig. 8 separates the yield of products obtained by epoxidation and by radical mechanisms. For all Fe-based MOF catalysts, the products via radical mechanism dominates over the epoxidation products. Nevertheless, the epoxidation proportion is much higher with this series of catalysts than with M-MOF-74 catalysts (with M being a redox metal like Cu, Co, Mn or Ni) under similar conditions [73–75]. Although the reason behind such radical/epoxidation ratio is not clear at this moment, it could be related to the higher accessibility of reactants to the active sites (framework iron) in these MOFs materials, as the epoxidation requires that both substrate (cyclohexene) and oxidant (molecular oxygen) become coordinated to a given active center.On the other hand, the structure of the sample Fe-BDC-DMF is basically maintained after reaction (Fig. S3), whereas Fe-BTC seems to suffer severe degradation under similar reaction conditions.This work presents the discovery of the semiamorphous MOF material Fe-BDC. Just like the widely-used catalysts Fe-BTC and MIL-100(Fe) form an unprecedented semiamorphous/full-crystallized pair of MOFs, the new material Fe-BDC is highly related to MIL-101(Fe) in a similar way. The semiamorphous Fe-BDC can be prepared in ethanol or in N,N-dimethlyformamide at room temperature by simply contacting both iron and terephthalic acid sources, with no extra chemical species (modulator, deprotonating agent, etc.). However, it could not be prepared in water or in methanol probably due to the negligible solubility of the organic linker in these solvents. The semiamorphous Fe-BDC and MIL-101(Fe) possess many key properties in common such as the same metal clusters, almost equal linker environments and conformations and similar thermal stability. However, the semiamorphous Fe-BDC, despite its acceptable microporosity, basically lacks both the crystalline nature and the mesoporosity of MIL-101(Fe). The similarities are accentuated when Fe-BDC is synthesized at higher temperatures (for instance, 100 ºC), which potentially allows to prepare on demand Fe-BDC-based materials having the same building units and different long-range order and porosity. These materials were catalytically tested in the solvent-free aerobic oxidation of cyclohexene, showing catalytic performance at least of the same order than that given by the commercial Fe-BTC catalyst. Such catalytic performance becomes higher when the semiamorphous Fe-BDC is prepared at high temperature.All authors have read and agree to the published version of the manuscript. J. Gabriel Flores Aguilar: Experimental work and design, Results discussion, Figures for the first draft of the manuscript. Rafael Delgado-García: Experimental work and design, Results discussion. Manuel Sánchez-Sánchez: Conceptualization, Resources, Writing – original draft, 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.Authors thanks Dr. Carlos Marquez-Álvarez for his help in registering the FTIR-ATR spectra. This work has been partially financed by CSIC through the programs: (i) CVCSIC-AEPP-Ayudas Extraordinarias para preparación de proyectos 2019 (2019AEP076) and (ii) i-COOP-2018 (COOPA20271). This work has been also financed by the CONACyT project A1–5-30646. J.G.F. acknowledges a Ph.D. CONACyT grant (687839). R.D.G. thanks the JAE-Intro CSIC grant (JAEINT18_EX_0043).Supplementary data associated with this article can be found in the online version at doi:10.1016/j.cattod.2021.11.004. Figure S1 Supplementary material .

Porous Fe carboxylates are amongst the most promising MOF-based materials due to their low price, low toxicity, metal environments (including open metal sites) and remarkable (meso)porosity variety. Fe-MOFs based on the cluster [Fe3O(X)(solvent)2]6+, that is, MIL-101(Fe), MIL-100(Fe) and semiamorphous Fe-BTC, are of particular interest. These three materials are quite related each other: (i) MIL-100(Fe) and MIL-101(Fe) have the same zeolitic topology MTN and two types of mesocavities, whereas (ii) Fe-BTC and MIL-100(Fe) form an unprecedented semiamorphous / fully-crystallized pair, having in common the metal cluster, the composition, one mesocavity, etc but without becoming a nano- / micro-crystalline pair. This work describes the room-temperature synthesis, characterization and catalytic performance in the aerobic cyclohexene oxidation of the semiamorphous Fe-BDC, which together with MIL-101(Fe) would form the second semiamorphous / crystallized pair in MOFs. Unfortunately, Fe-BDC could not be prepared in water as solvent, but in either ethanol or in N,N-dimethylformamide. It possesses relatively high textural properties (above 500 m2g-1) and key common features with MIL-101(Fe): XRD reflections at the same 2θ positions, similar thermal stability, almost equal linker conformations, etc. Fe-BDC became quite active in the solvent-free aerobic oxidation of cyclohexene under mild conditions, surpassing the activity performance of the well-known commercial Fe-BTC catalyst in the same reaction under the same mild conditions.

With the increasing number of motor vehicles, the harmful automotive exhaust becomes a major source of pollution in the urban environment all over the world. During combustion of fuel, in which sulfur-containing compounds are presented with trace amount, SO x is inevitably produced, leading to that 1) air pollution, and 2) irreversible poisoning of the three-way catalysts that are used in the emission after-treatment system. Therefore, more stringent environmental legislations have been introduced worldwide, which are especially crucial for the oil refinery industry. In Euro V and Tier III legislations, sulfur content in gasoline is required to be less than 10 μg g−1 [1]. The National V standard, implemented in China, also requires the sulfur content in gasoline to be lower than 10 μg g−1. Hydrodesulfurization (HDS) is the prevailing approach for sulfur removal in hydrocarbons, where H2 acts as the major reductant [2]. The HDS process using Co-Mo/Al2O3 or Ni-Mo/Al2O3 catalysts [3–6] is highly efficient in removing thiols, sulfides, and disulfides, but less effective for aromatic thiophenes and thiophene derivatives. In order to achieve overall high HDS efficiency, it requires high H2 pressure, simultaneously leading to the saturation of olefins. Such hydrogenation of unsaturated hydrocarbons will reduce the octane number of fluid catalytic cracking (FCC) gasoline.As the catalytic hydrodesulfurization (HDS) process itself is difficult to compromise the requirement of ultra-deep desulfurization and retaining high octane number [7–9], many new approaches, such as catalytic oxidative desulfurization [10], adsorption desulfurization [9], reactive adsorption desulfurization [1,11–13] ionic liquid extraction desulfurization [14,15], biocatalytic treatment [16], have been developed for sulfur removal from hydrocarbons. The reactive adsorption desulfurization (RADS) is one of the promising alternative approaches for deep desulfurization. It combines the advantages of HDS and adsorptive desulfurization, enabling the rapid removal of formed H2S in order to 1) improve the desulfurization efficiency, and 2) lower the H2 partial pressure thus beneficiary for retaining high octane number of resulted hydrocarbons.The S Zorb process of SINOPEC Corp has been proved to be effective for the production of low-sulfur gasoline via RADS by using a solid material [14,16–18]. The process was carried out in a fluidized-bed reactor at 375–450 °C, under 1.0–3.0 MPa H2 pressure. In the S Zorb process, Ni/ZnO is the major body of catalyst, where Ni functions as hydrodesulfurization site, while ZnO traps sulfur and transforms into ZnS [18]. During the process, the S atom is retained on the solid material, while the hydrocarbon with reduced sulfur content is released back into the reaction stream. The unique reaction pathway does not generate gas-phase H2S, and, therefore, prevents the recombination of H2S and olefins, which mainly produces mercaptans. In addition, it shifts the equilibrium of the desulfurization reaction and achieves ultra-deep desulfurization under low H2 pressures, thus limiting the olefin saturation [19,20].Many efforts have been made to unravel the mechanism of RADS. In a model reaction mimicking RADS, where thiophene is used as representative sulfur-containing molecule, the whole process comprises 5 steps, i.e. 1) localization and adsorption of thiophene molecule, 2) the interaction of S atoms with Ni followed by the removal of sulfur from thiophene molecule, 3) the formation of nickel sulfide species, 4) the S-transfer from Ni to ZnO forming ZnS on the catalyst surface, 5) the diffusion of ZnS from surface to bulk. Above 5 steps can be further classified into two categories, i.e. the catalytic HDS steps (1, 2 and 3) and S-transfer steps (4 and 5). Above network was generally accepted in a series of reports by Bezverkhyy and co-workers [21–24].Regarding the HDS steps, Robert and Angelici [25] summarized configurations of 8-coordinated organometallic complexes formed by thiophene interaction with transition metal, and classified those complexes into three major types, i.e. π-complexation mode of thiophene, nickel sulfurthiophene with SM (σ) bond mode, and a combination of the former two adsorption modes. Some theoretical studies were performed on the thiophene desulfurization mechanism. Mittendorfer et al. [26–28] studied the preferential adsorption state of thiophene molecule on the surface of Ni (100). It was pointed out that the adsorption of thiophene molecule was preferred to be adsorbed as π-complexation mode. As soon as the C-S bond breaks, the S atom will fill the Ni atomic gap, forming NiS. Gao et al. [29] argued that both π-complexation mode and direct sulfur metal (S–M) bonds mode were possible and likely to be presented simultaneously. Following the thiophene adsorption, based on the previous studies, two basic hydrodesulfurization (HDS) reaction routes were proposed for the sulfur removal: direct desulfurization (DDS) and hydrogenative sulfur removal (HYD) [30]. The major difference between HYD and DDS lies on whether H2 is involved in the elementary steps for C-S bond cleavage. Despite numerous experimental works [31–35] on the HDS mechanism, it has, however, not been possible to clarify some fundamental arguments, due to the current limitations of the characterization techniques. Thus, the mechanism of thiophene transformation on Ni surface is still under debate.Regarding the S-transfer steps, Li et al. [36] studied the effect of ZnO particle size on the desulfurization performance. It was found that the apparent activation energy of sulfur removal on Ni/ZnO was much lower for smaller ZnO particles. After the ZnO particle shell was sulfided, the smaller particle size of ZnO could effectively decrease the energy barrier of sulfur transport in ZnO, leading to high sulfur removal efficiency in the latter stage, where sulfur transport is the rate limiting step. Moreover, the ZnO with smaller particle sizes could result in higher dispersion of Ni species on the ZnO surface, beneficiary for the overall activity enhancement. However, the direct evidence for the S-transfer has still not been reported, limiting a thorough understanding of the process. Zhang et al. [37] studied the desulfurization mechanism of thiophene via reactive adsorption over a Zn3NiO4 cluster by DFT calculation and found that thiophene was first decomposed on the Ni site of Zn3NiO4 bimetallic oxide catalyst to form nickel sulfide, followed by the reduction of the nickel sulfide with two reaction pathways. However, above theoretical model was apart from the practical case, because the oxidation state and dispersion were not taken into consideration.In this work, the RADS mechanism of thiophene on Ni/ZnO sample was studied by a combination of DFT calculation and in-situ experiments. The detailed HYD and DDS pathways of thiophene over Ni (111) were simulated by DFT models. The phase transition during desulfurization process on Ni/ZnO was studied by in-situ XRD and STEM-EDS technology, in order to acquire direct evidence of S-transfer phenomenon between Ni and ZnO. As a result, a comprehensive overview of RADS mechanism, initiating with hydrogenative desulfurization followed by sulfur absorption and transfer, was proposed. It provides fundamental understanding for the site requirement of the S Zorb materials, which were demonstrated at industrial scale.The DFT calculations were performed with the program package of DMol3 in the Materials Studio of Accelrys, Inc. The exchange correlation energy was calculated within the generalized gradient approximation (GGA) using the form of the functional proposed by Perdew and Wang, usually referred as Perdew-Wang 91 [38]. The density functional semi core pseudo potential method was employed for the Ni atoms, and the carbon, sulfur, and hydrogen atoms were treated with an all-electron basis set. The valence electron functions were expanded into a set of numerical atomic orbitals by a double numerical basis with polarization functions (DNP). Fermi smearing of 0.136 eV and real-space cut off of 4.5 Å were used to improve the computational performance. All computations were performed with spin polarization.The configurations of the adsorption complex were optimized by relaxing all atoms of the thiophene and two uppermost layers of the surface. The spin-polarized calculations were performed for the adsorption on Ni (111). Relative adsorption energies of all the configurations were calculated as given in Eq. (1). (1) E ad = E th/surf − E surf − E th In Eq. (1), the adsorption energy was defined as E ad. The energy notation with other subscripts, i.e. “surf”, “th” and “th/surf”, represent the total energy of clean surface, of the thiophene molecule and of the thiophene adsorbed on the surface, respectively. A negative value of E ad indicates that the adsorbed system was energetically favored, as compared to the ground state.Transition state (TS) searches were performed with the complete LST / QST (linear synchronous transit maximization / quadratic synchronous transit maximization) method. In this method, the LST was first performed, followed by an energy minimization in directions conjugating to the reaction pathway to obtain approximated TS. After the approximated TS was roughly determined, QST was thereof performed, followed by another conjugated gradient minimization. The cycle was repeated until a stationary point was achieved. The convergence criterion for the TS searches was set to 0.272 eV Å−1 for the root-mean-square of atomic forces. Vibrational frequencies were calculated for the initial / final states (IS and FS, respectively) and the TS, which were obtained from the Hessian matrix based on the harmonic approximation. The zero-point energy (ZPE) was accordingly calculated from the above-mentioned vibrational frequencies.All chemicals used for catalyst preparation were of analytical grade. ZnO was purchased from Simopharm Chemical Reagent Co., Ltd. Ni(NO3)2 was purchased from Alfa Aesar. ZnO powders, calcined at 650 °C for 4 h before use, were impregnated in aqueous solution of Ni(NO3)2 with 20 wt% of NiO loading. The resulted samples were dried at 120 °C for 4 h and calcined at 550 °C for 4 h in static air, denoted as NiO/ZnO. Further reduction was conducted at T = 400 °C, under H2 flow with p = 1.38 MPa and 80 mL min−1 flow rate for 2 h.X-ray diffraction (XRD) was used to characterize the crystal structure. In this work, XRD patterns were obtained by using a Siemens D-500 X-ray diffractometer equipped with Ni-filtrated Cu-Kα radiation (40 kV / 100 mA). The 2θ scanning