CO2 Valorization Reactions over Cu-Based Catalysts: Characterization and the Nature of Active Sites

Active sites are the individual reactors at the molecular scale distributed on the heterogeneous catalyst surface. To a large extent, they determine the catalytic performances and the reaction pathway of a reaction. Therefore, understanding the nature and structure of the actives sites is crucial to improve and develop novel, robust and practical catalysts. The wide application of state-of-the-art characterization techniques these years makes it possible to obtain crucial information about the active sites for some catalysts. The Cu-based catalysts are widely used for water gas shift (WGS) and methanol synthesis from syngas (CO + H2). Although having some technical issues in the direct conversion of CO2 into value-added products, they are still promising for this reaction to mitigate CO2 concentration in the atmosphere. In the last several years, intensive efforts have been made to study Cu-based catalysts, and substantial progress has been achieved in understanding their active sites and the reaction mechanism. This review discusses the structure and nature of active sites of Cu-based catalysts for CO2 valorization in thermo-, photo-, and electro-catalysis. We present the characterization results of different types of Cu-based catalysts applied in these processes, unravel their active sites and structures, and figure out the most important and critical factors that drive the reactions on the sites. The principle and applications of various characterization techniques are also briefly analyzed and compared. It is expected to provide fundamental insights and perspectives for designing highly active and efficient catalysts for CO2 conversion.


Introduction
While utilizing fossil fuels through various industrial processes to generate energy and chemicals, it also emits carbon dioxide (CO 2 ), which has been recognized as one of the key greenhouse gases that result in the earth warmer than normal, causing imbalance and unsettling the ecosystem [1]. Human activities release an estimated 37.1 billion tons of CO 2 into the air each year, and about 80% of this amount is emitted directly from fossil fuels exploitation [2]. The increasing CO 2 concentration in the atmosphere will remain for the next few decades because fossil fuels will continue to be the primary source of energy [2]. The concentration is predicted to reach 550 ppm by 2050 [3,4] and 590 by 2100 [5] if no proper measures are implemented. This continuous rise in CO 2 concentration in the atmosphere is a big challenge to mitigating global warming, raising serious concern and calling for solutions.
Carbon capture, storage, and utilization are proven strategies for controlling anthropogenic CO 2 emissions [6,7]. The latter is pursued because of the possibilities of transforming the captured CO 2 into economically viable products like feedstocks for industrial processes. The economic benefit of converting CO 2 into chemicals and fuels is enormous. It helps in the net reduction of CO 2 emission, contributing to global emission control and decreasing dependency on fossil energy [8,9]. Figure 1 shows how far CO 2 conversion technology has developed. The past few decades have seen substantial growth in research devoted to technologies for converting CO 2 into chemical products such as hydrocarbons and methanol by direct or indirect routes. Characterization and the Nature of Active Sites Efforts were much dedicated to the development of catalysts with high conversion, selectivity, and stability. An example is the methanol synthesis catalyst which is still suffering from low equilibrium concentration under high temperature and pressure operating conditions [10]. Most early CO 2 conversion technologies have undergone major or marginal changes to the current status that have improved the process even further, paved the way for possible future improvements. However, many proposed technologies are still at the infant stage and require significant improvement to be on course to an industrial scale implementation. The catalysts developed in the past have been optimized to high selectivity and stability. But issues bothering on recyclability and rapid deactivation by coking under harsh reaction conditions remain. The captured CO 2 may be converted into fuels, construction materials, precursor chemicals for industrial products, such as fertilizers, plastics, adhesives, pharmaceuticals, and specialty chemicals like urea and salicylic acid [11]. For example, CO 2 can be converted into CO via the reverse water gas shift (RWGS) reaction and then hydrocarbons through the Fischer-Tropsch (FT) synthesis. CO 2 reformation of methane from natural or shale gas to produce syngas is also a well-established technology [12]. The syngas is a key starting material or feedstock of the present chemical industries. The direct conversion of CO 2 to methanol is still far from full maturity and profitable large-scale utilization. Also, enormous advances have been made in the catalytic conversion of CO 2 into short-chain products like lower olefins (C 2 −C 4 ) [13]. However, it is still a far reach from achieving the selective conversion into long-chain hydrocarbons (C 5+ ). Although highly active and selective catalysts for some of these reactions are available commercially, the rapid loss of their activities under the harsh reaction conditions due to deactivation of the catalytically active components by sintering and coke deposition is challenging [14,15]. The future holds even more opportunities for CO 2 conversion to other valuable materials. It has been found that striking CO 2 against an inert surface like Au foil can split the molecule to form molecular oxygen (O 2 ) and atomic carbon (C); the O 2 has many applications, such as in oxidation reactions. Oxygen is also safe as artificial air for astronauts traveling to and from space or other planets. This discovery is an inspiration to the Caltech reactor, which is proposed as a future reactor for removing CO 2 from the atmosphere.
Generally, methods, including thermo-, photo-, electro-catalytic reduction processes, and the combination thereof, have been adopted for CO 2 conversion [16][17][18][19][20][21][22][23][24][25]. The CO 2 reduction with H 2 gas, which can be obtained from the splitting/electrolysis of water or renewable energy sources, is a practical process for producing methanol using Cu-based catalysts, with demonstrations even at the commercial scale [10]. Numerous and extensive investigations have been conducted to synthesize, optimize, and develop catalysts to convert CO 2 to methane, methanol, ethanol, ethylene, and platform chemicals [26][27][28][29]. In many of the published studies, catalysts based on Cu nanoparticles (Cu NPs) have been found as the most active for CO 2 utilization by the reduction process [10,30]. These are often supported catalysts that have interaction amongst the various components [31]. Although studies have provided particular understandings into the working states of the catalysts, challenges associating with the accurate characterization of the complex system make the findings lacking and studies inconclusive. Thus, the particular structure and active site of this type of catalysts are not fully understood. The accurate mapping of the activity-structure relationships of heterogeneous catalysts requires adequate knowledge of the structure and active sites/phases. Clear identification of the catalyst structures remains a major challenge and requires an extensive and thorough characterization to clarify the fundamental bulk and surface chemistries. The basic understanding of catalysis of CO 2 reduction with Cu can be attained if the structure and active surface of Cu-based catalysts are determined.
This review discusses the structure and active sites of Cu-based catalysts investigated for converting CO 2 to value-added chemicals. A brief overview of the most important and commonly applied techniques for characterizing catalyst structure and active sites is also presented. We highlights primarily recent works applying a handful of both common and advanced techniques for unraveling the possible active sites, providing a holistic discussion to elucidate the influence of the structure and active sites on the mechanism of the CO 2 reduction over Cu-based catalysts. Providing a general insight into this is crucial for improving the performance of the Cu catalyst systems and gaining a fundamental understanding of their activity.    [10,42]. When CO 2 was used as the feed, formate was the dominant species, but methoxy species predominantly appeared with CO as the main reactant. Simplified mechanism for CO2 hydrogenation to methanol.

