Catalytic Hydrogenation of CO 2 to Methanol: A Review

: High-efﬁciency utilization of CO 2 facilitates the reduction of CO 2 concentration in the global atmosphere and hence the alleviation of the greenhouse effect. The catalytic hydrogenation of CO 2 to produce value-added chemicals exhibits attractive prospects by potentially building energy recycling loops. Particularly, methanol is one of the practically important objective products, and the catalytic hydrogenation of CO 2 to synthesize methanol has been extensively studied. In this review, we focus on some basic concepts on CO 2 activation, the recent research advances in the catalytic hydrogenation of CO 2 to methanol, the development of high-performance catalysts, and microscopic insight into the reaction mechanisms. Finally, some thinking on the present research and possible future trend is presented.


Background
CO 2 can be used as condensing agent (dry ice), protective gas for fruit preservation, fizzy gas in carbonated drinks, solvent for homogeneous catalysis (in a supercritical state), and reactant for the mineralization and carbonate-based chemicals [1,2]. In recent decades, booming research interests in CO 2 were stimulated by its greenhouse effect on global climate change [3]. The global atmospheric CO 2 has been speedily rising since the 1950s and has doubled nowadays (212 ppm in 1958 vs. 445 ppm in 2021) [4]. Another doubling was predicted, according to the model of the high emission scenario, to be reached by the year 2100 [5]. The greenhouse effect of atmospheric gases is originated from their absorption of infrared (IR) irradiation from the earth's surface. Any IR-active molecule such as H 2 O, CH 4 , CO 2 , and NO x can be a candidate for greenhouse gases with the ability to absorb photons depending on the dynamic transition dipole and the freedom of vibrations [5]. An exception is H 2 O, of which the natural greenhouse effect has made the earth's temperature appropriate for its biosystem, and equilibrium has been reached because of the huge surface of oceans [5]. Among other greenhouse gases, CO 2 plays the dominant role because of its overwhelming emission related to modern human life [6]. Control of CO 2 emission is urgent to alleviate the greenhouse effect. In recent years, carbon capture and storage (CCS) technologies have been developed to separate CO 2 from either chemical processes or combustion products, followed by recycling or storage [2]. Recently, CO 2 capture directly from the air also caught some scientific and practical interest [7,8]. However, though effectively reducing the emission of CO 2 to some extent, CCS does not provide the final solution because of the high cost and potential risk of leakage. Conversion of the captured CO 2 to value-added chemicals, in this situation, draws the CO 2 back to carbon recycling, providing a sustainable solution [9].
As the highest oxidation state of carbon element, CO 2 shows significant thermodynamic stability with the formation enthalpy of 396 kJ/mol [10]. Therefore, converting CO 2 Figure 1. Walsh diagram of CO2 orbital energies in linear and bent geometries. Reprinted with permission from [1]. Copyright 1996 Elsevier. CO2 molecules can be activated on either metal or oxide surfaces. On the metal surface, electrons in the metal are transferred to physisorbed CO2 resulting in a chemisorbed anionic CO2 δ− which was thought to be the precursor for CO2 dissociation and carbonate formation [29]. This process, however, is often accompanied by high activation barrier with the value depending on the metal's work function [1,30]. Steps [31,32], alkali metal promotions (e.g., K, Na) [33,34], coadsorption (e.g., H, O) [35,36] may facilitate the electron transfer. On oxide surfaces, CO2 adsorption becomes more complicated due to the amphoteric nature of CO2 and the structural and electronic complexity of oxide surfaces. In general, the carbon atom of the CO2 molecule, which is more electropositive, tends to bond to the Lewis basic sites (usually the oxygen sites), and the oxygen atoms of the CO2 molecule to the Lewis acidic sites (usually the cationic sites) [37,38]. As a result, a common adsorbed surface species, besides a physisorbed CO2, is carbonate with the geometry of monodentate, bidentate, or even a tridentate, depending on the surface atomic arrangement [39]. For instance, on an oxygen terminated surface, only monodentate carbonate is formed, while bidentate species is usually obtained on mixed-terminated or defective surfaces [40].
The first step of CO2 activation is typically the adsorption of CO2 on catalyst surfaces which lowers the total energy of the system. The binding energy of CO2 to the surface strongly influences the coverage of CO2 on the catalyst surface, as well as the reaction barrier [41]. Stronger binding of CO2 on the catalyst's surface promotes the surface coverage, but more energy must be injected to activate the intermediate. Therefore, the activation of CO2 makes sense only if a whole reaction is considered. Figure 1. Walsh diagram of CO 2 orbital energies in linear and bent geometries. Reprinted with permission from [1]. Copyright 1996 Elsevier.

Conversion Route of CO 2 to Value-Added Chemicals
The possible catalytic conversion routes are shown in Figure 2, and the representative reactions along with the typical catalysts are summarized in Table 1. The ultimate products are distributed from methane and formic acid to hydrocarbons (light olefins, liquefied petroleum gas (LPG), gasoline, aromatics, etc.). The most conspicuous intermediates are CH 3 OH and CO, owing to the relatively mature technologies of methanol to hydrocarbons (MTH) (including methanol to olefin (MTO), methanol to aromatics (MTA), methanol to gasoline (MTG), etc.) and Fischer Tropsch synthesis (FTS) adopted in the current chemical industry. Methanol often serves as the starting feedstock for several useful chemicals such as gasoline, dimethyl ether, oxymethylene ethers, or as an additive or fuel in engines and fuel cells [42,43]. Recently, methanol was proposed to be the candidate alternative to oil and gas to generate a "methanol economy" [44]. In this situation, the hydrogenation of CO 2 to produce methanol would be one of the key reactions for building energy recycling loops.

Conversion Route of CO2 to Value-Added Chemicals
The possible catalytic conversion routes are shown in Figure 2, and the representative reactions along with the typical catalysts are summarized in Table 1. The ultimate products are distributed from methane and formic acid to hydrocarbons (light olefins, liquefied petroleum gas (LPG), gasoline, aromatics, etc.). The most conspicuous intermediates are CH3OH and CO, owing to the relatively mature technologies of methanol to hydrocarbons (MTH) (including methanol to olefin (MTO), methanol to aromatics (MTA), methanol to gasoline (MTG), etc.) and Fischer Tropsch synthesis (FTS) adopted in the current chemical industry. Methanol often serves as the starting feedstock for several useful chemicals such as gasoline, dimethyl ether, oxymethylene ethers, or as an additive or fuel in engines and fuel cells [42,43]. Recently, methanol was proposed to be the candidate alternative to oil and gas to generate a "methanol economy" [44]. In this situation, the hydrogenation of CO2 to produce methanol would be one of the key reactions for building energy recycling loops.

Recent Advances in Catalyst Research for CO 2 Hydrogenation to Methanol
The main formal difference between the methanol synthesis from CO 2 hydrogenation and CO hydrogenation is the formation of H 2 O vapor in the products, as shown in Table 1. The catalysts for catalytic hydrogenation of CO 2 to methanol along with the recently reported reaction promoters are summarized in Table 2. For Cu/ZnO/Al 2 O 3 catalyst, the presence of water vapor was reported to cause a notable weakening of the stability of the Zn promoter or surface Zn-Cu alloy [61][62][63]. In addition, despite the fact that the WGS reaction is beneficial to the CH 3 OH synthesis from syngas (mainly CO), its reverse reaction converts CO 2 to CO, lowering the yield of methanol [61]. Therefore, a good catalyst for CO 2 hydrogenation to produce methanol must be stable under high water partial pressure and superior against RWGS reaction [64], though water vapor was thought to have a strong influence on the methanol selectivity [65].

Cu-Based Catalysts
Cu-based catalysts, especially Cu/ZnO/Al 2 O 3 (e.g., typical weight composition of CuO:ZnO:Al 2 O 3 being 60:22:8 adopted by ICI Ltd., London, UK), were originally used as the catalyst for industrial methanol synthesis from syngas (CO, CO 2 , and H 2 ) at~250 • C and 50-100 bar [96][97][98]. CO 2 was considered the main carbon source for methanol [99,100], which is possibly a hint that Cu/ZnO catalyst could also be active for methanol synthesis from the hydrogenation of pure CO 2 . Nowadays, Cu-based catalysts are still the hotspots in the research field of CO 2 hydrogenation to produce methanol [101] thanks to their allaround unique performance over others. Despite many controversies in understanding the reaction mechanism of commercial Cu/ZnO/Al 2 O 3 catalysts, progress in catalyst development in recent years has been conspicuous [98,[101][102][103][104][105][106].
The Cu/ZnO catalyst has been thought "a prototype for studying complex promotional interactions in catalysis" [106]. Indeed, both Cu and ZnO have certain activity for methanol synthesis from both CO and CO 2 , while only their mixture shows superior performance [97,107]. The consequence of the mixture can be: (1) gas-dependent morphological changes of the active component; (2) support-induced strain; and (3) strong metal-support interaction (SMSI) effect [106]. The consequence includes well-dispersed and stabilized Cu on ZnO, partially oxidized Cu species, creation of oxygen vacancies [108], and formation of ZnO-Cu interface and Zn-Cu alloy [98,106]. Number of strategies were designed revolving around modifying the intrinsic activity of copper in the catalyst, including development of new synthesis methods [109][110][111][112][113][114][115][116], adoption of various support materials [68,115,117,118], and promoters [87,119,120] to enable new structures and morphologies [68,112,121,122], more homogeneous and stable particles size [123,124], tunable metal-oxide interface [125][126][127][128], etc.