angle range was 10–70° with a step of 0.02° s−1. The average crystal size was estimated from the line broadening of the most intense XRD diffraction peak by the Scherrer equation. The in-situ experiments were performed by coupling an in-situ cell (XRK-900) to the diffractometer under N2, H2 or S-containing hydrocarbons. The detailed conditions are shown together with the results later in the manuscript.TEM analysis was conducted on an aberration-corrected microscope (JEM-ARM200F) working at an acceleration voltage of 200 kV. STEM EDS-mapping was acquired from selected areas of the HAADF-STEM images with continuous drift correction.As evidenced by XRD, among all the facets of metallic Ni, the major exposed facet of reduced Ni/ZnO is Ni (111) (SI, Fig. S1). Therefore, theoretical studies on the energy and structure of thiophene adsorption were performed on the surface of metallic Ni (111) for both π-complexation mode and S-M adsorption mode (Fig. 1 ), and the calculated thermodynamic parameters are listed in Table 1 .It can be seen from the table that the adsorption enthalpy values of π-complexation mode and S-M mode are −156 kJ mol−1 and −129 kJ mol−1, respectively. Compared to SM adsorption mode, the more negative adsorption enthalpy of π-complexation mode indicates that π-complexation is the preferred adsorption mode of thiophene on the Ni surface. Note that the above model is based on the low coverage assumption, i.e. the adsorption is without the influence of neighboring adsorbed thiophene or other molecules, in line with the practical condition of hydrocarbon desulfurization (the sulfur content is at hundreds of μg g−1 level). The preference of thiophene's π-complexation mode is further supported by our observation from IR spectroscopy (SI, Section 2).After the thiophene molecule is adsorbed on the metal Ni surface with the π-complexation mode, the C–S bond length of thiophene molecule increases from 0.178 to 0.183 nm, indicating that the C–S bond is weakened. On the other hand, the S atom begins to approach the Ni atom, and the distance between the Ni and S atom is shortened from initial 0.231 nm–0.217 nm, close to the NiS bond length (0.220 ± 0.003 nm) [39]. This observation was in good agreement with previous experimental studies of XAFS [40,41], XPS [42] and TPD [43].In order to further distinguish the two desulfurization pathways, i.e. DDS and HYD, the elementary steps involved in the two pathways are considered. It is worth mentioning that the activation barrier of H2 splitting is very low, so it is rational that the major species of H2 is H radicals over the Ni surface in the steady state kinetic model.In the DDS pathway, from the adsorbed thiophene of π-complexation form, the reaction is initiated with the formation of NiS bond and the first CS bond cleavage without involving H radicals, forming C4H4S as the final state of the first elementary step. As for the second step, the hydrogenation of the intermediate C4H4S takes place on the C radical formed via the CS cleavage. Simultaneously, the Ni-S bond has been weakened, leading to a total exothermicity of the second step. The third and fourth steps follow the same order, i.e. CS bond cleavage before hydrogenation. Noteworthy, the energy requirement for the last two steps is much less than the first two steps, and the formation of S-free hydrocarbon molecule is highly exothermic (thermodynamically favored). Fig. 2 shows the corresponding structures of IS, TS and FS of different elementary steps in the DDS pathway.In this scenario, the first CS bond cleavage (Fig.2a) leads to a downward movement of the S atom to form bond with a surface Ni atom. By forming the first TS (A-TS1), the CS bond is elongated to 0.219 nm from 0.183 nm as in the IS. An intermediate C4H4S is then formed (A-FS1) on the surface. The Ni atom is included in the adsorption structure and thus forms a 6-member ring analog. This is similar to the organometallic reaction of thiophene, in which one metal atom inserts into CeS bond [44]. The activation barrier and reaction enthalpy change are calculated as 78.1 kJ mol−1 and 28.5 kJ mol−1, respectively. The second transition state (A–TS2) is encountered upon the hydrogenation of the intermediate C4H4S. The activation barrier of forming A-TS2 (35.6 kJ mol−1) is much lower than of forming A-TS1 (78.1 kJ mol−1). The second CS bond cleavage is activated with assistance of the CS bond stretching. After that, the intermediate C4H5S reacts with the H radical, forming adsorbed butadiene molecule and the S atom removed from thiophene is bonded to the Ni (111) surface. The activation barrier of these two steps are 50.5 kJ mol−1 and 14.9 kJ mol−1, respectively. The results indicate that the first step of CS cleavage is the rate-determining step in the sulfur removal via DDS pathway.As for the HYD pathway, in contrast to the DDS, hydrogen is involved in the CS bond cleavage. It is still rational that the H radicals are presented on the surface with high coverage. The CS bond is only weakened upon the interaction between H radical and the neighboring C atom, forming an unstable 5-coordinated carbon (Step 1). The CS bond is subsequently split, leading to the formation of surface C4H5S intermediate (Step 2). Following the first CS bond cleavage, the second H radical hydrogenates another C atom besides the S atom, weakening another CS bond (Step 3). Finally, the second CS bond is split (Step 4) and the reaction sequence is then closed. The intermediate estimation and energy profile is shown in Fig. 3 .By now, the overall reaction network with different pathways, i.e. DDS and HYD of thiophene, has been thoroughly simulated by DFT modeling. The energy profile is compared for the HYD and DDS pathways, as shown in Fig. 4 .The activation barrier of “pre”-hydrogenation step in the HYD pathway is much higher than that of CS bond scission in DDS pathway (117.2 vs 78.1 kJ mol−1). Therefore, it can be confidently concluded that thiophene desulfurization on Ni (111) proceeds along the DDS pathway preferentially, thus the first half of sulfur trail has been revealed.Previous works suggested a general scenario that the removed sulfur undergoes a sulfur-transfer process from Ni to ZnO, and finally be absorbed by ZnO. However, there are still no direct evidences of the sulfur-transfer and the phase transformation via sulfur contact. In this section, in-situ XRD and STEM with EDS mapping were applied to deliver the information of key intermediates during the whole sulfur adsorption process, in order to provide convincing evidences of above-mentioned S-transfer scenario.The XRD patterns of the reduced fresh and used Ni/ZnO samples are shown in Fig. 5 . The reduced sample shows the characteristic diffraction peaks at 44.1° and 51.3°, which are correspondent to the Ni (111) and (200) planes of metallic Ni (JCPDS 65–2865), while the characteristic peaks for NiO phase (37.3°, 43.3° and 62.9°, JCPDS 65–5745) were not observed. Therefore, upon reduction in H2 for 2 h, NiO is completely reduced to metallic Ni. Furthermore, a weak diffraction peak appeared at 42.9° which was attributed to a metallic NiZn alloy and indicated that the reduction treatment of the sample may lead to a partial reduction of ZnO and the solid–state reaction between Zn and Ni, in line with what has been noticed by Bezverkhyy et al. [23] on the reduced Ni/ZnO.After RADS for 12 h, peaks at 21.8°, 31.1°and 49.7° that correspond to rhombohedral Ni3S2 were observed and the intensity of Ni3S2 peaks increased significantly with increasing time on stream. Simultaneously, peaks at 26.9° and 28.6° that correspond to cubic ZnS were also observed on the used sample, and the intensity of ZnS peak increased gradually with the time on stream. By far, it has been shown that Ni3S2 and ZnS are the major new phases formed during the sulfur exposure to the initial Ni/ZnO materials. However, there is still not enough evidence to demonstrate the sulfur transfer from Ni3S2 to ZnO. Wang and coworkers [45,46] reported a similar case of sulfur transfer on a desulfurization of dibenzothiophene (DBT) using ZnO-based sorbent. It was suggested that the Ni was responsible for sulfur removal from hydrocarbon, while the residual sulfur was transferred to the ZnO phase in the presence of H2. The previous study showed comprehensive evidences, e.g. XRD, XAS and XANES, on the transformation of Ni and ZnO phases during the reaction. Unfortunately, the characterization was focused on each phase, but the overall scenario of both phases was missing.Moreover, the mechanism was based on DBT desulfurization in the above study. DBT is the major sulfur-containing molecule in the gas oil, which is usually used for diesel production. It is well known that the DBT desulfurization reactivity is very low [47], so that the desulfurization step occurring on the Ni site is naturally considered as the rate limiting step. However, the RADS technology is currently applied for the raw gasoline refinery, where the major sulfur comes from thiophene with much higher reactivity. In such cases, whether the similar sulfur transfer mechanism is valid remains vague, so it is eventually worth revisiting with further insight.In order to further confirm the occurrence of S-transfer, a mechanical mixture of Ni3S2 and ZnO was prepared with the same Ni/Zn ratio as in the Ni/ZnO sample reported in the current work. 200 mg of the Ni3S2/ZnO mixture were exposed to hydrogen and nitrogen under 1.0 MPa for 24 h at different temperatures. The results of in-situ XRD are shown in Fig. 6 .The characteristic diffraction peaks of ZnO, ZnS and Ni3S2 did not change substantially after nitrogen treatment. In contrast, in the XRD patterns of the physical mixture Ni3S2/ZnO before and after H2 exposure, the intensity of Ni3S2 and ZnO peak decreases slightly, while that of ZnS peak increases slightly, indicating that sulfur is gradually transferred from Ni3S2 to ZnO to form ZnS in the presence of hydrogen. It provides direct evidence of the sulfur-transfer between Ni3S2 and ZnO under H2 atmosphere, implying that in the S Zorb process, the presence H2 is more essential for the sulfur-transfer than for the desulfurization over Ni.In addition to the above observation, there has still been one open issue to unravel. In previous work, the proposed mechanism was lack of direct evidence at the interface between Ni and ZnO, which is the key feature of the adsorptive desulfurization process. Thus, in order to more visibly describe the sulfur transfer process in the adsorptive desulfurization, and to further confirm the spatial distribution of sulfur content on the sample, the samples after 24 h of desulfurization reaction were analyzed by STEM-EDS.In combination with the EDS surface scanning element distribution mapping (Fig. 7 ), the spatial distribution of active components in Ni/ZnO samples after desulfurization can be further determined. It was further confirmed that Ni3S2 was formed mainly after the desulfurization reaction, which is consistent with in-situ XRD results. After desulfurization, S elements are mainly distributed in the outer surface of ZnO (Fig. 7e), while ZnS and Ni3S2 mainly appear at the interface of Ni and ZnO (Fig. 7g). In order to realize S transfer efficiently, the suitable accessible distance can effectively reduce the hindrance / energy barrier of mass transport of S to ZnO, so that the sample has higher desulfurization activity. If the Ni and ZnO phases are not with close vicinity, the mass transport becomes difficult, and ZnO will partially not be able to interact with Ni3S2, resulting in incomplete utilization of sorption materials. The results are consistent with experimental observation that the desulfurization activity can be enhanced by using ZnO with small particle size [36]. Similarly, it is also reasonable to propose that the mass transport sulfur transfer can be reduced by using highly dispersed Ni.Based on above experimental investigation, it is possible to complete a desulfurization scenario evidently.First, thiophene is adsorbed onto the metallic Ni sites of the reduced samples. Thiophene molecules move freely in the gas phase and approach to the surface of the sample, mainly as π-complexation adsorption. During the adsorption of thiophene, the S atom gradually approaches the nickel atom, and the electron transfer occurs. Ni weakens the CS bond of thiophene ring, the CS bond length increases and tends to be split while the Ni-S bond length shortens. In parallel, under the present of metallic Ni, the hydrogen molecules adsorbed on the surface of nickel dissociate to form active hydrogen atoms.After twice of CS bond cleavage and following hydrogenation, the adsorbed thiophene ruptures to form Ni3S2 and C4 hydrocarbons. The sulfur cannot be retained with Ni, but gradually transferred from Ni3S2 to ZnO, forming ZnS in the presence of H2. The metallic Ni is therefore regenerated during the sulfur-transfer process. After that, the regenerated Ni sites can participate in the adsorption of thiophene again. The possible reaction paths are shown in Fig. 8 .As a concluding remark for the site requirement, in order to achieve efficient sulfur transfer, the catalyst active components Ni and ZnO is preferred to have close vicinity, or Ni needs to be highly dispersed on ZnO with small particle size. The ideal sample is that Ni is highly dispersed on the surface of ZnO, and the contact surface between nickel and ZnO is enlarged, which promotes S transfer. With the formation of ZnS, the concentration of ZnO nucleus decreases. The nickel active center can be recovered until ZnO is completely transformed to ZnS. When the ZnO phase reaches the saturation of sulfur content in the bulk, the Ni/ZnO material is required to be regenerated.The adsorptive desulfurization process of thiophene on Ni/ZnO was studied, and the desulfurization mechanism was elucidated. Thiophene molecule is preferentially adsorbed on metallic Ni as π-complexation mode. The desulfurization follows a DDS pathway. Namely, under the reaction conditions, the CS bond is weakened and split with Ni-S bond formation, and H2 is spontaneously split into H radicals, leading to the hydrogenation of C atom besides S. This is evidenced by DFT study. After twice of above CS bond cleavage and following hydrogenation, the S-free C4 olefin is formed and can be further hydrogenated, while the sulfur is retained on the Ni surface, forming Ni3S2 phase. The resulted Ni3S2 phase after desulfurization, which is in close vicinity with ZnO phase in a typical S Zorb catalyst, undergoes a sulfur-transfer towards ZnO. As a consequence, the Ni active site is recovered and the sulfur moves towards ZnO, forming ZnS phase. Efficient sulfur-transfer is achieved at the interface of Ni/ZnO materials with the presence of H2. This is demonstrated by a multi-technique characterization of in-situ XRD, STEM and EDS mapping. For the S Zorb process which has been widely applied in large scale, the current study is specifically meaningful, as it provides convincing evidences of the essential materials transformation during practical operation.There are no conflicts to declare.The authors are grateful to Prof. Dr. Mingyuan He for fruitful discussion on the scientific scope of the manuscript. Technical supports of STEM experiments from Yanjuan Xiang (Sinopec RIPP) are highly appreciated. This work has been financially supported by research grant from Sinopec (Fund No. 118016-8).The following is the supplementary data related to this article: Multimedia component 1 Multimedia component 1 Supplementary data related to this article can be found at https://doi.org/10.1016/j.gee.2020.05.010.