Photocatalytic CO 2 Reduction
Interests in the photochemical reduction of CO 2 sprung up following the pioneering work of Inoue et al. [50] in the late 1970s. Also called artificial photosynthesis, the photocatalytic reduction of CO 2 with H 2 O using solar energy is an attractive route for transforming CO 2 into high-value products, although molecular H 2 can also be used [51][52][53][54]. Solar energy (light) plays a crucial role in this reaction by inducing charge excitation. The photocatalyst must possess the ability to absorb light and generate electron-hole pairs that are then separated and transferred to the surface of the catalyst surface to take part in the reduction and oxidation half-reactions [55].

Cu-Based Catalysts for Photocatalytic Reduction of CO 2
The photocatalysts investigated for the reduction of CO 2 are mostly semiconductors, including TiO 2 , ZnO, and CuO x [56,57], but our interest in the present work focuses on Cu-based semiconductors/photocatalysts. Cu can trap generated electrons, preventing the fast charge recombination. This function promotes the redistribution of charge on the surface of the photocatalysts which increases photoefficiency [30,57,58]. CuO and Cu 2 O have been extensively studied in the photoreduction of CO 2 to valuable products, such as methanol, methane, formic acid, and CO [59][60][61]. Cu 2 O is a p-type semiconductor with a wide bandgap of approximately 2.2 eV, which is very attractive as a photocatalyst for CO 2 reduction in the presence of visible light. However, Cu 2 O has issues related to its stability as it can be easily oxidized or reduced by the photogenerated charge carriers [62,63].
Generally, pure copper oxides exhibit a low selectivity for CO 2 reduction products due to fast charge recombination rates [64]. However, the selectivity can be improved by incorporating other materials as supports or using copper as a co-catalyst with other photocatalysts. As reported, Cu 2 O with Al-doped ZnO and TiO 2 as protective layers improved the stability of Cu 2 O for photo-electrochemical water splitting [65]. The photoactivity of Cu 2 O was improved when used as a co-catalyst with TiO 2 , carbon quantum dots (CQDs), RuO x , or SiC [61,66]. The improvement is generally a result of the cooperative effect between Cu particles and the co-catalyst, facilitating both trapping processes of photoexcited electrons and holes and decreasing the recombination rate [57]. CQDs and TiO 2 , for example, possess excellent charge separation and transfer capabilities. On the other hand, Cu ions (Cu 0 , Cu + , Cu 2+ ) can improve the formation of sites for electron trapping and facilitate charge transfer when incorporated into TiO 2 [54,66,67].
Tian et al. [68] demonstrated the photocatalytic activity of CuO/Cu 2 O/Cu nanorods-decorated rGO composite under visible light irradiation. The rGO assisted in the electrons transfer from and within the CuO/Cu 2 O/Cu nanorods. Cu 2 O/reduced graphene oxide (Cu 2 O/RGO) junction composite synthesized by a microwave-assisted in situ reduction was found to increase the photoreduction activity for reducing CO 2 to CO by twofold. Activity results correlated with the retarded electron-hole recombination, efficient charge transfer (Figure 3 a, b), and protective function of RGO [60]. CuO readily redistributed electric charge on the surface of the semiconductor support and trapped electrons, preventing the recombination of electron-hole pair and increasing its photoefficiency in photo-reduction of CO 2 to methanol [69]. It was found that photocatalysts promoted by CuO performed better than the materials promoted by Cu 0 and Cu + species [30,70] as opposed to other studies where either Cu 0 or Cu + or both species were believed to drive the activity [71,72]. It has also been suggested that during the photochemical reaction, Cu + species may be reduced by photogenerated electrons to Cu 0 , which can trap electrons. However, Cu 0 can easily be re-oxidized to Cu + in the air.  [60], (c) Schematic showing five fundamental steps in photocatalytic CO2 reduction [73], (d) Proposed photocatalytic reduction mechanism of CO2 catalyzed by CQDs/Cu2O [61].