Preparation Methods
A selective synthesis method was adopted by Tada et al. to prepare an amorphous Cu a Zr 1−a O b catalyst [111]. The method was originally based on an incipient wetness impregnation, followed by drying and calcination (see Figure 3). The main point was emphasized by low calcination temperature and low metal loading. The low calcination temperature enables the formation of interfacial sites between metal and support that were supposed to be the active sites for CO 2 -based methanol synthesis. The low metal loading (up to the solubility limit of metal in oxide support) can essentially prevent the formation of crystalline CuO particles and tune the interfacial sites between metal and supports during H 2 reduction. tion mechanism of commercial Cu/ZnO/Al2O3 catalysts, progress in catalyst developm in recent years has been conspicuous [98,[101][102][103][104][105][106].

Preparation Methods
A selective synthesis method was adopted by Tada et al. to prepare an amorpho CuaZr1-aOb catalyst [111]. The method was originally based on an incipient wetness pregnation, followed by drying and calcination (see Figure 3). The main point was e phasized by low calcination temperature and low metal loading. The low calcination te perature enables the formation of interfacial sites between metal and support that w supposed to be the active sites for CO2-based methanol synthesis. The low metal load (up to the solubility limit of metal in oxide support) can essentially prevent the format of crystalline CuO particles and tune the interfacial sites between metal and supports d ing H2 reduction.  Reprinted with permission from ref. [111]. Copyright 2018 American Chemical Society.
Yu et al. compared the performance of Cu/SiO 2 catalysts prepared by the ammonia evaporation (AE) method and flame spray pyrolysis (FSP) method [112]. The high temperature in the FSP method facilitates high specific surface area and the formation of mixed oxides and is therefore suited for modulating metal-support interactions [129,130]. Specifically, a distorted Cu-O-Si structure formed in the Cu/SiO 2 stabilizes Cu + species, which are closely correlated with the formation of intermediates and methanol. As a result, the FSP sample contains 5 times higher surface Cu + species than in the AE sample, leading to a significant enhancement of methanol selectivity. By comparison, Cu-based catalysts prepared through the aerosol-assisted sol-gel (AASG) process possess rich spherical and mesostructured microparticles (see Figure 4 for a schematic view of the AASG process), achieving highly dispersed open mesopores and promoters (e.g., Zn and Ga) [121]. The active component (Cu) was directly introduced by impregnation or incorporated in the precursor solution, which markedly simplifies the synthesis process. The following calcination leads to the enrichment of dispersed Cu species on the catalyst surface. The catalytic activity of resulted catalysts was ascribed to the formation of the sites that are active for methanol synthesis and inactive for reverse water gas shift reaction. The formation of the active sites was originally promoted by the structure promoter (e.g., Zn and Ga) and by the preparation method as well. Likewise, the precursor phase has a strong influence on the activity of the final catalyst. A Zincian georgeite precursor was adopted to synthesize Cu/ZnO catalysts and yielded a high content of exposed Cu sites and intimated Cu-ZnO interface [110].
sult, the FSP sample contains 5 times higher surface Cu + species than in the AE leading to a significant enhancement of methanol selectivity. By comparison, Cu catalysts prepared through the aerosol-assisted sol-gel (AASG) process posse spherical and mesostructured microparticles (see Figure 4 for a schematic view AASG process), achieving highly dispersed open mesopores and promoters (e.g., Ga) [121]. The active component (Cu) was directly introduced by impregnation o porated in the precursor solution, which markedly simplifies the synthesis proce following calcination leads to the enrichment of dispersed Cu species on the catal face. The catalytic activity of resulted catalysts was ascribed to the formation of t that are active for methanol synthesis and inactive for reverse water gas shift reacti formation of the active sites was originally promoted by the structure promoter ( and Ga) and by the preparation method as well. Likewise, the precursor phas strong influence on the activity of the final catalyst. A Zincian georgeite precurs adopted to synthesize Cu/ZnO catalysts and yielded a high content of exposed C and intimated Cu-ZnO interface [110]. . Schematic view of the AASG process that can be operated in a continuous mod one-pot production of nanostructured catalysts. The precursor solution contains a solven precursors, and a templating agent (e.g., a surfactant that will form micelles via evapor duced self-assembly, EISA). The droplets formed by atomization are transported in a dr towards a drying zone where the formation of structured particles takes place. Reprinted w mission from ref. [121]. Copyright 2020 Wiley.
Varied from the traditional way to stabilize the structure of Cu/ZnO catal compositions, sizes, and geometries of colloidal ZnO and Cu were tuned by the ad of phosphinate ligands during synthesis [113]. As a result, the structure of the ZnO catalyst was stabilized whilst keeping the surface activity, providing a simp ner to regulate the stability and activity of the catalyst. In addition, the particle si face area of Cu, and the strongly basic sites in the Cu/Zn/Al/Zr compound were tu the incorporation of different amounts of fluorine (F) ions [109]. With such an ap the activity and methanol selectivity can be systematically studied as a function o tent. An optimized result was obtained with F/Al = 0.83, where both the methan and methanol selectivity reach the maximum.
The activity of the catalyst is sensitive to the microstructure and morpholog et al. have prepared a Cu/ZnO@m-SiO2 catalyst with a core-shell structure. The m rous pore of silica confined the size of Cu nanoparticles to a narrow size distribu around 5 nm, which enables excellent stability against aggregation and deactiva well as superior catalytic activity [124], in line with the experimental findings by v . Schematic view of the AASG process that can be operated in a continuous mode for the one-pot production of nanostructured catalysts. The precursor solution contains a solvent, metal precursors, and a templating agent (e.g., a surfactant that will form micelles via evaporation-induced self-assembly, EISA). The droplets formed by atomization are transported in a drying gas towards a drying zone where the formation of structured particles takes place. Reprinted with permission from ref. [121]. Copyright 2020 Wiley.
Varied from the traditional way to stabilize the structure of Cu/ZnO catalyst, the compositions, sizes, and geometries of colloidal ZnO and Cu were tuned by the adoption of phosphinate ligands during synthesis [113]. As a result, the structure of the Cu and ZnO catalyst was stabilized whilst keeping the surface activity, providing a simple manner to regulate the stability and activity of the catalyst. In addition, the particle size, surface area of Cu, and the strongly basic sites in the Cu/Zn/Al/Zr compound were tuned by the incorporation of different amounts of fluorine (F) ions [109]. With such an approach, the activity and methanol selectivity can be systematically studied as a function of F content. An optimized result was obtained with F/Al = 0.83, where both the methanol yield and methanol selectivity reach the maximum.
The activity of the catalyst is sensitive to the microstructure and morphology. Yang et al. have prepared a Cu/ZnO@m-SiO 2 catalyst with a core-shell structure. The mesoporous pore of silica confined the size of Cu nanoparticles to a narrow size distribution at around 5 nm, which enables excellent stability against aggregation and deactivation, as well as superior catalytic activity [124], in line with the experimental findings by van den Berg [131]. Additionally, highly dispersed Cu/graphene catalysts were obtained by the adoption of metal-organic frameworks (MOFs) template, showing nice CO 2 hydrogenation activity. This was achieved by introducing MOFs (e.g., HKUST-1) agent into the graphene layers, followed by high-temperature calcination. The ordered structure of MOFs enables the periodic arrangement of Cu nanoparticles, and the sequent pyrolysis of MOF further promotes the formation of ultrafine metal nanoparticles [123].