The reactive adsorption behavior of thiophene on the reduced Ni/ZnO sample was investigated by a combination of theoretical and experimental study. It is widely accepted that Ni is responsible for the sulfur-removal of thiophene to release S-free hydrocarbons. Such surface reaction was simulated by DFT method. It is demonstrated that thiophene is mainly adsorbed as π-complexation mode over metallic Ni. During desulfurization, the SNi bond is formed and the CS bond is thus split without pre-hydrogenation, resulting in the formation of Ni3S2 phase and S-free C4 olefin which can be further saturated in the presence of H2. The S-transfer between Ni3S2 and ZnO was monitored by in-situ XRD and STEM with EDS mapping. Two essential features were identified for efficient S-transfer, namely, 1) the H2 atmosphere, and 2) the two phases are presented with close contact. Based on the acquired information, a general scenario of sulfur trail has been proposed for the desulfurization of thiophene on Ni/ZnO.

Diesel oxidation catalysts (DOCs) are used in the automotive industry to oxidize hydrocarbons and CO and convert NO to NO2, which is critical to oxidize soot collected on DPFs and improve SCR efficiency. Nowadays, the emphasis on lowering real world driving emissions requires active catalysts for CO and hydrocarbon oxidation at temperatures significantly lower than the current state of the art to meet future pollutant emission regulations, specially associated with city driving [1]. In this regard, the U.S. Department of Energy roadmap has set the goal to achieve greater than 90 % conversion of criteria pollutants at 150 °C or lower for the full useful life of the vehicle [2]. To tackle this challenge, new catalyst formulations are being developed in order to achieve advances in low-temperature DOCs.In search of low-temperature CO oxidation formulations, single atom catalysts (SACs) have reported high reactivity, which also provide efficient utilization of platinum group metals (PGM) [3–5]. For instance, Datye et al. have reported high performance of Pt single atom catalysts over ceria support for low-temperature CO oxidation, reaching 90 % of CO conversion at 64 °C [6]. Nevertheless, noble metal-based catalysts can still be considered expensive and scarce, continuously leading to the scientific community to the search for low cost and with good performance alternatives such as inexpensive transition metals based-catalysts (i.e. Fe, Mn, Cu, Co or Ni) [7–9]. Actually, Kim et al. reported in a recent review [10], that when active metals are loaded on the CeO2 surface, many active sites could be acquired by increasing the dispersion, and the catalytic activity can be dramatically improved by newly introducing the interfacial sites between the metals and the CeO2 support. Among these active metals, one of the most promising candidates for the CO oxidation reaction is the CuO-CeO2 system [11]. The reason lies in the fact that exists an important synergistic effect between copper and ceria to generate its exceptional catalytic activity [12].Nowadays, many progress has been made in order to understand the origin of these synergies that can be generated in copper-ceria systems [10,13–21]. More specifically, the key features of the CuO-CeO2 system which contributes to the denominated synergetic effect are: i) the facilitation of oxygen vacancies formation; ii) the redox interplay between copper and cerium pairs (Cu2+ + Ce3+↔ Cu+ + Ce4+); iii) the superior interfacial sites with enhanced reactivity; iv) the higher reducibility; and v) the enhanced oxygen mobility. Moreover, although the use of the undoped ceria as a support is well documented for copper-ceria binary systems, ceria-zirconia mixed oxide can be considered as a better alternative due to its enhanced thermal resistance and superior ability to promote the creation of oxygen vacancies and, thereby, enhancing oxygen mobility [22,23].In light of the above aspects, the peculiar reactivity of copper/ceria-based materials is due to many different physical-chemical contributions, which results in a much rather complex system in practice. In fact, the modulation of metal-support interactions has been deeply investigated by employing different: i) ceria-based support morphologies [24–26]; ii) synthesis routes [27–29]; iii) metal oxides precursors (for copper and cerium) [15,25,26]; and even iv) inverse configurations [30,31]. In that way, significant changes have been achieved in the composition, the shape, the size, and the electronic state of these copper/ceria-based systems, resulting in different catalytic performances. Given the current requirements for highly active, efficient and selective catalysts at very low temperatures, it is imperative to keep on investigating the nature of these synergistic effects on copper/ceria-based catalysts.This research aims to conduct a systematic study of CO oxidation reaction catalyzed by several copper/ceria-zirconia samples, evaluating factors affecting catalytic activity under different preparation routes to incorporate copper (co-precipitation, incipient wetness impregnation and physical mixing methods) and different copper contents (from 0.5–6 wt.%) for the synthesized catalysts. Focus will be put in this case linking CO oxidation catalytic results in parallel with CO-temperature programmed reduction profiles and selected characterization parameters in order to find out the correlation among catalyst’ properties/reducibility and catalytic behaviors, especially those corresponding to the nature and roles of the different CuOx species over ceria-based support on catalytic activity.The whole study will allow us not only to provide some insight into the nature and type/s of active site/s determining the CO oxidation, mainly at low temperatures, but also if the magnitude and extent of these active sites could be modulated or controlled by choosing a preparation route and a certain copper content. This approach is of paramount importance for the effective and rational design of last-generation catalysts.The Ce0.8Zr0.2O2 mixed oxide (denoted as CZ) has been synthesised by the co-precipitation method in alkaline medium, by using the cerium and zirconium precursors (NH4)2Ce(NO3)6 (supplied by Panreac with 99.0 % purity) and ZrO(NO3)2·xH2O (supplied by Sigma–Aldrich, with x ≈ 6, technical grade), respectively. The appropriate amounts of these precursors were dissolved in distilled water. The corresponding hydroxides of cerium and zirconium were co-precipitated by drop wise addition of a 10 % ammonia solution in water until pH = 9, under constant stirring. The solid obtained was then filtered under vacuum and the yellowish precipitate was washed with distilled water until neutral pH. Finally, it was dried overnight at 110 °C and calcined in air in a muffle at 500 °C for 1 h, with a heating rate of 10 °C/min. The composition of this mixed oxide (Ce0.8Zr0.2O2) was chosen in terms of optimal thermal resistance towards sinterisation and good response towards other catalytic reactions studied by some of these authors [32,33] if compared with bare ceria and other Ce/Zr compositions analyzed.Ce0.8Zr0.2O2-supported copper catalysts with Cu wt.% of 0.5, 1, 2, 4 and 6 (denoted as Cu0.5CZ-IWI, Cu1CZ-IWI, Cu2CZ-IWI, Cu4CZ-IWI and Cu6CZ-IWI, respectively) were prepared by incipient wetness impregnation with Cu(NO3)2·3H2O (supplied by Panreac with 99.0 % purity) solutions of different concentration. After impregnation, the samples were dried overnight in an oven at 110 °C and thereafter calcined under air at 500 °C for 1 h, with a heating rate of 10 °C/min. Due to the limitations of the Cu(NO3)2·3H2O solubility in water, the Cu 0.5 %, Cu 1% and Cu 2% catalysts were impregnated in a single step, while Cu 4% and Cu 6% were prepared by successive impregnations with the solution used to prepare Cu 2% (two and three consecutive impregnations, respectively).In order to study the influence of copper entities in contact with ceria-zirconia, which eventually yield synergistic catalytic effects, several samples with different preparation routes have been prepared with a same copper loading (2%).The Ce0.76Zr0.19Cu0.05O2 sample (named as Cu2CZ-CP) was synthesized by the co-precipitation method in alkaline medium, by using the following cerium, zirconium and copper precursors: (NH4)2Ce(NO3)6 (supplied by Panreac with 99.0 % purity), ZrO(NO3)2·xH2O (supplied by Sigma-Aldrich, with x ≈ 6, technical grade) and Cu(NO3)2·3H2O (supplied by Panreac with 99.0 % purity), respectively. The same amounts of precursors than those used for the Cu2CZ-IWI synthesis were dissolved in distilled water, and after jointly co-precipitation, a 2 wt.% of copper was achieved. This co-precipitation procedure continues exactly as described above for the CZ catalyst.Bulk CuO was obtained by calcination of Cu(NO3)2·3H2O in air, at 500 °C during 1 h with a heating rate of 10 °C/min. It was used to prepare the physically-mixed samples, and also as a catalyst itself.A physically-mixed sample of bulk CuO with ceria-zirconia was also prepared. Physical mixing between CuO and CZ have been conducted by tight contact mode in an agate mortar with pestle, and consisted of an intimate mixture of the proper amount of CuO and CZ, during 5−6 min, to yield a 2 wt.% of copper. Afterwards, it was calcined in air at 500 °C for 1 h, with a heating rate of 10 °C/min. The sample obtained after this treatment is denoted as Cu2CZ-PM.The commercial 1%Pt/Al2O3 catalyst was supplied by Sigma-Aldrich (with BET surface area of 160 m2/g).Catalytic tests for CO oxidation were carried out in a U-shaped quartz reactor (16 mm inner diameter), loaded with 50 mg of catalyst and 100 mg of silicon carbide. The total flow rate of the feed gas (1000 ppm CO and 10 % O2 in He) was 100 mL/min, corresponding to GHSV of 90000 h−1. The catalytic tests consisted of Temperature-Programmed Reactions, where the temperature was increased from room temperature up to 300 °C at 5 °C/min under the reactive atmosphere. Previously, the samples were pretreated in situ at 500 °C under a flow of 5% O2/He (100 mL/min) for 30 min. The outlet gases were analyzed using a gas chromatograph (HP model 6890 Plus Series) equipped with two columns: Porapak Q 80/100 for CO2 separation and Molecular Sieve 13X for O2 and CO separation, coupled to a thermal conductivity detector (TCD). The CO conversion (XCO, %) was calculated as follows (1): (1) X C O % = [ C O ] i n - [ C O ] o u t [ C O ] i n x   100 where [CO]in and [CO]out are the CO concentration (ppm) in the inlet and outlet gas streams, respectively.Reaction rates were estimated as μmol CO2 produced/gCu·s, at certain temperatures, and apparent activation energies (Ea) were calculated assuming differential conditions (CO conversions ≤ 20 %). Repeatability of the catalytic measurements considering different batches was quite good.Temperature-programmed reduction (TPR) measurements employing CO as a reductant were carried out using the same experimental setup than that employed for the catalytic tests for CO oxidation. The total flow rate of the feed gas (5% CO in He) was 35 mL/min. The temperature was increased from room temperature up to 650 °C at 5 °C/min under the reactive atmosphere. Prior to every run, the catalysts were pretreated in situ at 500 °C under flow of 5% O2/He for 30 min, and then, the catalysts were cooled down to room temperature in the gas flow and purged under inert gas. The whole details of the procedure are described elsewhere [34].A very complete description of the classical characterization techniques, used for the investigation of the physico-chemical features (surface, textural and structural properties), of the fresh catalysts used in this work is described in depth elsewhere [35,36]. The protocol for determining the copper dispersion data in selected catalysts, by means of H2 volumetric studies.Catalytic testing was performed to understand the effect of the amount and nature of copper species present on the catalysts on activity. CO oxidation was performed first on IWI-samples with various copper loadings, including the undoped Ce0.8Zr0.2O2 sample and a reference CuO sample calcined at the same temperature (Fig. 1 ). Complete conversion of CO is achieved for all the catalysts, at temperatures lower than 300 °C, with the exception of the undoped ceria-zirconia (CZ). The amount of copper significantly affects the ignition, that occurs at temperatures as low as 50 °C for the best catalysts, and the oxidation rate, that gradually increases with temperature. The slope of the light-off curve seems to be influenced by the copper contents onto the catalysts, the higher the copper loading, the more pronounced the slope. The shape of the curves, in general, differs from that of the 1%Pt/Al2O3 commercial catalyst (chosen as an effective benchmark, due to its high activity towards other oxidation reactions and high BET surface area), which is characterized by an abrupt ignition between 110 °C and 140 °C. This characteristic rapid increase from low to high conversion can be explained due to the first-step coverage of the active sites with CO at low temperatures, inhibiting initially the CO oxidation reaction [37,38]. Conversely, all the synthesised ceria-based catalysts are active at temperatures as lower at 75 °C, including the support. Interestingly, the bulk CuO seems to be quite active, if compared with the uncatalyzed reaction, and around 200 °C even outperforms the support’s activity. The fact that CuO entities (like CuO bulk) can be active towards CO oxidation under these experimental conditions, could partially account for the non-gradual increasing trend in activity with the highest copper contents of the catalysts. Fig. 2 depicts the effect of copper content on the CO oxidation activity expressed at the temperatures at which several CO conversions are reached: 10 % (T10), 50 % (T50) and 90 % (T90), respectively. As the copper loading increases, (in the range among 0.5 and 2%), those parameters are lowered indicating that the catalyst activity becomes better. However, the profile of these curves exhibits minor decrements of T10, T50 and T90 temperatures with the highest copper contents (4 and 6%). Zhu et al. have studied the CO oxidation behavior of related formulations, where copper was incorporated by incipient wetness impregnation over a Ce0.5Zr0.5O2 support, and have reported an optimal copper oxide loading around 5.25 % [39].Following with the ideas of presenting reliable comparisons with an effective benchmark, (the commercial platinum’s catalyst), Table 1 compiles the list of the intersection temperatures of the copper-containing catalysts’ curves with regard to the Pt’s curve. The fact that copper/ceria-zirconia catalysts prepared from IWI method outperforms Pt’s activity in a wide range of low temperatures, can point to high degrees of CuOx entities’ dispersion, and consequently, quite small average crystallite sizes, since will be discussed later. The more copper content on the catalysts, the higher the temperature should be reached by the platinum’s catalyst to achieve the same value of conversion than those of the corresponding copper catalysts.In our work, the copper-containing catalysts exhibit decrease of BET surface area with increase in metal loading due to partial blocking of porosity as a consequence of the IWI method, but, on the other hand, the incorporation of copper seems to be critical to define the activity. To ascertain and split the influence of both effects, Fig. 3 shows specific rates for this set of catalysts normalized to catalyst surface area. Surface-area normalized values were derived from the rate of CO2 production per second and per square meter of solids’ surface areas. By comparing Fig. 1 with Fig. 3, it can be said that the differences are becoming larger for the normalised parameters with the copper content. By comparing the lowest copper contents catalysts (0.5, 1 and 2% with regard to CZ), the values seem to increase proportionally with the copper content. This trend is quite interesting suggesting that the concentration of active sites per m2 increases in a gradual way with the copper loading, thus pointing out the goodness of the preparation method under low copper contents (providing this interesting trend). This tendency is also consistent with the extremely low (and similar) copper crystallite sizes (1.7 nm for Cu0.5CZ-IWI, 1.7 nm for Cu1CZ-IWI and 1.8 nm for Cu2CZ-IWI, respectively, see Table S2 on SI) which provides certain evidences of very well-spread CuOx entities able to create a relatively high population of active interfaces. When increasing the contents up to 4 and 6%, the increase is not gradual anymore. This is reflected in Fig. 4 for three representative temperatures of reaction (80, 100 and 120 °C, respectively). These trends could suggest the co-existence of several active sites of different nature/relevance for this catalytic process. It should be reminded that: i) both support and CuO bulk are active for the process and ii) the IWI process brings differences in the steps conducted, because for samples Cu4CZ-IWI and Cu6CZ-IWI successive impregnation steps were needed to incorporate the desired copper contents (two and three consecutive impregnations, respectively). Despite this consideration involved in the preparation method, Cu crystallite size is still very low for Cu4CZ-IWI (2.3 nm) and Cu6CZ-IWI (2.7 nm), which justifies that the CO oxidation activity remains high.It should be noted that crystallite size determined by H2 chemisorption method is a measure of metallic copper particle size. However, CuOx entities are the supposed active species towards CO oxidation. CuO particle size can be calculated from metallic Cu particle size by assuming that CuO will adopt a spherical or near-spherical shape when they are dispersed over ceria-zirconia. Therefore, the size of CuO crystallite is only slightly larger than that of metallic copper particle (Table S2).It is relevant in this context, as well, to compare the reactivity per total metal content on the catalysts. Therefore, the representation of reaction rate (μmol CO2/gCu·s) versus the inverse of reaction temperature is illustrated on Fig. 5 . Consistently with the information shown on Fig. 4, by increasing the copper contents in the range of 0.5, 1 up to 2%, the representation of the reaction rate expressed per gram of copper versus the inverse of temperature is almost identical for these three catalysts indicating that the number of active sites participating in the process seem to grow in a gradual way when adding more copper contents in this range. However, the representation of 4 and 6% Cu contents falls below the trend traced by the mentioned three catalysts, thus suggesting a lowest effectivity for copper metal atom into these two catalysts. All these evidences are in general agreement with the trend in copper dispersion data.The composition of 2% of copper loading was selected to analyse the relevance of the preparation method thus unraveling if CuOx entities of different nature can exist and which is the relationship among their type and amount with their own catalytic activity. Indeed, given the complexity of the copper/ceria-zirconia system, it is challenging to shed light on the preliminary identification of the active sites and their correlation with catalytic activity. For this purpose, three different preparation routes (very different among them, regarding their physico-chemical fundamentals) were approached. It is supposed that different procedures of copper incorporation onto the catalyst will yield different degrees of contact/distribution/nature of copper species onto the ceria-zirconia catalyst.In line with these ideas, three catalysts with the same 2% copper content were synthesized. The corresponding nomenclature is: Cu2CZ-IWI, Cu2CZ-CP and Cu2CZ-PM. The second solid was prepared by a combined co-precipitation procedure of the metallic precursors and the third solid was obtained by an intimate physical mixture of the support and the copper precursor and, subsequent calcination a 500 °C (see Section 2 for additional details). Besides, the BET surface areas obtained for the three catalysts are highly similar (71, 72 and 70 m2/g, respectively), thus allowing us a reliable comparison of the influence of the different nature of the copper species, since both parameters (copper content and exposed surface areas) are practically identical in the three samples considered. The corresponding CO conversion curves are depicted on Fig. 6 . These three copper/ceria-zirconia catalysts were found to be more active than unsupported ceria-zirconia and bulk CuO. It is relevant to point out that the route of procedure significantly affects the catalytic activity, both the onset reaction temperature and the slope of the curve. The order of activity is as follows: Cu2CZ-PM < Cu2CZ-CP < < Cu2CZ-IWI In accordance with the exposed ideas, reaction rates against the reciprocal of temperature for these three catalysts show the highest values for the catalyst prepared by the IWI method on Fig. 7 . To better understand the differences in reactivity provided with the different variables studied (copper content and type of preparation procedure), CO-TPR experiments are presented below, with the aim to determine the distinctions among all the catalysts in the reaction between CO and oxygen from the surface/lattice of the catalysts. The combined analysis of all the experimental data (catalytic tests, CO-TPR results and characterization of the catalysts) will allow us to ascertain if the nature and type of copper species can play a role during the promotion/activation of oxygen from the catalysts, taking part in the catalytic reaction.CO-TPR has been widely used to characterize reducibility of the CuO-CeO2 systems [40]. Additionally, with the general purpose to infer some arguments which can assist the understanding of the trends in the catalytic activity towards CO oxidation of the different sets of catalysts, the use of CO is preferred as a probe molecule instead of H2 [41]. Nevertheless, the discussion of the influence of copper contents and preparation procedure on the H2-TPR profiles for the catalysts studied was presented elsewhere [35,36] and the corresponding data and additional interpretation for selected catalysts is reported on the next section.CO-TPR experiments have been performed in order to investigate the influence of copper entities of different possible nature and interaction degree with the support on the catalysts’ reducibility [42]. In fact, if several CuOx species co-exist onto these samples, they could play different roles during the activation of surface/sub-surface oxygen from the ceria-based support, which could produce, eventually, characteristic CO2 emission profiles for this type of copper/ceria-zirconia systems [13,40,42,43]. Besides, by comparing the corresponding profiles of the different catalysts prepared with those of bulk CuO and the bare support, possible synergetic effects among CuOx species well spread onto the catalysts’ surface and the cerium centers in close vicinity with them (thus facilitating ceria’s reduction due to close interfacial interactions, well commented on literature [18,20]) could be evidenced. Fig. 8 A and B shows CO2 emission profiles during CO-TPR experiments for the two sets of copper/ceria-zirconia catalysts analyzed in this work. The reduction profile of pure CuO, is characterized by a single and broad peak in a temperature range from 120 to 260 °C, indicating the temperature window where the unsupported CuO tenorite-like species reduction is expected to take place. In this case, CuO sample’s profile is presented on the Fig. 8 like a calculated curve corresponding to 2% of bulk CuO profile, for comparative purpose. Considering the conditions employed, this broad peak can be ascribed to direct reduction to metallic copper, in agreement with previous studies [13,44,45]. The undoped ceria-zirconia profile exhibits an asymmetric first broad peak starting at 80 °C, which can be also considered a shoulder of a second one centered at around 430 °C. Unfortunately, the interpretation of its CO-TPR profiles becomes extraordinarily complex due to the simultaneous occurrence of the water-gas shift (WGS) reaction and Boudouard reaction, according to these Eqs. ((2) and (3), respectively): (2) CO + OH− → CO2 + 1/2 H2 (3) 2CO → CO2 + C thus providing a constant CO2 emission level at high temperatures during the experiments, which makes that the profile does not return back completely to the baseline. Many authors have revealed the contribution of these side reactions catalyzed by ceria-based catalysts, under CO-TPR experiments [13,46–48], mainly at medium temperature range (> 275 °C) for WGS reaction and high temperature range (> 400 °C) for Boudouard reaction. For the sake of a reliable comparison and taking into account that the most interesting/representative peaks or contributions appear on the corresponding patterns of copper/ceria-zirconia catalysts at low/medium temperatures, the subtraction of the whole profile of the bare support to those corresponding to the catalysts prepared, was conducted in an attempt to remove the side effects of the WGS and Boudouard reactions (whatever the extension at which both reactions take place). Additionally, this subtraction highlights in an adequate manner: i) the CuOx entities’ reduction patterns and ii) possible synergetic effects arisen from the copper entities in close contact with ceria-zirconia’s surface, consisting of “extra” cerium centers’ reduction as a consequence of promoted interfacial interactions.In line with these exposed ideas, Fig. 9 A and B depicts the “treated” profiles after conducting these subtractions. Assuming that the Boudouard reaction occurs in similar extent for all the catalysts, Fig. 9A shows two different low-temperature peaks or contributions, that can be ascribed to the CuOx species reduction or to the mentioned interfacial interactions which could yield to additional cerium centers reduction. Cu2CZ-IWI and Cu2CZ-CP presents a very low temperature broad contribution, centered at around 80 °C and completely absent both in the catalyst prepared by physical mixing method and in the CuO bulk sample. Some authors reported in the literature that this first contribution could be ascribed to very finely CuO dispersed onto ceria-based materials or to the reduction of copper species strongly interacting with ceria [40]. The experimental fact that this α contribution arises at lower temperature than those reported in other papers is consistent with the very high activity of these catalysts towards CO oxidation (even outperforming that of Pt’s catalyst) [40]. This comparison also suggests that the physical mixture method yields a poorer copper/support contact or a lack of these very finely dispersed CuOx species and for this reason this catalyst presents much lower activity in the range of low temperatures compared with their counterparts having the same copper loading. This behavior is consistent with the high copper crystallite size for Cu2CZ-PM (32.7 nm) and low dispersion data (3%), estimated from H2 adsorption volumetric studies.Conversely, the second contribution (β) appears as a clear peak (specially for Cu2CZ-IWI and for Cu2CZ-PM) in the temperature window of the unsupported CuO. Interestingly, by comparing the relative areas of the CuO bulk (2% of the whole profile) against these catalysts, the inferior area of the reference CuO can be clearly seen whatever the catalyst considered, but specially for Cu2CZ-IWI. Moreover, the profiles are negative at medium temperatures. A reasonable explanation for this observation is that a concomitant reduction of cerium centers is taking place in this temperature range (occurring at higher temperatures in the bare support). In order to tentatively distinguish if it is motivated by an “anticipated” reduction of the support or if, additionally, “extra” reduction of cerium centers, in close vicinity with the several active CuOx species, takes place, additional quantifications were estimated and compiled on Table 2 , as commented on below.Regarding the effect of the copper content for IWI-catalysts (Fig. 9B), three contributions (α, β and γ) are presented in different extensions and relative importance, being γ specially relevant for the catalyst prepared with 6% of copper loading (this peak appears like a very low and broad contribution or is absent for the rest of catalysts). All the IWI catalysts show the contribution at low temperatures (α), which seems to reach maximum values for Cu2CZ-IWI and Cu4CZ-IWI. For these two catalysts, the second contribution (β), emerging in the temperature window of the unsupported CuO tenorite-like species, presents the highest values of the series as well. Conversely, the Cu6CZ-IWI catalyst exhibits less intensity for these two first contributions (α and β) in favor of a very prominent and sharp peak at higher temperature (γ). It is worth reminding that this catalyst was prepared by three successive impregnations of the copper precursor solution.As advanced, Table 2 compiles additional quantifications estimated from Fig. 9A and B. The data on the second column correspond to the integrated amounts of CO2 emitted by each catalyst after subtraction of the CO2 emission corresponding to the ceria-zirconia’ support. The third column collects these values after subtraction of the theoretical CO2 emissions corresponding to the stoichiometric reduction of the CuO molar contents of the different catalysts, according to this global stoichiometry (4), (by assuming a general formula of CuO for all the CuOx entities): (4) CuO + CO → Cu + CO2 Assuming the validity of these subtractions, the “extra” amounts obtained as μmol CO2/g of every catalyst (Table 2) suggest that whatever the method of preparation of the catalyst or the amount of copper introduced by the IWI method, “additional” or “extra” cerium centers reduction can be measured, thus suggesting a good degree of interaction copper/ceria-zirconia and an excellent reducibility under CO in the systems prepared. Following with this discussion and in an attempt to rationalize these “extra” amounts