Role of Supports
The role of supports was originally thought to provide a carrier for well-dispersed metal particles and structural separation against sintering [97,132]. However, the SMSI effect was intensively reported in many supported catalysts having a strong influence on the chemical state, morphology, stability, activity, etc., of the catalysts [132,133]. For CO 2 hydrogenation to methanol catalysts, the frequently studied supports were ZnO [98,103], Al 2 O 3 [97,134], ZrO 2 [118], CeO 2 [117,130], SiO 2 [124], and their mixed oxides [115,135].
For Cu/ZnO catalysts, the deactivation was proposed to follow the oxidation of Cu 0 to Cu 2+ and the agglomeration of ZnO species [136]. To prevent the possible deactivation, measures must be adopted to stabilize the structure of ZnO and the oxidation state of copper, which can be achieved by either regulation of the structural promoter or the addition of a hydrophobic promoter. Other problems of the Cu-based catalyst are the sintering of the Cu component during reaction due to the relatively low Hüttig temperature (407 K) [137,138] and adsorbate-induced dynamic shape changes [139,140]. In this case, Al 2 O 3 was usually adopted as a structural promoter to elevate the stability, as well as the dispersion of active components [138,141]. Comparatively, ZrO 2 appears to work in multiple ways [142]. A parallel comparison of Cu supported on CeO 2 , ZnO, and ZrO 2 showed that the SMSI between Cu and CeO 2 leads to efficient dispersion of CuO and sufficient oxygen vacancies on the oxide support that stabilizes the intermediates (e.g., formate and methoxide) to produce methanol [117]. High coverage of formate species inhibits the RWGS reaction and therefore results in a high methanol selectivity [68]. Besides the SMSI effect, the apparent interactions between nitrogen species and Cu species have some impact on the microstructure of the resulted catalyst. As a result, the Cu-ZrO 2 catalyst supported on N-doped carbon nanotubes (CNTs) exhibits enhanced CuO dispersion that consequently promotes the reduction of Cu and hence the catalytic performance [118].
In addition, because of the good thermal stability and flexibly controlled microstructure, spinel structure materials have attracted research interest to tune the properties of Cu-based catalysts for methanol synthesis [135,[143][144][145]. For Cu supported on ZnFe 2 O 4 spinel catalyst, the size of Cu nanoparticles was adjustable according to the Cu/Zn and Zn/Fe molar ratios. Compared with the other Cu/ZnM-0.5 (M = Co, Ni, Sn, Cr, Al, Ga, and Mn) catalysts prepared with the same procedures, Cu/ZnFe-0.5 exhibits better methanol synthesis performance. The optimized 33Cu/ZnFe-0.5 catalyst exhibits a high methanol selectivity of 71.6% and a CO 2 conversion of 9.4% at 260 • C and 4.5 MPa. Therein, the spinel structure of ZnFe 2 O 4 was thought to tune the formation of small ZnO adjacent to surface Cu NPs as well as the SMSI. The ZnFe spinel structure enhances the adsorption of CO and inhibits the weak and medium-strong CO 2 adsorption, resulting in improved methanol selectivity. The spinel structure catalysts, however, usually have relatively low surface area because of the high calcination temperature during preparation [135]. Combined with the microwave-hydrothermal synthesis method [116,135], a Cu 1−x Zn x Al 2 O 4 precursor with a high specific surface area was obtained. Thanks to the separation of the spinel lattice, very small Cu particles (<4 nm) were obtained with the exposed Cu surface area of 8 m 2 /g. The catalyst also exhibited good structural stability against aggregation in the following reduction process. Similarly, tuning of the microstructure and intrinsic catalytic properties of Cu/ZnAl 2 O 4 was also achieved by adding Cr or Ga [145].

Promoters
The effect of the promoter in the heterogeneous catalyst is structural and electronic in nature [146,147]. For instance, superior methanol selectivity has been achieved by potassium (K) promoted Cu-Zn-Al catalyst (CZA-K) [87]. Compared with sodium (Na), the Cu + /Cu ratio of CZA-K is higher, which enables better CO 2 hydrogenation activity. The formation of surface K-O-(CO)-O species prevents the dissociation of CO 2 , leading to a limited RWGS reaction and higher methanol selectivity. Ga 2 O 3 added to Cu-ZnO/HZSM-5-based catalyst can effectively increase the specific surface area, reduce the size of the metallic component and maintain a relatively high Cu + content, all of which were thought crucial for the hydrogenation of CO 2 [148]. Besides, indium (In) shows a good promotion effect for Cu/CeO 2 catalyst. Loading of 1 wt% indium in Cu/CeO 2 leads to a prominent improvement in methanol yield rate. The promotion effect herein includes smaller Cu particle size and improved dispersion and stability of Cu [119].
With a surface organometallic chemistry (SOMC) approach, Cu supported on SiO 2 with different promoters (Ti, Zr, Hf, Nb, Ta) was successively prepared. Importantly, the prepared catalysts have the same physicochemical properties, in which condition the role of the promoter can be compared in parallel. As the promoters were proposed to modify the Lewis acid strength of interfacial metal sites [120] that correlates with the stability of the reaction intermediates (formate and methoxy) at the periphery of Cu nanoparticles, a promoter with stronger Lewis acid metal sites will enable a better catalytic activity over Cu-based catalyst.
In addition to the exotic promoters, it was also reported that the methanol itself may act as a promoter over Cu/ZnO/MgO catalyst. With the presence of a small amount of methanol in the reactant steam, the apparent activation barrier was notably reduced (from 117.9 kJ/mol to 67.9 kJ/mol) [149]. The origin of the methanol promotion effect was not discussed in the publication but is possibly related to the interaction between methanol and the catalyst with the self-limiting activity expected.

In 2 O 3 -Based Catalysts
In 2 O 3 was initially reported to show excellent catalytic activity for methanol steam reforming with extremely high selectivity of CO 2 relative to that of Cu/ZnO catalysts [150]. In 2012, Ge's group published a theoretical paper predicting the methanol synthesis activity of In 2 O 3 from CO 2 hydrogenation [151]. In the following paper, oxygen vacancy (D4) was deemed to play a key role in the activation of CO 2 and stabilizing the key intermediates involved (see Figure 5 for the structure of ideal and defective In 2 O 3 (110) surface) [42]. Since then, a number of works have been published reporting the superior activities and selectivities of pure In 2 O 3 or In 2 O 3 -based catalysts [83,[152][153][154][155][156][157][158][159][160][161]. A very early experimental test was performed on commercial In 2 O 3 powders after simple calcination in air at 500 • C for 5 h [158]. The catalyst showed somewhat comparable activities with the reported Cu-based catalysts. This could be a promising result since only pure In 2 O 3 was applied, not to mention the unknown specific surface area. Various strategies have been adopted to improve the performance, including: ble metal loading to facilitate H2 activation and formation of oxygen vacancies; (2) duction of oxide support to elevate the dispersion and resist sintering of the activ ponent; (3) adoption of different synthesis methods to achieve novel morphologica Various strategies have been adopted to improve the performance, including: (1) noble metal loading to facilitate H 2 activation and formation of oxygen vacancies; (2) introduction of oxide support to elevate the dispersion and resist sintering of the active component; (3) adoption of different synthesis methods to achieve novel morphological, electronic, or interfacial effects.
Recently, Liu's group has reported an In 2 O 3 supported Ir catalyst (Ir/In 2 O 3 ) showing enhanced activity, methanol selectivity as well as stability under the reaction condition of 300 • C, 5 MPa, and 2100 h −1 while keeping the H 2 /CO 2 ratio of 4 [83]. The catalyst was prepared by precipitation and impregnation method giving a specific surface area of 68 m 2 /g. Therein, the interaction between Ir and In 2 O 3 has proposed the key role in determining the structural characteristics of the catalyst, including a high dispersion of Ir, avoidance of over-reduction of In 2 O 3 , and stabilized oxygen vacancies that were thought to be the activation center for CO 2 molecules. Similarly, Rh supported on In 2 O 3 prepared by the deposition-precipitation method also exhibited good activity (space-time yield (STY) of methanol up to 0.5448 g MeOH ·g cat −1 ·h −1 ) and selectivity (56.1%) for CO 2 hydrogenation to produce methanol [16]. Higher selectivity was achieved at a lower reaction temperature due to the unfavorable RWGS at a lower temperature. The introduction of Rh gives rise to an enhanced H 2 dissociation efficiency, and the sequent spillover to In 2 O 3 further promotes the formation of surface oxygen vacancies. Similar effects were also found for Ni, Pt, Rh, Au, and Ru supported on In 2 O 3 [16,80,83,155,156,159,161].
Supported catalysts, however, sometimes suffered from sintering even though the metal nanoparticles were separated physically by support materials [136,152]. Appropriate interaction between support materials and metal nanoparticles may hinder the sintering to some extent. Au/In 2 O 3 catalyst prepared by deposition-precipitation method has exhibited the consequence of the interactions [80]. The Au/In 2 O 3 catalyst yields a methanol selectivity of 67.8% with the STY of 0.47 g MeOH ·g cat −1 ·h −1 . After a 12 h on-stream test, the conversion of CO 2 dropped by only 0.5%, and the mean particle size of the Au nanoparticles was slightly changed from 1.0 nm to 1.3 nm, indicating a nice structural stability that was not seen on other oxide support [80,157]. In addition, anchoring of Pd to the lattice of In 2 O 3 leads to superior catalytic performance, as well as structural stability. By controlled coprecipitation, Pd was atomically dispersed in the lattice matrix of In 2 O 3 , and the growing up of the Pd cluster was efficiently hindered [152].
Another challenge of the In 2 O 3 catalysts for CO 2 hydrogenation is the structural instability of In 2 O 3 during reaction [160], which can be overcome by choosing appropriate support materials [78,79] [78]. After ZIF-67(Co) was synthesized, it was impregnated with aqueous indium nitrate solution and then dried under vacuum and calcined under different thermal conditions (including pyrolysis at 600 • C in nitrogen flow and calcination at 400 • C in airflow). The obtained 3In@4Co (20) catalyst showed methanol selectivity of 75%, CO 2 conversion of 17.3% and methanol STY of 0.65 g MeOH ·g cat −1 ·h −1 at 285 • C, 50 bar and gas hourly space velocity (GHSV) = 15,600 h −1 with H 2 /CO 2 = 4:1. On the one hand, the In distribution all over the metal-organic matrix, micropore, and interparticle surface area can be optimized through tuning the textural properties of ZIF-67(Co); on the other hand, the formation of mixed metal carbide, namely Co 3 InC 0 . 75 , formed during the pyrolysis-calcination process can effectively stabilize high dopant distribution and prevent the formation of large oxide domains, which is beneficial to shorten the induction time during CO 2 hydrogenation.
Martin et al. [79] compared the performance of pure In 2 O 3 , In 2 O 3 supported on monoclinic ZrO 2 (In 2 O 3 /m-ZrO 2 ) and Cu/ZnO/Al 2 O 3 . 100% methanol selectivity was achieved under industrial conditions (T = 473-573 K, P = 5.0 MPa, GHSV = 16,000 h −1 ) over the pure In 2 O 3 ( Figure 6). The STY of methanol varies with reaction temperature as well as the CO concentration in the feed (keeping H 2 /CO 2 = 4:1) (Figure 6), which was ascribed to the evolution of oxygen vacancies. By comparison, the In 2 O 3 supported on monoclinic ZrO 2 significantly improves the yields of methanol because of the promoted formation of oxygen vacancies, the better dispersion of In 2 O 3 , the resistance against sintering, and possibly the modulation of the In 2 O 3 activities by a catalyst-support interactions. In addition, the supported In 2 O 3 showed excellent stability over 1000 h under reaction conditions. For both In 2 O 3 and In 2 O 3 /ZrO 2 , the oxygen vacancies were emphasized to play a key role in the adsorption and hydrogenation of CO 2 molecules.
tively stabilize high dopant distribution and prevent the formation of large oxide domains, which is beneficial to shorten the induction time during CO2 hydrogenation.
Martin et al. [79] compared the performance of pure In2O3, In2O3 supported on monoclinic ZrO2 (In2O3/m-ZrO2) and Cu/ZnO/Al2O3. 100% methanol selectivity was achieved under industrial conditions (T = 473-573 K, P = 5.0 MPa, GHSV = 16,000 h −1 ) over the pure In2O3 ( Figure 6). The STY of methanol varies with reaction temperature as well as the CO concentration in the feed (keeping H2/CO2 = 4:1) (Figure 6), which was ascribed to the evolution of oxygen vacancies. By comparison, the In2O3 supported on monoclinic ZrO2 significantly improves the yields of methanol because of the promoted formation of oxygen vacancies, the better dispersion of In2O3, the resistance against sintering, and possibly the modulation of the In2O3 activities by a catalyst-support interactions. In addition, the supported In2O3 showed excellent stability over 1000 h under reaction conditions. For both In2O3 and In2O3/ZrO2, the oxygen vacancies were emphasized to play a key role in the adsorption and hydrogenation of CO2 molecules. To further understand the role of ZrO 2 support, Frei et al. have prepared a series of In-Zr mix-oxide (synthesized by coprecipitation method with the molar content of In from 0 to 100%) and supported In 2 O 3 catalysts (on monoclinic and tetragonal ZrO 2 by wet impregnation method) [162]. The mix-oxide showed strong electronic interaction between In 2 O 3 and ZrO 2 while displaying inferior CO 2 hydrogenation activity, indicating a negligible electronic contribution of ZrO 2 to the activity enhancement. For supported catalysts, however, the methanol yield was about one order higher for In 2 O 3 supported on m-ZrO 2 than that on t-ZrO 2 , although on both supports, nuclear magnetic resonance (NMR) results showed ordered fine epitaxial In 2 O 3 . The ultraviolet-visible (UV-vis) spectrum of In 2 O 3 /m-ZrO 2 shows more absorption in the band gap, indicating a higher density of oxygen vacancies in the oxide. The m-ZrO 2 and t-ZrO 2 are slightly different in lattice constant, which makes the lattice mismatch to In 2 O 3 of 1.2% and 0.5%, respectively. The slightly larger lattice mismatch between In 2 O 3 and m-ZrO 2 tends to generate tensile strain that favors the formation of more and possibly diverse oxygen vacancies on In 2 O 3 [154,162]. The large discrepancy in catalytic activity was also attributed to the formation of electron-rich In 2 O 3 because of the electron transfer from ZrO 2 to In 2 O 3 [163], which could be from the same origin. The work highlighted the role of imperfect epitaxy in the formation of vacancies (defects) and consequent catalytic activities. Also, one must note that the peculiarity of m-ZrO 2 over t-ZrO 2 , in this case, cannot be simply transplanted to other systems, e.g., Cu, Ag, and Au supported on t-ZrO 2 were reported to show either better [164,165] or worse [166] CO 2 hydrogenation performance than on m-ZrO 2 , depending on the preparation method and pretreatment conditions. Further modification of In 2 O 3 /ZrO 2 system was achieved by loading active metals, e.g., via a coprecipitation method to obtain Au/In 2 O 3 -ZrO 2 catalysts [84]. While the Au nanoparticles (NPs) in the solid solution facilitate H 2 dissociation, the presence of Zr stabilizes the structure of In 2 O 3 and increases the number of oxygen vacancies, both as the result of SMSI.
Indeed, the SMSI between Pd and In 2 O 3 may lead to a bimetallic species that causes weakening of H 2 dissociation ability and hence the attenuation of the catalytic activities [167]. To avoid such alloy formation, Liu's group prepared the supported catalysts by introducing Pd-peptide composite, followed by a thermal treatment [168]. Compared with the conventionally prepared supported catalyst by the impregnation method, the resulted catalysts showed much better metal dispersion with a mean particle size of 3.6 nm and improved catalytic activities. The author attributed the improved performance of Pd-P/In 2 O 3 catalysts to the enhanced H 2 activation and more pronounced CO 2 adsorption ability as the result of highly dispersed Pt nanoparticles and the interfacial effect associated with the preparation method. In addition, atomically dispersed Pt/In 2 O 3 [169] was also reported to show good stability due to the higher barrier against sintering. A small amount of Pt loading (0.03-0.58 wt%) by coprecipitation results in anatomically dispersed doping of Pt in the In 2 O 3 lattice. The Pt n+ , though part of which is initially sintered under reaction conditions, facilitates the formation of oxygen vacancies and also the activation of hydrogen molecules.

Nanoalloy Catalysts
Nanoalloy catalysts often exhibit unique electronic structures distinguished from either component [106,[170][171][172]. As a result, the bonding properties of reactants, intermediates, or products can be tuned, which finally yields tunable activity and selectivity of the catalysts. Nanoalloy is formed either in the preparation process or under reaction conditions. The existence of nanoalloy can be validated by the finger-printed diffraction angle (2θ) in X-ray diffraction (XRD) [173], the alloy state of an element in X-ray photoelectron spectroscopy (XPS) [174], the specific vibrational mode of probe molecules binding to the alloy sites [175], or the lattice constant value shown in transmission electron microscopy (TEM) [88,176]. For a specific reaction, the turnover frequency (TOF) can be tuned by either the composition or the relative content [177]. Studt et al. compared the catalytic activities of three Ni-Ga catalysts via incipient wetness impregnation method and Cu/ZnO/Al 2 O 3 catalyst via coprecipitation route as reference [102,177]. The better activity and methanol selectivity of the Ni 5 Ga 3 /SiO 2 catalyst were attributed to the suppressed RWGS activity. While supported on mesoporous nitrogen-rich carbon, Ni 5 Ga 3 was also found to have good activity for CO 2 hydrogenation, though the activity is sensitive to the preparation method [176]. The author highlighted the freeze drying method that enables uniformly distributed metal nanoparticles with an average size of 2-5 nm and correlated the formation of NiGa alloy with the suitable local environment realized by the preparation method.
Shi et al. [88] synthesized Cu-In intermetallic catalyst and investigated the effect of reduction temperature on the reaction activity. It was found that reduction temperature exerts a notable influence on the formation of alloy, the crystallite size, and the adsorption of gases. For instance, CuO was reduced to metallic Cu at 250 • C, and the Cu 11 In 9 alloy appeared when the reduction temperature was above 300 • C. With higher reduction temperature, the crystallite size of Cu 11 In 9 increases slightly, accompanied by an attenuated H 2 adsorption ability. The CO 2 adsorption ability varies notably with increasing the reduction temperature, and the maximum value was obtained with CuIn-300 (reduced at 300 • C), the trend of which agrees well with the STY of methanol tested at 240 • C and 3 MPa (for the same reason, the CO 2 adsorption ability was thought to be the key factor for the design of Cu/In 2 O 3 catalysts). Besides the temperature, reduction pressure also affects the formation of alloy. An X-ray absorption near edge structure (XANES) analysis on the reduction process of commercially available Cu/ZnO/Al 2 O 3 catalyst at different pressure (1 mbar-10 bar) showed that Cu-Zn alloys were formed only under reduction pressure of 100 mbar or above [178]. The maximum reduction rate (simultaneous formation of copper (0)) is gradually shifted to low temperature by elevating the reduction pressure. The catalysts started to show methanol synthesis activity when the total pressure was above 1 bar, along with the increased formation of oxygen vacancies and other structural distortions in the ZnO phase.
Note that, in this section, we do not really attempt to emphasize the alloy formation in the activity modification for the metal supported on oxide since their interactions are not necessarily accompanied by the formation of nanoalloy. For Ni supported on In 2 O 3 , the authors emphasized the role of Ni-In alloy formation in determining its activity [153,172], while Hensen and coworkers [179] claimed the activity of Ni/In 2 O 3 system stems from the synergy effect with no alloy formation evidenced. In any case, the formation of nanoalloy in modifying the electronic structure and hence the catalytic performance provides a novel idea for catalyst design.