along the sets of catalysts, the CO2 μmols/g estimated values were referred to the CuO molar content of the different catalysts. Assuming the general stoichiometry of cerium cations’ reduction on ceria’s surface (5): (5) 2CeO2 + CO → Ce2O3 + CO2 the last column on Table 2 compiles the potential number of cerium centers surrounding the CuOx entities which are supposed to suffer reduction according to these quantifications. It is clearly demonstrated that more cerium centres are affected due to improved synergetic interactions according to this order in terms of the method of preparation: Cu2CZ-PM < Cu2CZ-CP < Cu2CZ-IWI This was the order observed for the catalytic activity, and as well, the order also agrees with the contribution of the α peak (at the lowest temperature), which is more prominent for the IWI-catalyst.Regarding the effect of copper content, generally speaking, the approximate calculations report a trend consisting of more cerium centers affected by CO reduction in the vicinity of CuOx entities as the copper loading decreases. However, this seems not to be a linear effect with the copper loading. This observation can be tentatively explained by the variety of CuOx entities which seem to coexist on this set of catalysts.In recent years, copper/ceria-based catalysts have attracted considerable attention due to their low cost and excellent catalytic performance in many oxidation reactions [49]. Nevertheless, due to their inherent reactivity, the use of classical methods to determine copper entities’ dispersion by means of traditional chemisorption procedures is very problematic due to the existence of large spillover phenomena, (in the case of H2 chemisorption methods), or the possibility of concomitant oxidation of cerium centers if other probe molecules (CO and N2O, as examples) are used for chemisorption studies after a previous reduction step. In this context, H2 adsorption isotherms at sub-ambient temperatures or sophisticated imaging analysis techniques such as HAADF-STEM might be applied to determine reliable copper dispersion data in recent years (but with certain amount of difficulties) [50].Since it is widely recognized that the dispersion states, redox properties and catalytic performances of CuO/ceria-based catalysts are critically dependent on the preparation methods [49], this section will be devoted to a comparative analysis considering the characterization’s features of some representative catalysts of this study. On this basis, implications will be intended to be extracted concerning the nature and roles of the different CuOx species on catalytic activity. By a combined discussion considering the whole characterization data obtained from catalysts prepared by different methods, CO-TPR results and CO oxidation activity data, an interesting interpretation of the correlation among catalyst’ properties and catalytic behaviors can be obtained.By having a look at the rate of CO2 production during the catalytic tests of CO oxidation at a representative temperature (e.g. 80 °C), it can be noted that the catalyst prepared by physical mixture presents a rate of CO2 production of 0.020 μmol/g·s (being an intermediate value among that of the bare support -0.011- and that of the lowest copper content catalyst prepared by IWI method (0.037), but quite far from that of the catalyst with a same copper content, (2%), prepared by IWI method as well (0.290). On the other hand, the coprecipitation method yields a catalyst whose rate of CO2 production (0.094) is much more similar to that of the 0.5 % Cu (0.037) and far from that of its counterpart 2% copper content (0.290). These data provide some signs about the relevance of the preparation method on the type, amount and nature of active sites on these complex systems.In order to have in mind the complete picture, Fig. 10 A and B illustrates the representations of the CO oxidation activities and the CO2 production profiles (obtained from the CO-TPR profiles) for selected catalysts. It is interesting to note that the nature and type of CuOx species generated onto the catalysts, modulated by the different methods, seem to exhibit more influence on catalytic activity values and catalysts’ reducibility than the whole content of CuO onto them. According to these results, Cu0.5CZ-IWI (prepared by incipient impregnation method) presents the same activity and similar redox properties than Cu2CZ-CP, prepared by a coprecipitated method of the three metal precursors, in spite of having 4 times less copper content.In order to shed light on these experimental observations, comparative characterizations results will be presented now. As commented earlier, the whole characterization results, dealing separately the effect of copper content and the influence of preparation method on the activity towards diesel soot combustion, were presented elsewhere [35,36]. For the discussion presented in this section, only representative characterization results of selected catalyst will be shown.First of all, XPS analyses will be presented. Cu-2p3/2 photoelectron spectra of selected catalysts (Cu0.5CZ-IWI, Cu2CZ-CP and Cu2CZ-IWI) were depicted on Fig. 11 . As illustrated, Cu0.5CZ-IWI with Cu2CZ-CP present much more similarities on their XPS spectra than Cu2CZ-CP with Cu2CZ-IWI (its counterpart with the same copper content). To deepen into the mode in which copper distribution takes place among the surface with regard to the bulk of the catalysts in terms of the synthesis route and the amount of copper for representative catalysts, Table 3 collects the corresponding Cu/(Cu + Ce + Zr) surface atomic ratios (designated as Cu/(Cu + Ce + Zr)sur) and the estimated bulk atomic ratios (designated as Cu/(Cu + Ce + Zr)nom). The catalysts prepared by incipient wetness impregnation present higher Cu surface atomic ratios than the theoretical bulk values, which means that copper is finely dispersed onto the support’s surface, partially blocking porosity, as inferred from previous publications [35,36]. Cu2CZ-CP has a lower surface atomic ratio than the theoretical bulk value, thus indicating that copper has been (at least in part) incorporated into the ceria-zirconia lattice or subsurface due to the combined coprecipitated method of preparation used. This makes that the fraction of Cu atoms exposed for Cu2CZ-CP remains much lower than that obtained by incipient wetness impregnation method, and yields values close to those shown by Cu0.5CZ-IWI.To approach the discussion about reducibility of the catalysts and trying to complement the data obtained by means of CO-TPR, results obtained by H2-TPR for selected catalysts are presented on Fig. 12 , whose detailed discussion was presented elsewhere [35,36]. Again, a relevant similarity can be found by comparing the profiles of Cu0.5CZ-IWI and Cu2CZ-CP (in terms of intensity and shape of the patterns), however, the catalyst prepared by incipient wetness impregnation presents the profile a little bit moved forward to lower temperatures, indicating a promoted reducibility at lower temperatures. Anyway, both profiles are considerably lower than that of the Cu2CZ-IWI, in agreement with the rest of catalytic and characterization data presented so far.With the aim of supporting the discussion regarding the extension of the reduction process occurring onto the copper-containing catalysts (if only “anticipated” cerium centers reduction, compared with the reduction of bare support, takes place, or if, additionally, “extra” or “new” cerium centers, not susceptible to be reduced in the case of the bare support, can be affected by the reduction process due to an excellent quality of the interphase copper species/ceria-zirconia, the same estimations, based on subtractions conducted for CO-TPR profiles, were carried out, now, with the data obtained from H2-TPR profiles, considering the following stoichiometries ((6) and (7)): (6) CuO + H2 → Cu + H2O (7) 2CeO2 + H2 → Ce2O3 + H2O In this case, the results (not shown for the sake of brevity), yield that, approximately, the same amount of cerium centers that suffer reduction onto the bare support are reduced “in advance”, but not “extra” cerium centers reductions are evidenced after the corresponding subtractions. These discrepancies can be motivated by two reasons: i) the estimations are subjected to several source of errors and, as a consequence, they need to be viewed as “approximate results” and ii) CO can be considered a compound with a more pronounced reducing character than H2 towards these catalysts (actually, having been conducted under the same experimental conditions, CO-TPR profiles are moved toward lower temperatures with regard to H2-TPR ones).This study has revealed that very high but different ranges of catalytic performances can be reached by means of different CuO/ceria-zirconia catalysts and that CO-TPR curves seem to be very sensitive to the presence of several types of CuOx entities onto this support, thus displaying several CO2 contributions/peaks due to the optimal interaction of CO molecules with the oxygen from the own reservoir of every catalyst. Both CuOx entities and ceria-zirconia were more readily reduced than, at least, the corresponding independent components. In addition, it is suggested that “extra” cerium cations could be reduced in variable amounts due to the excellent qualities of the interfaces created. Even though the whole estimations should be considered taking into account that some assumptions or oversimplifications were taken, attempts were tried to estimate the “extra” cerium centers affected by the reduction under CO, which is a primary trial of “titrating” the quality and extension of the interphase CuOx/ceria-zirconia, responsible of improvements of redox properties and catalytic responses in these systems with regard to CuOx well-dispersed onto different nature supports. These values are suggested to be influenced by copper content, (linked to dispersion of CuOx species) but more importantly to the method of preparation.To clarify the connection or coupling between the profiles obtained by CO oxidation (CO2 formed by reacting 1000 ppm of CO with 10 % of O2) and by CO-TPR patterns (CO2 originated by reacting 5% of CO with the oxygen coming from the own catalysts), a combined representation of both profiles expressed as μmol CO2/gcat·s is presented versus temperature for selected samples on Fig. 13 . Dotted lines correspond to the rate profiles from CO-TPR and solid lines to those of CO oxidation in presence of O2. The catalysts prepared by “chemical” routes (co-precipitation or incipient wetness impregnation) exhibit a remarked ability to activate and oxidize CO molecules with their own oxygen (coming from CuOx entities or surrounding ceria-zirconia’s surface). Conversely, the sample prepared by physical mixing between copper precursor and ceria-zirconia (and subsequent calcination) does not possess the “active sites” responsible of oxidizing CO at very low temperatures (α contribution is absent for this catalyst) and, accordingly, the CO2 production curve is delayed with regard to the rest of the catalysts. From the combined analysis illustrated on Fig. 13, the best catalytic response shown by Cu2CZ-IWI can be attributed to a joint presence of a relevant α contribution (probably CuOx entities very well-dispersed onto ceria-zirconia, proved by extremely low size of copper crystallite) and accessible CuO-like tenorite entities, (β contribution) similar to those presented by bulk CuO and Cu2CZ-PM (with high average crystallite size). These entities are probably at a surface level and very accessible and are able to promote reduction of cerium centers in close vicinity, as inferred for the respective comparison with bulk CuO area presented above. Conversely, the sample prepared by co-precipitation, shows much lower β and a slightly lower α contribution than its counterpart prepared by IWI. Part of its CuO species seem to be more “buried” onto the catalyst’s sub-surface, and as a consequence of this different copper distribution/accessibility, the resultant catalytic behavior reaches the same level than that of a catalyst prepared with much lower copper content, but more accessible on surface (Cu0.5CZ-IWI), as previously pointed out.It is assumed that oxidation of CO onto these catalysts proceeds though a Mars-van Krevelen mechanism, which has been previously invoked by ceria-based materials with very positive effect of facile oxygen transfer from the catalysts due to a favorable formation of oxygen vacancies. This is clearly promoted by the presence of CuOx species, because the relevant step during CO oxidation is the reaction between adsorbed CO and oxygen from the own reservoir of the catalyst. In this sense, CO-TPR studies were approached in order to analyze possible differences in the catalysts with the different amount and nature of CuOx species. Nevertheless, whatever the copper-containing catalyst studied, CO conversion is coupled with CO2 production during CO-TPR, evidencing a fast and facile transfer of oxygen from the catalyst (CuOx species/ceria-zirconia’surface) towards CO molecules in the absence of O2-gas phase, which could, eventually replenish the own reservoir of oxygen under oxidant atmosphere.These experimental evidences are congruent with the values of apparent activation energies (Ea) estimated from CO oxidation catalytic tests. The corresponding values are presented on Table 4 for all the catalysts studied. For most of the cases, CZ presents a higher apparent activation energy (68 kJ/mol) than that of copper-containing catalyst. It is worth noting that the couple of catalysts previously discussed because of similarities among physico-chemical features and CO-TPR profiles, (Cu0.5CZ-IWI and Cu2CZ-CP) are characterized by presenting a same value of this parameter (52 kJ/mol, lower than that of the bare support), supporting the idea that a similar number and type of active sites are present on these catalysts. On the contrary, Cu2CZ-IWI, showing a highlighted ability to transfer oxygen to CO molecules (prominent α and β peaks) presents the lowest Ea values (38/42 kJ/mol) in agreement with the idea of a higher number of active sites on this catalyst. This Ea value is even lower than that estimated for 1%Pt/Al2O3 catalyst, in agreement with a poorer reducibility of this catalyst. In this sense, the CO oxidation mechanism of alumina-supported platinum catalyst is known to take place via a single-site competitive Langmuir–Hinshelwood mechanism (suprafacial mechanism) [37]. The non-participation of oxygen lattice from the support in this mechanism provokes the low reducibility of 1%Pt/Al2O3 under CO-TPR conditions [51]. This contrasts with copper-containing catalysts, in which the high participation of the lattice oxygen atoms from these catalysts seen under CO-TPR conditions is indicative for the dominating Mars-van Krevelen mechanism (intrafacial mechanism). However, Cu2CZ-PM presents even a higher value of Ea (73 kJ/mol) than 1%Pt/Al2O3 and, accordingly, this catalyst is characterized by a reduced capacity to transfer oxygen to CO molecules, since the α contribution is completely absent for this catalyst.This research has been dedicated to the preparation and study of the catalytic activity of different copper/ceria-zirconia catalysts in order to understand the nature of the active sites and the generated synergies for CO oxidation reaction at low temperature. The general conclusions that have been drawn are the following: - All the catalysts obtained by incipient wetness impregnation method are more catalytically active towards CO oxidation than 1%Pt/Al2O3 at low temperatures (T < 130 °C), due to the high synergies created among CuOx species and the ceria-zirconia support, even at copper loading as low as 0.5 %. This seems to be connected with a very high copper’s dispersion degree reached with this procedure, yielding significantly low copper crystallite sizes. - Regarding the IWI-catalysts, the CO oxidation activity increases linearly with Cu loading up to 2 wt.