Other Catalysts
MoS 2 and MoS 2 -based materials have been widely used as lubricants [180], transistors [181], heterogeneous catalysts [182][183][184], and gas sensors [185]. Due to the special lattice structure, MoS 2 is easily peeled into thin layers or even a single atomic layer [186]. The electronic structure of two-dimensional MoS 2 is sensitive to the status of surface vacancies. The sulfur (S) vacancies especially located at edges or in-plane, exhibit completely different catalytic activities [90,187,188]. The edge S vacancies were thought to catalyze the CO 2 hydrogenation to methane, while the in-plane S vacancies were proven to be ideal active sites for CO 2 hydrogenation to methanol [90]. Over the in-plane sulfur vacancyrich MoS 2 nanosheets, a methanol selectivity of 94.3% at a CO 2 conversion of 12.5% was achieved at 180 • C, and the catalyst was stable for over 3000 h without any deactivation. The findings enlightened the potential role of the in-plane vacancies in catalysis, and further modification of the vacancies-controllable MoS 2 material or two-dimensional material is meaningful.
Due to the unique structural characteristics, MOF materials were featured with highly ordered and tunable porous structures, high surface area, flexible organic linkers, and metal centers [189][190][191]. Accordingly, MOFs can be a template for porous material synthesis [123], supporting materials for nanocatalysts [191,192], or act as catalysts individually by introducing active metal centers as nodes or located on the MOF linkers [193,194], as shown in Figure 7. Although many MOFs suffer from high temperature and moisture conditions, some selected MOF-based catalysts, including UiO-bpy [191], MOF-74 [192], UiO-67 [194], and UiO-66 [195], were reported to have good thermal stability even under moisture condition, of which further tuning of the catalytic properties is likely foreseen.
Another promising catalyst catalog for CO 2 hydrogenation to methanol is solid solution catalysts [74,[196][197][198]. Wang et al. have prepared a series of ZnO-ZrO 2 solid solution catalysts through the coprecipitation method [196]. Both methanol selectivity and CO 2 conversion reach maximum over the catalyst when Zn/(Zn + Zr) ratio is around 13%. Methanol selectivity of 86% to 91% with CO 2 conversion of 10% was achieved under the condition of 5.0 MPa, 24,000 mL·g −1 h −1 and 320 • C. The catalyst was proved to show long-term thermal and chemical stabilities against sintering and poisoning by, e.g., SO 2 or H 2 S. Density functional theory (DFT) simulation results suggested that Zn and Zr provide the adsorption sites for H 2 and CO 2 , respectively. Therefore, a solid solution catalyst takes advantage of both components to achieve the synergetic effect. Such synergetic effect was also observed in other solid solution catalysts such as M a ZrO x (Ma = Cd, Ga) [197]. Another interesting aspect of the ZnO-ZrO 2 solid solution catalyst is that its activity was reported to be sensitive to the preparation method (the microstructure) rather than the ZnO/ZrO 2 ratio. The 20% ZnO-ZrO 2 catalyst prepared by the evaporation-induced self-assembly process exhibited better methanol synthesis activity than that of the coprecipitation method [74]. The enhanced activity of the former was ascribed to its larger specific surface area related to the mesoporous structure and more active sites for CO 2 and H 2 adsorption (which are possibly correlated with the larger surface area). Due to the unique structural characteristics, MOF materials were featured with highly ordered and tunable porous structures, high surface area, flexible organic linkers, and metal centers [189][190][191]. Accordingly, MOFs can be a template for porous material synthesis [123], supporting materials for nanocatalysts [191,192], or act as catalysts individually by introducing active metal centers as nodes or located on the MOF linkers [193,194], as shown in Figure 7. Although many MOFs suffer from high temperature and moisture conditions, some selected MOF-based catalysts, including UiO-bpy [191], MOF-74 [192], UiO-67 [194], and UiO-66 [195], were reported to have good thermal stability even under moisture condition, of which further tuning of the catalytic properties is likely foreseen. Another promising catalyst catalog for CO2 hydrogenation to methanol is solid solution catalysts [74,[196][197][198]. Wang et al. have prepared a series of ZnO-ZrO2 solid solution catalysts through the coprecipitation method [196]. Both methanol selectivity and CO2 conversion reach maximum over the catalyst when Zn/(Zn + Zr) ratio is around 13%. Methanol selectivity of 86% to 91% with CO2 conversion of 10% was achieved under the condition of 5.0 MPa, 24,000 mLg −1 h −1 and 320 °C. The catalyst was proved to show longterm thermal and chemical stabilities against sintering and poisoning by, e.g., SO2 or H2S. Density functional theory (DFT) simulation results suggested that Zn and Zr provide the adsorption sites for H2 and CO2, respectively. Therefore, a solid solution catalyst takes advantage of both components to achieve the synergetic effect. Such synergetic effect was also observed in other solid solution catalysts such as MaZrOx (Ma = Cd, Ga) [197]. Another interesting aspect of the ZnO-ZrO2 solid solution catalyst is that its activity was reported to be sensitive to the preparation method (the microstructure) rather than the ZnO/ZrO2 ratio. The 20% ZnO-ZrO2 catalyst prepared by the evaporation-induced selfassembly process exhibited better methanol synthesis activity than that of the coprecipitation method [74]. The enhanced activity of the former was ascribed to its larger specific surface area related to the mesoporous structure and more active sites for CO2 and H2 adsorption (which are possibly correlated with the larger surface area).

Mechanistic Understanding
As shown in the above text, large numbers of catalysts have been reported on their superior activities for CO2 hydrogenation to methanol. They may be synthesized with different methods, from different sorts and proportions of raw materials, treated with different parameters, or activated under different reaction conditions. Consequently, the resulted catalysts are featured with varied compositions (at surface region), microscopic morphologies, particle sizes, surface area, bonding properties, etc., finally leading to unique activity, selectivity, and stability. One of the practically noticeable consequences is the gradually improved STY and methanol selectivity in the newest publications. Parallel to the improvement in catalyst development, a comprehensive understanding of the

Mechanistic Understanding
As shown in the above text, large numbers of catalysts have been reported on their superior activities for CO 2 hydrogenation to methanol. They may be synthesized with different methods, from different sorts and proportions of raw materials, treated with different parameters, or activated under different reaction conditions. Consequently, the resulted catalysts are featured with varied compositions (at surface region), microscopic morphologies, particle sizes, surface area, bonding properties, etc., finally leading to unique activity, selectivity, and stability. One of the practically noticeable consequences is the gradually improved STY and methanol selectivity in the newest publications. Parallel to the improvement in catalyst development, a comprehensive understanding of the structure-activity relationship has made a lot of progress based on systematic kinetic analysis, surface science study, operando techniques, theoretical simulations, etc., although one must note that some reported mechanisms might be case-dependent and not expanded to other systems.
An active site is one of the most important concepts in the study of catalysis [199][200][201]. It functions in multiway, providing adsorption sites for reactants, diffusion routes for adsorbate, suitable microscopic geometric matching for the reaction, appropriate bonding to intermediates, etc. Therefore, an active site is generally not a single point, a special micro or macrostructure, or a certain element with a specified chemical state, but the combination of a series of microscopic locations with distinctive functions and special elemental, electronic and geometric matching for the certain specific catalytic reaction [201]. It is an arduous task to build up a panorama of an active site, not to mention the difficulties superimposed by the structural complexity of nano-sized particles and numerous interferential issues. An alternative and feasible way is to simplify the catalyst system, e.g., to perform reactions on structurally well-defined model catalysts, such as single crystals, polycrystalline, and thin films [202][203][204]. In this section, we summarize the works about active sites for catalytic hydrogenation of CO 2 to CH 3 OH on both modeled single crystals and realistic powder catalysts.

The Chemical State of Copper
As early as the 1990s, Goodman et al. [205] have studied the hydrogenation of CO 2 to CH 3 OH over Cu (100) surface under the condition of 500-550 K and 44-102 kPa, with the feed composition of CO 2 /CO/H 2 being 1/2/12. The activation energy of CO 2 hydrogenation to CH 3 OH was estimated to be 73.4 ± 6.0 kJ/mol. No metallic copper was detected after the reaction, indicating that copper may take effect in an ionic state. They, therefore, concluded that CO 2 only plays a positive role in keeping the copper in a suitable oxidation state, and CO provides the carbon source for the reaction. In contrast, Chorkendorff et al. [206] worked on the hydrogenation of pure CO 2 on Cu(100) under the reaction conditions of 543 K and 2 bar (CO 2 /H 2 = 1:1) and did not find the oxidized state of copper after the reaction. However, they obtained similar apparent activation energy of 69 ± 4 kJ/mol. The introduction of CO into the reaction system significantly enhanced the reactivity by lowering the reduction barrier, where CO was thought not to participate in the reaction [171,207]. The discrepancy in the carbon source was likely unified by a DFT and kinetic Monte Carlo (KMC) study [208], which revealed that the main carbon source on the Cu(111) (metallic Cu) surface is CO 2 , whereas on Cu 2 O(111) (ionic Cu) surface is CO.
Besides the chemical state, CO 2 hydrogenation to CH 3 OH is also sensitive to the coordination state of copper. Compared with Cu(100) [205] and polycrystalline Cu foil [209], Cu(110) [210] exhibits higher CO 2 hydrogenation activity in its metallic state. The apparent activation energy for the synthesis of CH 3 OH on Cu (110) was measured to be about 67 kJ/mol, close to that on Cu(100) [205] and polycrystalline Cu foil (77 ± 10 kJ/mol) [209]. However, the apparent activation energy for RWGS on Cu(110) was measured to be 78 ± 14 kJ/mol, which is significantly smaller than that on polycrystalline Cu (135 ± 5 kJ/mol) [209]. Likewise, not all surface Cu atoms are active in the reaction. DFT results of CO 2 adsorption over Cu (100) surface have shown that CO 2 only dissociates at step sites [32]. While Kim et al. [211] found that the dissociation occurs only when H 2 is coadsorbed on step sites, where CO 2 is activated and converted to CO, surface oxygen (O*), and surface hydroxyl (HO*). Moreover, these active species were subsequently converted into carbonate (CO 3 2− ), bicarbonate (HCO 3 − ), and formate (HCOO*) species. Formate species formed on terrace sites were then converted to CH 2 O(O) on step sites that were subsequently converted to methoxy and methanol.
In numerous cases, the phase and microstructure are varied dynamically with the ambient environment, such as pressure, temperature, proportion, duration, etc., making the determination of the active site unconvincing. In such a situation, in situ or operando spectroscopic investigation may tell some more detailed information [212,213]. Ren et al. [214] provided direct spectroscopic evidence for the microscopic process of CO 2 hydrogenation to CH 3 OH on Cu(111) surface at room temperature by near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS), ultraviolet photoelectron spectroscopy (UPS), and low energy electron diffraction (LEED). They found that CO 2 dissociates on Cu(111) surface to form atomic oxygen, and hence Cu-O species under CO 2 + H 2 atmosphere, which consequently promote CO 2 hydrogenation to CH 3 OH through formate pathways, as shown in Figure 8.
Catalysts 2022, 12, x FOR PEER REVIEW 17 of 36 [214] provided direct spectroscopic evidence for the microscopic process of CO2 hydrogenation to CH3OH on Cu(111) surface at room temperature by near ambient pressure Xray photoelectron spectroscopy (NAP-XPS), ultraviolet photoelectron spectroscopy (UPS), and low energy electron diffraction (LEED). They found that CO2 dissociates on Cu(111) surface to form atomic oxygen, and hence Cu-O species under CO2 + H2 atmosphere, which consequently promote CO2 hydrogenation to CH3OH through formate pathways, as shown in Figure 8.

Strong Metal-Support Interaction
The concept of strong metal-support interaction (SMSI) was initially put forward to explain the attenuated adsorption ability (e.g., to CO and H2) of the group VIII metals supported on TiO2 after reduction at high temperature (e.g., 500 °C) [215]. The concept was extended to metal particles supported on reducible transition metals. Even more gen-

Strong Metal-Support Interaction
The concept of strong metal-support interaction (SMSI) was initially put forward to explain the attenuated adsorption ability (e.g., to CO and H 2 ) of the group VIII metals supported on TiO 2 after reduction at high temperature (e.g., 500 • C) [215]. The concept was extended to metal particles supported on reducible transition metals. Even more generally, the support materials were believed not to be limited to transitional or reducible if the surface reduction actually occurs (e.g., the oxides of niobium, manganese, and lanthanum) [216,217]. Later, the SMSI effect was also observed under oxidative treatment conditions, further extending the generality of the concept [218]. Although the driving force for SMSI is energetic in nature, it is often accompanied by electron transfer between metal and support, formation of an intermetallic bond, coating of metal by support or of support by metal, and behaviors related to them (e.g., formation of oxygen vacancies, presence of interface and synergetic effect, stabilization of metal component against thermal and chemical sintering, tunable bonding properties of the catalyst, etc.). Herein, one must note that the consequences induced by SMSI are not necessarily positive for catalytic properties, just like the case it came from, but provide the possibility to tune the properties of catalysts.
For the industrially available Cu/ZnO/Al 2 O 3 catalyst, it is generally agreed that ZnO acts as either support to separate Cu nanoparticles physically or an electronic and structural promoter to give rise to synergy effect with the latter mainly induced by SMSI [133,215]. The effect of SMSI may include the formation of the positively charged metal component, oxygen vacancies (with different types), Cu-Zn alloy, electronic doping of ZnO, and possibly the consequent stabilization of copper nanoparticles and metal-oxide interfaces [98,106,108,219]. A direct demonstration of the SMSI effect on catalytic activity was carried out by Behrens et al. [98], where Cu steps were thought to be one of the key factors to facilitate the formation of methanol. The presence of Zn δ+ at defect sites was verified to be the result of the dynamic SMSI effect leading to the partially covered metal particles with ZnO x . The dynamic change was also accompanied by the alloy formation through substitution of Zn into the Cu steps, the presence of partially oxidized Zn δ+ , and the geometric matching between Cu steps and adjacent ZnO x for stabilizing the intermediates to methanol.
Fujitani et al. [220] have studied the influence of Zn deposition on the activity of Cu at 523 K and 18 atm and found that moderate coverage of Zn could promote the methanol synthesis of Cu(111) while poisoning the Cu (110). The promotion effect of Zn on Cu (111) was ascribed to the preferential formation of ZnO x species on Cu(111) over Cu (110) surface. This could be indirect evidence to show the effect of SMSI on the activities of Cu/ZnO catalysts. A further study by Nakamura et al. [221] showed that Cu + -O-Zn species are formed during CO 2 hydrogenation over polycrystalline Cu surface with Zn deposition. The presence of Cu + -O-Zn species stabilizes the intermediate species (formate and methoxy). On the physically mixed Cu/SiO 2 and ZnO/SiO 2 powder catalysts [222], they also found that ZnO moiety partly migrates onto the Cu surface, creating Cu + -O-Zn active species and partly dissolves into Cu particles to form Cu-Zn alloys. After deposition of Cu clusters on ZnO(0001)-Zn terminated surface, a small amount of Cu promoted the reduction of ZnO by triangular reconstruction for syngas hydrogenation to CH 3 OH [223]. The metallic zinc leads to the formation of Cu-Zn alloys, which exhibit an excellent activity for CH 3 OH formation.