%. When increasing the contents up to 4 and 6 wt.%, respectively the increase is not gradual anymore. - The synthesis method significantly affects the CO oxidation activity. The catalytic activity increases along the series: Cu2CZ-PM < Cu2CZ-CP << Cu2CZ-IWI. - Different CO-TPR peaks/contributions were observed in different extensions and relative importance for copper/ceria-zirconia catalysts. Remarkably, Cu2CZ-IWI and Cu4CZ-IWI exhibit the most intense low-temperature contribution (α peak), as well as β contribution. - Quite similar CO oxidation activities can be reached from different synthesis methods and different copper contents (Cu0.5CZ-IWI and Cu2CZ-CP), evidencing the importance of the nature and the type of CuOx species generated over the catalysts’ surface, which can be modulated by the synthesis procedure. These features seem to be more relevant than the own CuO content on catalysts. - From the combined study of CO-TPR and H2-TPR profiles, it can be seen that CuOx entities are reduced at low temperatures (very dependent on the preparation method) onto ceria-zirconia, but, importantly all the cerium centers susceptible to be reduced under CO and/or H2 in the support, are reduced in an anticipated way, with regard to the interval of temperatures where this reduction process takes place in the bare support. The detailed investigation of the several contributions and peaks that appear on the CO-TPR profiles, compared with the bare support and with the CuO bulk profile has been revealed as very useful for a first approach into the understanding of the synergies created on this system. Finally, the evidences provided by the reducibility and catalytic profiles of the catalysts prepared by the different procedures can contribute, interestingly, to the clues about the features which should be determinant for these catalyst to present very high catalytic activity towards CO oxidation, mainly at low temperatures. All the catalysts obtained by incipient wetness impregnation method are more catalytically active towards CO oxidation than 1%Pt/Al2O3 at low temperatures (T < 130 °C), due to the high synergies created among CuOx species and the ceria-zirconia support, even at copper loading as low as 0.5 %. This seems to be connected with a very high copper’s dispersion degree reached with this procedure, yielding significantly low copper crystallite sizes.Regarding the IWI-catalysts, the CO oxidation activity increases linearly with Cu loading up to 2 wt.%. When increasing the contents up to 4 and 6 wt.%, respectively the increase is not gradual anymore.The synthesis method significantly affects the CO oxidation activity. The catalytic activity increases along the series: Cu2CZ-PM < Cu2CZ-CP << Cu2CZ-IWI.Different CO-TPR peaks/contributions were observed in different extensions and relative importance for copper/ceria-zirconia catalysts. Remarkably, Cu2CZ-IWI and Cu4CZ-IWI exhibit the most intense low-temperature contribution (α peak), as well as β contribution.Quite similar CO oxidation activities can be reached from different synthesis methods and different copper contents (Cu0.5CZ-IWI and Cu2CZ-CP), evidencing the importance of the nature and the type of CuOx species generated over the catalysts’ surface, which can be modulated by the synthesis procedure. These features seem to be more relevant than the own CuO content on catalysts.From the combined study of CO-TPR and H2-TPR profiles, it can be seen that CuOx entities are reduced at low temperatures (very dependent on the preparation method) onto ceria-zirconia, but, importantly all the cerium centers susceptible to be reduced under CO and/or H2 in the support, are reduced in an anticipated way, with regard to the interval of temperatures where this reduction process takes place in the bare support. The detailed investigation of the several contributions and peaks that appear on the CO-TPR profiles, compared with the bare support and with the CuO bulk profile has been revealed as very useful for a first approach into the understanding of the synergies created on this system. Finally, the evidences provided by the reducibility and catalytic profiles of the catalysts prepared by the different procedures can contribute, interestingly, to the clues about the features which should be determinant for these catalyst to present very high catalytic activity towards CO oxidation, mainly at low temperatures. J.C. Martínez-Munuera: Conceptualization, Methodology, Investigation, Validation, Visualization, Writing - original draft. V.M. Serrano-Martínez: Investigation. J. Giménez-Mañogil: Investigation. M.P. Yeste: Investigation, Validation, Visualization. A. García-García: Conceptualization, Methodology, Writing - original draft, 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.The authors gratefully acknowledge the financial support of Generalitat Valenciana (PROMETEO/2018/076 project) and the Spanish Ministry of Science and Innovation (PID2019-105542RB-I00 project) and the UE-FEDER funding. Martínez-Munuera also acknowledges Spanish Ministry of Science and Innovation for the financial support through a FPU grant (FPU17/00603).

The aim of this research is an attempt to shed some light on the understanding of the nature of the active sites and the generated synergies in the copper/ceria-zirconia formulations for low temperature CO oxidation by means of the creation of copper entities with different physico-chemical nature. For this reason, several CuOx/ceria-zirconia catalysts, with different Cu contents and different methods to incorporate copper species, were synthesized. Focus was specially put in this case trying to link the results of CO oxidation catalytic tests with the CO-temperature programmed reduction profiles/approximate estimations and selected characterization parameters in order to find out correlations among catalysts’ properties/reducibility and catalytic behaviors, especially those corresponding to the nature and roles of the different CuOx species in contact with ceria-based support on catalytic activity. Results reveal a significant improvement in CO conversion compared to the ceria-zirconia support by adding a small amount of copper loading (as low as 0.5 %), emphasizing the paramount role of copper incorporated by the method of IWI. From 0.5 up to 2% of copper loading, an interesting increase gradual trend in activity and reducibility can be noted. It should be mentioned that all the catalysts obtained by this procedure are more catalytically active towards CO oxidation than 1%Pt/Al2O3 at low temperatures (T < 130 °C). CO-TPR results show that the reducibility of these catalysts is in line with their CO oxidation activity. The method of preparation has been revealed as a critical variable in the catalytic performance, and quite similar catalytic activities can be reached from different synthesis methods and different copper contents, due to the similar nature and type of CuOx species generated over the catalysts’ surface, identified by the CO-TPR profiles and the rest of characterization data. Finally, IWI method seems to be the best one among those tested, thus combining superior areas of both α and β contributions assigned on CO-TPR profiles, which seem to be critical in the interpretation of the catalytic behaviors.

Al–Ni reactive materials, including composite particles and multilayered foils, are a class of energetic materials (EMs) with high-energy content [1–3]. These materials can undergo intermetallic reaction with a significant amount of heat release and forming composites with high mechanical strength [4,5]. Due to this promising feature, Al–Ni reactive materials have been widely used in various energetic applications such as reactive fragments for warhead [6], reactive shaped charge liner [7], insensitive penetrator based on nano-structured EMs [8,9], However, Al–Ni mixtures are not able to be readily ignited especially when their sizes are in micron scale. The relatively low exothermicity of reactions between metallic reactants may results in a high energy barrier for reliable ignition. In addition, the naturally formed oxide shell on the surface of Al particles is likely to lead to the lowered reactivity between the Al and Ni, thereby limiting its broader application [10–12]. In order to improve the ignition and energy release efficiency of Al–Ni, the novel design and relevant advanced preparation of Al–Ni materials have shown to be promising according to recent literature summarized as follows.Various methods used for the preparation of Al–Ni reactive materials with lower ignition threshold and higher combustion performance have been proposed [13–18]. For instance, Hadjiafxenti et al. produced Al–Ni nanocomposite powders by low energy ball milling (LEBM), which exhibit a lower ignition temperature below 600 K [19–21]. The increased reactivity of Al–Ni nanocomposites produced by ball milling is likely due to the fact that the nano-Ni was embedded in Al matrix without formation of oxide barriers or intermediate layers [1,21,22]. Except for composite particles, Al–Ni foils with varied bilayer thicknesses could be fabricated at different atomic ratios by using sputtering method. The onset temperatures of those multilayered foils are below 800 K, higher than its nanoparticle style [23]. Mukasyan et al. proposed that the combustion wave in the Ni/Al nano-foil appears to be a sequential two-stage process, which involves the chemical and physical exothermic transformations [24]. Gunduz et al. reported another two-stage reaction in their experimental and modeling work, which includes the flame front propagates near the reverse peritectic transformation temperature of Ni2Al3 into NiAl and melts at 1406 K. The reaction continues with the growth of NiAl until the melting temperature of 1911 K [25–27]. The above-mentioned preparation methods may provide Al and Ni with more intimate and high surface area contacts, which are critical to the self-propagation combustion of the solid-state reactions. In this way, reactions that normally require high heat input for initiation can be realized at lower temperatures.In addition to the preparation methods that were used to reduce ignition threshold and improve the combustion performance of Al–Ni by increasing the intimate contact, the additives can also be used to enhance the reactivity of Al–Ni by accelerating of the reaction rate and reducing the agglomeration. Researchers have managed to use metallic additives to prepare Al–Ni/M (M: molybdenum, copper and magnesium). The initial temperatures of Al–Ni were found to be increased to various levels depending on the type of metals used. Among them, Cu shows the most significant effect on the combustion process of Al–Ni with a remarkable increase in the flame temperature from 2000 K to 3000 K [28]. For fluoropolymers, the addition of PTFE to Al–Ni may decrease the critical shock pressure for initiation of shock-induced chemical reaction, due to a lowered the apparent activation energy and increased the chemical reaction efficiency of Al–Ni in the presence of fluorine as a highly oxidative element. In particular, with the addition of PTFE, the pre-ignition reaction (PIR) occurs between Al2O3 and PTFE, so that the heat release from PIR reaction plays a positive role in the promotion of intermetallic reaction between Al and Ni [11]. Besides, the agglomeration was greatly reduced due to generation of gaseous product (AlF3) during PIR, so that the reaction efficiency may also be enhanced. In addition to PTFE, other types of fluorine-containing polymers could also react with Al2O3 passivation layer that facilitates the exposure of the active Al, thereby promoting the reactivity [29–31].Besides fluoropolymers, the same effect can be observed for the transition metals used as coating agents. A thin Ni coating layer on the surface of Al particle was shown that the agglomeration of the CCPs would be prevented. In this way, the ignition temperature of Al was reduced to 760–950 K and the front velocity was increased by a factor of 4 as compared to the unmodified ones [2,32]. The ignition mechanism was found to be correlated directly with intermetallic exothermic reactions between Al and Ni [2,33,34]. It was explained that the improvement in combustion performance is a result of the cracking Al2O3 shell due to thermal stress that promotes ignition of Al. In order to further improve the combustion performances of Al–Ni, it would be desirable if the coating agents can react with the passivation layer of Al and forming gaseous products. Except for fluoropolymer, the coating of halogen-containing EMs on the surface of Al–Ni has the great potential to meet both requirements.It has been shown that coating modification with halogen-containing oxidants may lead to a significant enhancement of ignition and combustion by improving the reactivity of Al [35–38]. In addition, the coating technique has a certain positive effect on improvement of the mechanical properties of Al-based composites, which is, however, not in the scope of this study. Our group have conducted various investigations on the coating of Si with optimized ECs. As typical examples, the Si@PVDF/CL-20 (PVDF/CL-20 with the mass ratio of 1:6), Si@AP/NC, (AP/NC with the mass ratio of 2:1) composites with a core-shell structure have been successfully synthesized by using spray-drying technology [39,40]. The results showed that Si@ECs can undergo a more complete reaction between Si and the decomposition products of ECs during the combustion process. The ECs have relatively lower ignition threshold, higher reactivity and better stability. Those ECs could be an appropriate candidate for enhancing the combustion performance of Al–Ni. Therefore, it is of great interest to investigate the effect of energetic composites AP/NC and PVDF/CL-20 on the combustion performance of Al–Ni.In this work, the Al–Ni@ECs at various Al–Ni atomic ratios have been prepared by arrested high energy ball milling (AHEBM) followed with spray-drying technique. The morphologies and compositions of prepared composites and the CCPs were characterized by scanning electron microscopy (SEM) and X-ray diffraction technique, respectively. The thermal reactivity and combustion