Interfacial Effect
The Cu-ZnO interface was believed to be a key factor for Cu/ZnO catalysts, which has been evidenced by the metastable "graphite-like" ZnO x layers on the top of Cu nanoparticles observed by transmission electron microscopy (TEM) [126]. Palomino et al. [224] also found that the reactivity of Cu single crystal surfaces was significantly enhanced after the deposition of ZnO nanoparticles. The activity follows Cu(111) < Cu(100) < ZnO/Cu(111) < ZnO/Cu(100), where the role of Zn was thought to stabilize the formate intermediate. Kattel et al. [103] have compared the activity of Zn-Cu alloy and ZnO/Cu model catalysts for methanol synthesis and found that metallic zinc nanoparticles were transformed to ZnO after running CO 2 hydrogenation reaction over the Zn/Cu(111) catalyst at 525 K and 550 K. They, therefore, concluded that the synergy of ZnO and Cu at interface catalyze the hydrogenation of CO 2 , although Nakamura et al. thought the positive valence state of Zn after reaction could be the result of formate species binding to Zn [104,105,173].
Despite the unsettled divergence in the XPS assignment, the model proposed by Muhler et al. [125] about the structural change over time on stream (TOS) possibly provided a unified understanding (Figure 9). The Zn-Cu alloy was initially formed during the reduction process on Cu/ZnO prepared either by coprecipitation method [225] or by a physical mixture of Cu/SiO 2 and ZnO/SiO 2 [221]. The Zn species was then partially oxidized to Zn δ+ at the defective Cu sites under CO 2 hydrogenation conditions due to the SMSI effect [103,104]; Along with diffusion and oxidation of Zn, there may form a graphitic-like ZnO x layer on Cu [126,226,227]; and finally, thick ZnO layers are formed after long-term reaction under CO 2 hydrogenation conditions [228]. In the whole process, the initial fast-decreased activity was thought to be the result of the partial restructuring of ZnO, which influences the stability of the Cu-ZnO interface; the long-term deactivation was ascribed to the loss of specific surface and Cu-ZnO interface due to the crystallization of ZnO. bar. Recorded degrees of conversion (black points) under differential controlled conditions. The dashed red curve describes the intra and extrapolation of the experimental data, which were calculated with the MATLAB ® software according to the studies of Fichtl et al. [230]. Illustrations from left to right: Cu 0 -Zn 0 surface alloy according to Nakamura et al. [231]. Zn δ+ species at the defective Cu 0 surface according to Behrens et al. [98], graphitic-like ZnOx layer on Cu 0 according to Lunkenbein et al. [126]. ZnO layer on Cu 0 according to Lunkenbein et al. [229] as well as Fichtl et al. [230] and ZnO on the top layer of Cu 0 according to Kattel et al. [103]. Reprinted with permission from ref. [125]. Copyright 2020 Springer.
The interface (or synergetic) effect of Zn-Cu alloy was predicted by DFT simulations as well, where Zn atoms are located on the Cu surface in the form of single atoms or clusters [232]. At the Cu-Zn interfaces, both single-atom Zn and Zn clusters (Zn3) can catalyze CO2 to methanol with the activation barrier notably lower than that on a pure Cu surface [233]. While the Zn single atom/Cu interface is thermodynamically and kinetically resistant to oxidation, the Zn clusters can be easily oxidized through the dissociation of CO2 at interfaces (which is likely the case under industrial reaction conditions), leading to an obvious reduction of the reaction barrier. Similarly, the rate of methanol formation over CeOx/Cu (111) was found about 200 times faster than that on Cu (111), which was ascribed to the synergetic effect of the interface between two complementary components of copper and ceria [234]. The ceria components stabilize CO2 − and OH species, providing complementary chemical properties for Cu catalysts. Such unique matching was barely ob- Figure 9. Long-term methanol synthesis over the industrial Cu/ZnO/Al 2 O 3 catalyst at 210 • C and 60 bar. Recorded degrees of conversion (black points) under differential controlled conditions. The dashed red curve describes the intra and extrapolation of the experimental data, which were calculated with the MATLAB ® software according to the studies of Fichtl et al. [229]. Illustrations from left to right: Cu 0 -Zn 0 surface alloy according to Nakamura et al. [230]. Zn δ+ species at the defective Cu 0 surface according to Behrens et al. [98], graphitic-like ZnO x layer on Cu 0 according to Lunkenbein et al. [126]. ZnO layer on Cu 0 according to Lunkenbein et al. [228] as well as Fichtl et al. [229] and ZnO on the top layer of Cu 0 according to Kattel et al. [103]. Reprinted with permission from ref. [125]. Copyright 2020 Springer.
The interface (or synergetic) effect of Zn-Cu alloy was predicted by DFT simulations as well, where Zn atoms are located on the Cu surface in the form of single atoms or clusters [231]. At the Cu-Zn interfaces, both single-atom Zn and Zn clusters (Zn 3 ) can catalyze CO 2 to methanol with the activation barrier notably lower than that on a pure Cu surface [232]. While the Zn single atom/Cu interface is thermodynamically and kinetically resistant to oxidation, the Zn clusters can be easily oxidized through the dissociation of CO 2 at interfaces (which is likely the case under industrial reaction conditions), leading to an obvious reduction of the reaction barrier. Similarly, the rate of methanol formation over CeO x /Cu (111) was found about 200 times faster than that on Cu (111), which was ascribed to the synergetic effect of the interface between two complementary components of copper and ceria [233]. The ceria components stabilize CO 2 − and OH species, providing complementary chemical properties for Cu catalysts. Such unique matching was barely observed on pure Cu or Cu-Zn alloys [98,234], which emphasizes the uniqueness of the metal-oxide interface in property tuning. Such effect was also observed on Cu-CeO 2 [68], Pd-Ga 2 O 3 [235], Au-CeO x /TiO 2 [236], Cu/ZrO 2 [111,237], Cu/Al 2 O 3 [128], CuO-In 2 O 3 [88], Au 4 /In 2 O 3 [156], etc. Overall, the interfacial effect in supported catalysts is not only a geometric matching but also an electronic effect, just as the fact that it is often induced by SMSI.

Size Effect
In some earlier years, the size-dependent activity has been largely reported on gold and other catalysts [238,239]. The single-atom catalysts reported in recent decades could belong to the catalog as well. The size effect is mainly an electronic effect, with no exclusion of the possibility of structural effect since the convergence of surface energy may lead to different geometry and lattice parameters of the nanoparticles [240]. Pacchioni et al. [241] have investigated the electronic structure of rutile TiO 2 (110) thin films ranging from two to ten layers and observed oscillated band edges (bottom of conduction band (CB) and top of valence band (VB)) and varied band gaps. When reducing the particle size of metal clusters, the energy levels of the frontier electron orbitals become more localized, and loss of metallic properties may occur at some critical dimension. Che and Bennett [242] roughly estimated the size effect of metal particles by comparing the spacing of electronic levels of the particle with thermal energy kT and obtained the critical dimension of about 2 nm. Despite the unknown error bar, it gave a reasonable coincidence with the reported work where the physical and catalytic properties start to change near this dimension. Goodman and coworkers [239] have studied the size-dependent activity of Au/TiO 2 for CO oxidation and found the activity reaches the maximum at two-atom-thick clusters in which nonmetallic properties start to appear. The size dependency of Au activity was thought insensitive to the support materials [238,243].
For small Cu clusters (n = 3, 4, 20) supported on Al 2 O 3 thin films, the catalysts exhibit strong size-and temperature-dependent activity for CH 3 OH synthesis [244]. Cu 3 /Al 2 O 3 shows maximum activity at 175 • C, while Cu 20 /Al 2 O 3 is more active at higher temperatures (275-325 • C). For all temperatures tested, Cu 4 /Al 2 O 3 was the most active one. The finding differs from the quasi-linear relationship between methanol synthesis activity and specific Cu surface area for each class of the catalysts (e.g., Cu/ZnO, Cu/Al 2 O 3 , and Cu/ZnO/Al 2 O 3 ) reported by Kurtz et al. [138]. The discrepancy is caused by the size effect, possibly through the change in electronic properties, since the Al 2 O 3 is only a structural promoter in the Cu/Al 2 O 3 catalyst (a synergetic effect is ruled out). Although the grouping of different catalysts proposed by Kurtz may not represent a general case, the work evidently exhibited that the size effect is not isolated but correlated with the microenvironment of the active components.
Zhang et al. [245] have worked on a series of Cu clusters with DFT simulations and observed a linear relationship between the CO 2 hydrogenation activity and the adsorption energies of CO and O on Cu clusters. The latter were thought to correlate with the location of the d band center that varies with the size of clusters. Note that the maximum cluster adopted in the calculations consisted of 79 Cu atoms which correspond to a diameter of about 1.5 nm. For large clusters, Karelovic et al. [246] claimed that the intrinsic activity for methanol synthesis is independent of the particle size (from 8.5 to 37.3 nm), and the enhanced methanol selectivity over the catalyst with a large particle size stems from the suppression of RWGS reaction. In comparison, Van den Berg et al. [131] have studied the size effect of Cu particles supported on different carriers with the particle size ranging from 2 to 15 nm and concluded that the catalytic activity for CO 2 hydrogenation to methanol is close correlated with the particle size for each group of the catalysts. Smaller particles show even worse activity because the reaction takes place only at Cu sites with a special configuration, e.g., step edge sites, which are not favorable on smaller particles [131]. This work again demonstrated the impact of the microenvironment that superimposes the size effect.
Although the size of metal particles can be tuned by changing the loading of metal content by wet impregnation [246,247] or choosing a porous host with suitable pore size [248], the possible morphological evolution under reaction conditions (e.g., with high pressure, high temperature and the presence of reactants) may lead to even more uncontrollable results [140,142,249]. From this aspect, the MOF template could be a choice to stabilize the metal particles [123,189].

Oxygen Vacancies
The less coordination of the atoms at defect sites often makes them more reactive than bulk or facet atoms [250][251][252]. The role of such sites has been extensively reported, such as the stepped Cu in the formation of Cu-Zn alloy, the oxygen vacancies on ZnO for CO hydrogenation, the in-plane sulfur vacancies for the CO 2 dissociation and hydrogen activation, and oxygen vacancies in In 2 O 3 for CO 2 adsorption.
For CO 2 hydrogenation over Cu/ZnO catalysts, although the active site was proposed to be the Cu-ZnO interface, the formation of such interface is always accompanied by the formation of oxygen vacancies. In other words, it seems that the latter triggers the formation of Zn-Cu alloy and then the Cu-ZnO interface. Different from the Cu/ZnO system, for Cu supported on ZrO 2 , oxygen vacancies functionalize through activating CO 2 [118,253] (One may note that the statement that the activation of CO 2 is the rate-determining step for CO 2 hydrogenation or enhanced CO 2 adsorption will benefit the hydrogenation activity is restricted to specific catalyst system and cannot be extended to a general case). On pure In 2 O 3 catalyst, the oxygen vacancies were thought to activate the CO 2 molecule and stabilize the intermediates [79,83,162]. For the recently reported two-dimensional MoS 2 nanosheet catalysts, the in-plane sulfur vacancies were verified to dissociate CO 2 to CO, which is further hydrogenated to methanol. On Cu/CeO 2 catalyst [117], excess surface oxygen vacancies enhanced the stability of Cu in different oxidation states, which leads to the stabilization of the COOH intermediate at the active sites.
The formation of oxygen vacancies is sensitive to the environment. For instance, the formation energy of oxygen vacancies on Cu/CeO 2 (111) ranges from 3.8 to 86.8 kJ/mol depending on the positions, significantly smaller than that on stoichiometric CeO 2 (111) surface [254]. The formation is promoted by loading selected metals such as Ni, Ir, Rh, Ru, etc., in In 2 O 3 [16,83,161]. The SMSI further stabilizes oxygen vacancies, leading to enhanced activity and stability. Moreover, the formation of oxygen vacancies on In 2 O 3 is also facilitated by slightly imperfect lattice mismatching, e.g., supported on monoclinic ZrO 2 [154,162,163]. Figure 10 presents the proposed reaction pathways for methanol synthesis over Cubased catalysts. CO 2 is hydrogenated through two competing pathways via either formate or carboxyl intermediates [255]. CO was considered as starting material as well since methanol can be synthesized via CO intermediate. The dissociation of H 2 was proposed to be the initial stage of the reaction, and then CO 2 reacts with preadsorbed H, where the preadsorbed H indeed enhances the binding energy of CO 2 on Cu and improves the probability for further elemental steps.  [257]. They exposed 0.2 bar H2 to CO2 pre-exposed Cu/Pt(111) surface and obs significant loss of surface carbonate species (formed by pre-exposed CO2). An infer hydrogenation of carbonate was then drawn based on the experimental findings. thors explained the missing formate species in the postreaction XPS spectrum by sile strain, which lowers the decomposition barrier of formate species and possi plies structure-sensitive intermediates and reaction pathways.

Reaction Intermediates and Reaction Pathways
On Ni/In2O3 catalyst, a typical reaction pathway via formate pathway and pathway is shown in Figure 11 [258]. For both Ni4/In2O3_P and Ni4/In2O3_P surfa RWGS pathway is the preferable route. The introduction of oxygen vacancies signi reduces the activation barrier. Hydrogen molecules dissociate over Ni species an spill over to the interfacial sites for further reaction. Varied from the theoretical prediction, Schumacher et al. claimed that carbonate is an important intermediate for CO 2 hydrogenation on Pt(111) surface strained Cu overlayers [256]. They exposed 0.2 bar H 2 to CO 2 pre-exposed Cu/Pt(111) surface and observed a significant loss of surface carbonate species (formed by pre-exposed CO 2 ). An inference of hydrogenation of carbonate was then drawn based on the experimental findings. The authors explained the missing formate species in the postreaction XPS spectrum by the tensile strain, which lowers the decomposition barrier of formate species and possibly implies structure-sensitive intermediates and reaction pathways.
On Ni/In 2 O 3 catalyst, a typical reaction pathway via formate pathway and RWGS pathway is shown in Figure 11 [257]. For both Ni 4 /In 2 O 3 _P and Ni 4 /In 2 O 3 _P surfaces, the RWGS pathway is the preferable route. The introduction of oxygen vacancies significantly reduces the activation barrier. Hydrogen molecules dissociate over Ni species and then spill over to the interfacial sites for further reaction. Catalysts 2022, 12, x FOR PEER REVIEW 23 of 36 Figure 11. Gibbs energy profiles for CO2 hydrogenation to methanol over Ni4/In2O3 via the (a) formate pathway and (b) RWGS pathway. The orange line and blue line represent the Ni4/In2O3_P and Ni4/In2O3_D, respectively. Reprinted with permission from ref. [258]. Copyright 2022 Elsevier.

Conclusions and Outlook for Future Work
A catalytic reaction usually consists of a number of essential steps that occur in series. A good catalyst is not necessary to catalyze each of the steps but to uniquely reduce the energy barrier of the rate-determining step or to show the best statistical performance for all steps, which is generally performed by the active site. The rate-determining step, however, is not immutable but correlated with a specific catalyst. This is probably one of the reasons why most of the reported catalysts revolved around a limited number of composites or elements. Microscopic understanding of the structure-activity relationship at the elemental level is worthwhile for the rational development of CO2 to methanol catalysts. However, one must note that, in most cases, an active site is not a single atomic site with a specific element and chemical state but a series of microscopic locations with distinctive functions and special elemental, electronic and geometric matching for certain specific elemental steps. In other words, catalysis is a multidimensional issue. Therefore, the ultimate aim of the catalyst study is to control the microscopic structure and electronic structure by varying macroscopic parameters (e.g., temperature, pressure, composition, syn-

Conclusions and Outlook for Future Work
A catalytic reaction usually consists of a number of essential steps that occur in series. A good catalyst is not necessary to catalyze each of the steps but to uniquely reduce the energy barrier of the rate-determining step or to show the best statistical performance for all steps, which is generally performed by the active site. The rate-determining step, however, is not immutable but correlated with a specific catalyst. This is probably one of the reasons why most of the reported catalysts revolved around a limited number of composites or elements. Microscopic understanding of the structure-activity relationship at the elemental level is worthwhile for the rational development of CO 2 to methanol catalysts. However, one must note that, in most cases, an active site is not a single atomic site with a specific element and chemical state but a series of microscopic locations with distinctive functions and special elemental, electronic and geometric matching for certain specific elemental steps. In other words, catalysis is a multidimensional issue. Therefore, the ultimate aim of the catalyst study is to control the microscopic structure and electronic structure by varying macroscopic parameters (e.g., temperature, pressure, composition, synthesis method, etc.). It seems that we are facing a black box on the way of catalyst development. Catalytic properties are correlated with the composition, geometry, particle size, shape, acid-base properties, chemical states, type of promoters, bonding properties, etc. Particularly, these factors are not isolated but correlative tight or weak, random or regular. This means, especially for the study on the reaction mechanism, the rule drawn on a catalyst through systematically changing certain factors might be misled by the concomitant factors. Conclusions must be drawn with care. In addition, the dynamic structural evolution of catalysts under reaction conditions brings more complexity to the issue (note that the structural evolution is not always negative for activity). At this point, operando techniques would benefit the mechanistic study [258][259][260][261][262]. For the same reason, DFT-derived energy and activation barriers are sensitive to the coverage of surface species. Kinetic analysis may reveal more details about the reaction [263][264][265]. Besides, machine learning has shown its unique advantages in material science as well as the mechanistic understanding of active sites, which could be a new branch of catalyst development [266][267][268].
Significant progress has been made on the development of the catalyst for CO 2 hydrogenation to methanol, including improving activity, selectivity, stability, microscopic understanding of the structure-activity relationship, prediction of new catalysts, etc., on the laboratory scale. The reported STY CH3OH and selectivity per se are high enough for the industrial application (e.g., a typical STY of methanol from syngas is roughly estimated to be about 0.5 g·g cat −1 ·h −1 [97]). Catalyst development at a larger scale, including synthesis of catalysts, evaluation of activity and stability (e.g., chemical, mechanical, and thermal stability), etc., may fasten the industrialization objective.  Acknowledgments: This work was supported by the National Natural Science Foundation of China (Project No. 21972128). We thank Yu Liu for carefully reading the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.