performances of Al–Ni@ECs composites were evaluated by thermal analyses and customized combustion diagnostic system. The thermodynamic calculation of the full-range chemical equilibrium of the Al–Ni@ECs was conducted by HSC software as a supporting information to elucidate the mechanisms of the enhanced combustion.The micro-spherical aluminum powder (μ-Al) with an average diameter of 1 μm, nano-spherical nickel particles (n-Ni) with a mean diameter of 100 nm, acetone, and dimethylformamide (DMF) were purchased from Sigma-Aldrich company. NC with the nitrogen content of 12.6 wt%, AP (≥99.5%), CL-20 (≥99.5%), PVDF (≥99.9%) were supplied by Xi’an Modern Chemistry Research Institute.In order to obtain the maximized energy output, the ECs including AP/NC (NA) and PVDF/CL-20 (PC) were optimized at mass ratio of 2:1 and 1:6, respectively. The maximum energy release of NA and PC were 5919 J/g and 3536 J/g, respectively [39,40]. Coating of ECs on Al–Ni particles were prepared by spray-drying technique. For the practical applications, it is beneficial to use the least mass content of ECs in the reaction mixture, which is capable of yielding a sufficiently low ignition temperature and initiating the rapid self-propagating combustion. To achieve a thermo-chemical activation mode, the smallest quantity of ECs used for the activation of Al–Ni in this system that was experimentally obtained is 10 wt%. The process for preparation of Al–Ni@ECs composites is illustrated in Fig. 1 .Ball milling time is the major controlled factor variable in preparation of Al/Ni and Al/3Ni composites at molar ratios of 1:1 and 1:3, respectively. It has been shown that the heat release of Al/Ni starts to decrease when the milling time increases up to 2 h, whereas it is 6 h for Al/3Ni (more details are shown in Section 3.1). Therefore, the milling times used for preparation of Al/Ni and Al/3Ni are fixed to be 2 h and 6 h, respectively. The preparation of Al/3Ni reactive composite is briefly introduced as follows: taking Al/Ni as an example, 3.15 g Al and 6.85 g Ni powders are milled in a 250 mL stainless steel jar by using a planetary ball milling facility (XQM-2-DW, China) for 2 h. The rotation speed was 300 rpm, and the diameter of stainless steel ball is 5 mm. The mass ratio of ball to powder was 10:1. In this process, 20 mL mixed solution of DMF and acetone with the volume ratio of 4:1 was used as the processing media. It is also the case for preparation of Al/3Ni, but the only difference is that the milling time was increased to 6 h.Afterwards, 10 wt% of ECs as the coating agents are introduced to two typical Al–Ni composites as shown in Table 1 . The uniformly ball milled Al–Ni reactive materials are firstly collected into a beaker, and then 10 wt% of ECs with 50 mL mixed solution are added. The resulted precursor solution was stirred for 2 h to ensure sufficient dispersion of the Al–Ni powders. Finally, the precursor solution was spray-dried to obtain the final composite coated products. The parameters for the spray-drying process are as follows: diameter of feed well is 1 mm, and fluid flow rate is 3 mL/min. The inlet and outlet temperature are kept at 170 °C and 110 °C, respectively.The spray-dried powders are enclosed in a cylindrical mould with the internal dimension Ф 10 × 45 mm2 and uni-axially pressed at a pressure of 2 MPa for 5 min to make a dense sample. It was then placed in the sealed chamber with pressurized Ar for the combustion diagnosis. The CCPs were collected from DSC experiments and then characterized in terms of phase compositions and morphologies by using XRD and SEM techniques. The details of such characterizations are provided in the Supporting Information.The DSC experiments for all involved Al–Ni composites with milling time have been conducted, and the resulted heat flow curves as a function of temperature are shown in Fig. 2 .As shown in Fig. 2(a), there are two exothermic peaks at 579.9 °C and 618.1 °C for Al/Ni composite, when it was prepared by ball milling of 0.5 h. However, the second exothermic peak disappeared when the ball milling time was in the range of 1 –6 h. Interestingly, the second exothermic peak appears again for the composite after ball milling of 9 h. Moreover, the peak temperature of the exotherm for Al/Ni composite decreases from 579.9 to 562.6 °C as the ball milling time increases. However, when the milling time further increases (e.g. to 9 h), the exothermic peak shifts to a higher temperature, and it is also the case for Al/3Ni (shown in Fig. 2(b)).The detailed DSC parameters obtained for Al/Ni and Al/3Ni composites are summarized in Table 2 . It can be noticed that the measured heat flows (Q) increase first and then decrease with the increase of ball-milling time for both composites. There is an appropriate milling time, when the Al and Ni is homogeneously mixed and well contacted, so that the initial reaction temperature reach the lowest point. It also suggests that excessive ball milling leads to the partial intermetallic reaction between Al and Ni, so that the reactivity would be decreased and the measured heat of reaction becomes lower. According to the heat of reaction, the appropriate ball-milling time for Al/Ni should be about 2 h (e.g. 832.0 J/g). In comparison, the best ball-milling time for Al/3Ni could be around 6 h, where the maximum heat release was 598.8 J/g, lower than that of Al/Ni.The morphologies of Al, Ni, Al–Ni and Al/Ni@ECs are shown in Fig. 3 . For Al and Ni as the starting materials (Fig. 3(a) and Fig. 3(b)), their surfaces are neat and smooth. In case of the ball-milled Al–Ni composites, the surface of Al particle (Fig. 3(c) and Fig. 3(d)) becomes uneven, where the n-Ni particles are randomly distributed. The surfaces of Al/Ni@NA and Al/Ni@PC composites show a relatively rough morphology (Fig. 3(e) and Fig. 3(f)). The initial particle size of Al is about 2 μm, and it increases a little once Ni is covered. Once ECs is included, the Al/3Ni or Al/Ni seem aggregated to about 3–5 μm (Fig. 3(e) and Fig. 3(f)). Meantime, the surface of these large particles is defected with lots of pores, showing increased specific surface areas.The obtained element mapping results for Al/Ni@NA and Al/Ni@PC are shown in Fig. 3(g) and Fig. 3(h), where the distributions of Cl/N/O and F/N/O illustrate the AP/NC and PVDF/CL-20 are uniformly coated on the surface of assembled aggregated Al/Ni particles.The DSC experiments have been performed to investigate the thermal behaviors of Al–Ni@ECs composites and the corresponding heat flow curves are plotted in Fig. 4 .As shown in Fig. 4(a), the first exothermic peak of AP/NC at 213.2 °C is mainly due to the decomposition of NC, and the followed endothermic peak at 246.8 °C is due to the polymorphic transition of AP from orthorhombic to cubic form [41,42]. The second exothermic peak of AP/NC is due to low temperature decomposition of AP with release of NH3 and HClO4 intermediates, but this peak temperature is slightly lower than that of AP (299.3 vs. 312.9 °C). In addition, the third exothermic peak of AP/NC at 386.3 °C could be a result of the high-temperature decomposition of AP, where the oxidation of absorbed NH3 or NH4 + by·ClO3 radical [42]. Obviously, this decomposition process of AP is accelerated, likely attributed to the catalytic effect by the condensed decomposition products of NC (e.g., hydrocarbon chains) [43]. For PVDF/CL-20, the first endothermic peak representing the melting process of PVDF becomes very weak at about 202.1 °C due to relative lower content, as compared to that of pure PVDF at 164.6 °C. It means that there might be interaction between F and NO2 group, which may stabilize both components. Therefore, the first exothermic peak of PVDF/CL-20 at 238.2 °C is due to the decomposition of CL-20, which is slightly higher than that of pure CL-20. The second exothermic peak caused by decomposition of PVDF is shifted from 503.4 to 496.6 °C, due to the catalytic effect by condensed decomposition product of CL-20 as the case for NC towards AP shown above [43].An endothermic peak at 661.1 °C for pure Al can be clearly seen in Fig. 4(b), which is corresponding to the melting of Al [43]. Exothermic peaks displayed at 572.2 °C and 601.9 °C are observed for Al/Ni and Al/3Ni, respectively. It reveals that ECs decomposes much earlier than the intermetallic reaction. Those two exothermic peaks are attributed to formation of AlNi and AlNi3, respectively [17,19]. The detailed reaction mechanisms of Al–Ni are discussed in the following Section 3.3.2.As the coating layer of Al/Ni, the first exothermic peak of AP/NC appears approximately at 205.4 °C, which is surely caused by the decomposition of NC [44] and it is 7.8 °C lower than that of pristine AP/NC. The following exothermic peak at 303.6 °C is due to decomposition of AP. In presence of Al/Ni, the two decomposition peaks of AP are merged into one, and the peak temperature is 9.3 °C lower than that of the second T p of AP. The third exothermic step associated with a small peak at 506.7 °C may be caused by the reaction between the acidic condensed products of AP and the Al2O3 passivation layer, with heat release of 116.9 J/g [45]. The final exothermic process should be attributed to the intermetallic reaction between Al and Ni, which has a peak at around 573.9 °C, the heat release is 771.0 J/g and the condensed alloy product was confirmed by XRD spectrum (see in Section 3.3.2). The total energy release of thermit and intermetalllic reactions of Al is 887.9 J/g, which is improved by 6.7% in comparison with pure Al/Ni. A very similar exothermic reaction process is shown for Al/3Ni@NA composite, the energy release of the Al-related reaction was 860.8 J/g, which was 28.8 J/g higher than Al/Ni. Obviously, the energy release of Al–Ni was increased when it is coated with ECs, the increased energy is due to the preheating and coupling effects of ECs thermolysis. In comparison, for Al/Ni@PC, the first exothermic peak at around 237.9 °C can be assigned to CL-20 thermolysis, which is only 0.3 °C lower than that of pristine PC. The second small exothermic peak at around 418.8 °C was due to the pre-ignition reaction (PIR) between PVDF and Al2O3 passivation layer [46]. It is followed by the main intermetallic exothermic reaction with a peak at 596.5 °C. It is also the case for Al/3Ni@PC, where the first two exothermic peaks at 233.4 °C and 465.4 °C are due to the decomposition of CL-20 and the PIR reaction, respectively. The third and last reaction steps are partially overlapped with a final peak at 628.5 °C, respectively. Such a two-step exothermic pattern is covered with a heat release of 507.0 J/g, which is 353.8 J/g lower than that of Al/Ni@PC. The energy release of Al-related reaction of Al/3Ni@ECs is lower than that of Al/3Ni, which maybe due to the reduced reaction rate between Al/3Ni and the condensed products of ECs, so that part of the heat release is covered by the baseline and overlooked.The above results imply that the improved reactivity of Al–Ni@ECs could be due to a synergistic effect, where Al–Ni catalyzes the decomposition of ECs and the heat release of Al–Ni is promoted in presence of condensed thermolysis products of ECs, especially when the products are acidic and could easily react with the oxide layer of Al and Ni.In order to understand the condensed phase reaction mechanisms of Al–Ni@ECs composites, the XRD was implemented to identify all possible intermediate products quenched at different temperatures from DSC experiments. The obtained XRD spectra are shown in Fig. 5 . As expected, a strong diffraction peak of intermetallic compound is detected for the product collected at 800 °C, but the diffraction patterns are different depending on the types of ECs and atomic ratio of Al–Ni.As shown in Fig. 5(a), only the diffraction peaks of Al and Ni can be observed for the Al/Ni at room temperature. At 620 °C, the diffraction peaks of AlNi, Al3Ni2, Al0.9Ni4.22, and unreacted Al and Ni are shown together. At further elevated temperature of 800 °C, the diffraction peaks of unreacted Al and Ni disappear and the relative diffraction intensities of AlNi and Al0.9Ni4.22 are increased as well. Additionally, a new diffraction peak of Al3Ni is observed at this temperature. The results indicate that the intermetallic exothermic reactions between Al and Ni occurs in between 620 °C and 800 °C for Al/Ni.For Al/Ni@NA (Fig. 5(a) and Fig. S1), the diffraction peaks of Al, Ni, and AP can be clearly seen in the XRD pattern at room temperature. At 520 °C, only the diffractions of Al and Ni are observed. It indicates that AP completely decomposed at this temperature, which is consistent with the DSC results. At 800 °C, the diffraction peaks are dominated by the intermetallic phase of AlNi, suggesting the intermetallic reaction between Al and Ni dominates at this temperature. For Al/Ni@PC, the XRD patterns are almost identical to those of Al/Ni@NA at each attempted temperature.For Al/3Ni (Fig. 5(b)), when the temperature elevates to 800 °C, the phase compositions of the final condensed product are dominated by AlNi3 and which is contaminated with small amounts of Al3Ni2 and AlNi. With the addition of ECs, a new phase of AlN is formed at 800 °C for Al/3Ni@ECs implying that the reaction takes place between N element from ECs and Al. This reaction partially consumes the Al, which is supposed to participate in the intermetallic reaction later, and thereby resulting in decreased practical reacting ratio between Al and Ni.The maximum energy releases of Al–Ni@ECs during the combustion process have been measured by using a bomb calorimeter. The heats of reaction are shown in Fig. 6 and summarized in Table S1.The variation in the measured heat of reaction for Al–Ni@ECs composites is shown in Fig. 6. The highest energy generation is achieved by Al/Ni@NA, indicating that a higher combustion efficiency of Al/Ni could be realized by a minor use of NA. Compared with Al/Ni, the reaction heat of Al/Ni@NA and Al/Ni@PC are increased by 108% and 53%, respectively. It is strange that pure Al/3Ni cannot be easily ignited under 3 MPa of Ar. However, with the introducing of ECs, the ignition of Al/3Ni@ECs composites can be easily achieved. Herein, the theoretical energy release value of Al/3Ni (1230 J/g) is used as reported by Fischer et al. [47]. to assess the effect of coating ECs on Al/3Ni. Compared with the theoretical energy release value of pure Al/3Ni, the reaction heat of Al/3Ni@NA was increased by 42%, whereas that of Al/3Ni@PC was increased by 22%. Obviously, the energy release of Al–Ni was greatly improved with the inclusion of ECs.Meantime, the reaction heat of Al/Ni@ECs is higher than that of Al/3Ni@ECs, suggesting that Al–Ni with the atomic ratio of 1:1 has higher energy content than that of 1:3. The results indicate that both the atomic ratio of Al–Ni and the type of ECs have significant effects on the energy content and release rate of Al–Ni. The Al–Ni with the inclusion of PC seem to be less reactive as compared to NA. The possible reason could be related to the relative low exothermicity from the interfacial reaction between PC and Al–Ni. In fact, the reactivity of Al/CuO nanothermite composites with fluoropolymers has been measured with similar findings [48]. They showed that hydrogen fluoride (HF) released from PVDF may react with Al/CuO, resulting in less energy due to higher heat of formation of fluorides than oxides. This less exothermic reaction may be responsible for the lower combustion rate and less heat release. Therefore, it can be concluded that the exothermicity of the reactions between ECs and Al–Ni plays a critical role in tunning energy contents and heat release rates of Al–Ni the involved composites.The combustion behaviors of Al–Ni@ECs composites have been studied using our customized combustion diagnostic system. The sequential snapshots of all samples burning in Ar have been done using a high-speed camera through a transparent window (Fig. 7 and Fig. S2). All the samples except Al/3Ni were successfully ignited and proceeded to a self-sustainable combustion. The low ignitability of Al/3Ni is likely due to high chemical stability of Ni, especially when the fraction of Ni in the Al/3Ni composite is surpassing a certain threshold. The difficulty in ignition suggests that this composite at atomic ratio of 1:3 is insensitive to heat. Thus, the reactivity of the Al/3Ni@ECs may also be relatively lower compared to that of Al/Ni@ECs samples.For Al/Ni, the light emission of the burned sample lasts for ∼2 s, which mainly involves the intermetallic reaction between Al and Ni. When the intermetallic reaction was completed, the brightness of the burned Al/Ni gradually reduces during cooling process. When it was coated with NA, the flame propagation process of Al–Ni@NA presents significant visible spots as well as irregular cracks on the surface of burned samples as shown in Fig. 7(b) and d. At the same time, an axial elongation of Al–Ni@NA increases with the increase of burn time. The same phenomenon was shown in the case of Al/3Ni@NA. Such a behavior could be attributed to the effect of a large number of gaseous products (e.g., HF, NO and CH4) released from ECs, which were ejected from the sample surface and generating porous structures in the residues. Furthermore, the violent combustion reactions were observed for Al–Ni@NA, indicating that the reactivity of intermetallic reaction between Al and Ni has been greatly enhanced with the inclusion of NA. The flame front of Al/Ni@NA takes about ∼450 ms to reach the bottom of charge, since the flame propagation rate of NA is much faster than PVDF/CL-20. Moreover, the self-sustained combustion rate of Al/Ni@PC is smaller than that of Al/3Ni@PC, which is probably caused by the less exothermic reaction between the condensed phase products decomposed from PC and Al–Ni.To further investigate the combustion behaviors of the prepared composites, the flame propagation rates have been calculated based on the recorded images by using high-speed camera in Fig. S2, and the results are summarized in Table 3 .From Table 3, it can be seen that the flame propagation rate of Al–Ni was increased with the inclusion of ECs. The flame propagation velocity of Al/Ni is 15.8 mm/s. In case of Al/Ni@NA, the flame propagation rate was increased by 30.0% from 15.82 mm/s to 20.6 mm/s under the effect of the NA coating. The same positive effect of PC coating on Al/Ni is obtained, where the flame propagation rate of Al/Ni@PC was 11.6% higher than pristine Al/Ni. For Al/3Ni@ECs composites, the propagation rate is nevertheless reduced with the increase of Ni. It suggests that despite Ni contributes to catalytic effect on ECs, excessive Ni would reduce the overall energy content in comparison to Al, thereby reduce the flame propagation rate.Combustion wave temperature usually demonstrate the efficiency of the heat generation and the heat capacity of the combustion products. The temperature distribution and its dependence on burn time have been obtained by using the high-speed infrared camera. The recorded thermal images are displayed in Fig. S3, that allows one to preliminarily judge the relative difference in surface temperature distribution for the involved samples, the actual temperature could be much different, since the radiation parameter is very difficult to be accurately obtained. Anyway, these values are meaningful for a systematically comparing the effect of ECs on the improving the reactivity of intermetallic reaction of Al–Ni system.As shown in Fig. S3, the maximum flame temperature of Al/Ni@ECs is about 1800 K, which is ∼500 K higher than that of Al/3Ni@ECs with less energy content. The bright red and blue vapors indicate that large amounts of gases are produced by ECs during their combustion processes. It is clear that Al/Ni@ECs composites release more gaseous products than Al/3Ni@ECs. The fast heat generation by gaseous production of ECs could be absorbed by Al–Ni for preheating, which has positive affect on the combustion efficiency/rate of Al–Ni. Furthermore, the secondary reaction of Al/Ni@ECs as depicted in Fig. S3(b) and Fig. S3(c) during the combustion process is featured with a secondary temperature rise of the burned sample, which was not observed for Al/3Ni@ECs. The secondary heat release process of Al/Ni@ECs is mainly dominated by the intermetallic reaction, which is greatly enhanced by the presence of the decomposed gaseous products by ECs as the catalysts or reactive sites, where porous structure has been formed. The absence of this phenomenon of Al/3Ni@ECs might be responsible for the relatively low reaction temperature change. It further confirms that an enhanced reactivity and higher combustion efficiency can be obtained for Al/Ni with an atomic ratio of 1:1.For a better comparative analysis, the average combustion wave temperature of Al–Ni@ECs have been evaluated. The corresponding temperature vs. time profiles and its derivatives are shown in Fig. 8 (a) and Fig. S3(b), respectively. All the curves are shifted manually along the X-axis in order to avoid overlapping. In general, the average temperatures of the composites with ECs coating are several hundred Celsius higher than that of the reference sample of Al/Ni. Moreover, an average maximum combustion temperature (T max) obtained for both Al/Ni@ECs composites are higher than that of Al/3Ni counterparts.The temperature rise rate (γ t), serves as the parameter indicating the intensity and efficiency of self-sustained combustion. It is affected by the gases production rate, thermite reaction rate and intermetallic reaction rate during the combustion process. Fig. 8(b) shows the temperature rise rate derived for Al–Ni@ECs composites, where γ t is basically increased with the addition of ECs regardless of their types. For the best scenario observed, it is over 11 times higher than that of the reference sample. Furthermore, the temperature rise rate of Al/3Ni@ECs is relatively lower than that of Al/Ni@ECs, which can also be supported by the fact of the lower exothermicity of Al/3Ni.The morphology and composition of the CCPs were characterized aiming to reveal the combustion reaction mechanisms. The CCPs of Al–Ni@ECs composites were collected for SEM, EDS and XRD analyses and the results are shown in Fig. 9, Fig. S4 and Fig. 10 . Compared with Al/Ni shown in Fig. S4(a), the holes of the CCPs from Al–Ni@ECs (Fig. 9, Fig. S4(b) and Fig. S4(c)) are increased. It can be obtained that the gaseous products of Al–Ni@ECs are increased in comparison to that of Al/Ni without the ECs coating. The gaseous products on the interface layer can exclude the sintering and form lots of pores as literature reported [37]. Those pores also provide new channels for the further reaction between Al–Ni and the condensed products of ECs.The surface of the CCPs from Al/Ni@PC contains many hollow spheres with varied diameters ranging from 5 to 200 μm as shown in Fig. 9(a). According to the previous studies, the hollow structure provides the penetrating channels for better heat and mass transfer, which leads to an enhanced intermetallic reaction rate of Al/Ni. The CCPs of Al/Ni@PC were examined by powder XRD techniques, where AlNi and a small amount of Al5O6N (Fig. 10) were discovered, indicating that the reaction between the gaseous products of ECs and Al occurred before intermetallic reaction of Al and Ni.For Al/3Ni@PC, a plenty of cubic crystals with smooth surfaces can be easily observed in their CCPs as shown in Fig. 9. The EDS results show that the element of fluorine is dominated in the crystals, suggesting that they are F-containing compounds. As mentioned above, the reactivity of Al/3Ni@PC is relatively lower, which is responsible for the low reaction temperature as well as low burning rate of Al/3Ni@PC. Thus, the lowered reaction temperature and reduced burning rate may collectively be in favor of the growth of this F-containing crystal. However, such a crystal is unknown in the chemical database, and the specific formation mechanism needs to be further studied. Moreover, the CCPs of Al/3Ni@PC display a less porous structure in comparison to that of Al/Ni@PC. Since the formation of F-containing crystals would reduce the availability of F, so that the gaseous products are consequently reduced. The reduced gaseous products have a detrimental effect on the formation of porous structure, and thereby it significantly decreases the reaction rate of Al/3Ni. The CCPs of Al/3Ni@PC mainly contain AlNi3, a small amount of AlN and Al5O6N (Fig. 10). The presence of some other unknown diffraction peaks in the CCPs of Al/3Ni@PC may be associated with the F-containing cubic crystals.In order to further understand the reaction process, the equilibrium compositions of Al/Ni@NA at different temperatures were calculated by using HSC software. Fig. 11 shows the equilibrated species of Al/Ni@NA vs. temperature. The quantities of HCl and AlNi decrease with the increase of temperature due to the formation of AlCl, suggesting that the reaction between HCl released from AP and shell of Al has occurred, which is beneficial to improve the combustion performance of the composite [49]. With the increase of AlCl content, the AlO appears at higher temperature. The AlO is a primary combustion intermediate of Al, indicating that the reaction between ECs and Al occurs after the etching of Al2O3 shell. Furthermore, the formation of Al5O6N determined by XRD analysis further confirms that the reaction is most likely to occur between gaseous products of ECs and Al. In addition to the CCPs of AlNi, the calculation also shows that gaseous species such as NH3, Ni, and AlCl are generated during the combustion process. The formation of gaseous substances would greatly change the reaction pathways, by shifting it from a solid-solid to solid-gas or even liquid-gas modes. The combustion performance was significantly improved under the combined combustion mode [50].According to the above results, the overall reaction processes of Al–Ni@ECs are proposed as follows. First, ECs decompose and ignited with combustion products uniformly cover the particle of Al–Ni, and then the products or intermediates of ECs would react with preheated Al. The fast heating by ECs combustion can promote the flame propagation rate of Al–Ni. Additionally, the acidic products such as HCl and HF generated from ECs may etch the Al2O3 shell, so that the inner active Al to be exposed and easily react with Ni. Therefore, more efficient energetic systems can be obtained by introducing these halogen containing ECs.In this work, two types of ECs (NA and PC) were used to coat Al–Ni reactive materials by ball milling followed with spray-drying technique. The effects of ECs on the heat release, combustion characteristics of Al–Ni reactive materials, morphologies and compositions of the corresponding combustion products have been comprehensively studied. It has been shown that the combustion performances of those composites are greatly affected by introducing different types of halogen-containing ECs as coating agents. Particularly, the combustion performance of Al–Ni can be significantly improved by coating of NA. The flame propagation rate was increased from 15.8 mm/s to 20.6 mm/s, which was 30.0% higher than that of the reference. In addition, the combustion wave temperature of the corresponding surface modified composites was ∼500 K higher than that of the reference without surface modifications. The acidic gaseous products decomposed from halogen-containing energetic composites can react with Al2O3 passivation layer, which make the inner active Al to be exposed and easily react with Ni. Therefore, the intermetallic reaction between Al and Ni was greatly enhanced.These results presented in this paper demonstrate that the halogen-containing energetic composites are the promising candidate for tuning the reactivity and combustion characteristics of the reactive intermetallic materials. Further efforts can be made on the clarification of the detailed combustion mechanisms of such composites with advanced techniques such as Time of Flight/Mass Spectroscopy (TOFMS).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 Nation Natural Science Foundation of China (Grant No. 51776176) and the Fundamental Research Funds for the Central Universities, China (Grant No. G2017KY0301). This paper was also partially funded by NSAF project (Grant No.2030202) and sponsored by Innovation Foundation for Doctor Dissertation of Northwestern Polytechnical University (Grant No. CX2021048).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.dt.2022.01.007.

In this paper, various core-shell structured Al–Ni@ECs composites have been prepared by a spray-drying technique. The involved ECs refer to the energetic composites (ECs) of ammonium perchlorate/nitrocellulose (AP/NC, NA) and polyvinylidene fluoride/hexanitrohexaazaisowurtzitane (PVDF/CL-20, PC). Two Al–Ni mixtures were prepared at atomic ratios of 1:1 and 1:3 and named as Al/Ni and Al/3Ni, respectively. The thermal reactivity and combustion behaviors of Al–Ni@ECs composites have been comprehensively investigated. Results showed that the reactivity and combustion performance of Al–Ni could be enhanced by introducing both NA and PC energetic composites. Among which the Al/Ni@NA composite exhibited higher reactivity and improved combustion performance. The measured flame propagation rate (v = 20.6 mm/s), average combustion wave temperature (T max = 1567.0 °C) and maximum temperature rise rate (γt = 1633.6 °C/s) of Al/Ni@NA are higher than that of the Al/Ni (v = 15.8 mm/s, T max = 858.0 °C, and γt = 143.5 °C/s). The enhancement in combustion properties could be due to presence of the acidic gaseous products from ECs, which could etch the Al2O3 shell on the surface of Al particles, and make the inner active Al to be easier transported, so that an intimate and faster intermetallic reaction between Al and Ni would be realized. Furthermore, the morphologies and chemical compositions of the condensed combustion products (CCPs) of Al–Ni@ECs composites were found to be different depending on the types of ECs. The compositions of CCPs are dominated with the Al–Ni intermetallics, combining with a trace amount of Al5O6N and Al2O3.