Mechanism of Photocatalytic Reduction of CO 2
Typically, the heterogeneous photocatalytic process consists of five sequential steps -light absorption and excitation to generate electron and hole, charge separation, CO 2 adsorption, surface redox reaction, and product desorption [60,74,75] (Figure 3 c). The photocatalyst first absorbs the solar energy in a step known as the "light harvesting step". When incoming light illuminates the photocatalyst, electrons are excited from the valance band (VB) to the conduction band (CB), leaving an equivalent number of holes in VB [76,77]. This process requires energy equivalent to or greater than that of the semiconductor catalyst's bandgap energy. However, photocatalysts cannot have a very large bandgap since this would limit their ability to use the solar spectrum effectively. The light-harvesting step depends on the morphology and structure of the catalyst used, and could be improved using meso-or microporous catalytic structures and band engineering [77]. The next step is the spatial separation of photo-generated electrons and holes involving excitation of the electrons in the valence band to the conductance band. This results in proton generation in the conduction band and holes in the valence band [78]. The excitation can be improved by tapering the band gap. The electrons tend to move toward the outer region of the semiconductor but can be restricted by the catalyst active sites present at the surface. The recombination of electrons and holes (volume) may take place. Thus, the process of charge separation is directly competing with the charge recombination. Generally, the recombination stages should be prevented because it causes energy loss [79]. The third step is CO 2 adsorption and reduction to the CO 2 •− radical anion. In this step, electrons must be transferred from the photocatalyst to the CO 2 molecule. Photocatalysts with large surface typically have more active sites for CO 2 adsorption. Furthermore, modification of the photocatalyst surface with alkali could improve the CO 2 adsorption, as in the case for TiO 2 [51,75]. This is because the acidity of CO 2 molecules would promote the reaction between CO 2 and the alkalinic photocatalyst surface, resulting in intermediates generation and favouring CO 2 molecule activation and subsequent reduction. The surface redox reaction is the fourth stage, in which electrons and holes migrate to the surface, causing surface recombination. Both the oxidation and reduction The holes catalyze water oxidation to molecular O 2 , while the electrons reduce CO 2 to different photocatalytic products [80,81]. This step is followed by the final product desorption.
Some specific examples of the photoreduction mechanism of CO 2 to specific molecules are briefly discussed. In the formation of methane over (Au, Cu)/TiO 2 under visible light in the presence of H 2 O, the reduction of CO 2 proceeds by adsorption of both reactants, CO 2 and H 2 O, leading to the formation of surface adsorbed CO 2 •species upon their activation by electron-hole transfer. This is swiftly followed by the cleavage of the C-O and formation of C-H bonds in a pathway termed "the carbine pathway" [82]. Li et al. [61] proposed a mechanism for reducing CO 2 to methanol under visible light over Cu 2 O/CQDs. Electron-hole pairs are produced upon excitation of Cu 2 O by visible light; the produced electrons are consumed for reducing CO 2 to methanol on the Cu 2 O surface [61]. The holes transferred to the surface of CQDs, where they oxidize H 2 O to O 2 . White et al. [83] described a mechanism involving the proton transfer to an oxygen atom on CO 2 •− assisted by water, the loss of an OH group, and the subsequent second electron transfer to produce an adsorbed CO species as demonstrated in the reduction of CO 2 to formate on transition metals. The addition of a second electron also led to the formation of the formate ion [83]. The surface intermediates, particularly CO 2 •-, Cu-CO, and carbon deposits on the surface, can be detected by surface-sensitive characterization tools, such as in-situ FTIR [82,84]. Kang et al. [84] used in situ FTIR experiments to detect the surface species during the CO 2 reduction to CH 4 . A series of adsorption bands around 1330-1590 cm -1 and 1600-1890 cm -1 , whose intensities increased with time appeared under the UV/Vis irradiation, attributed to the bending vibration of C-H in CH 4 , and the stretching vibration of C=O and the asymmetric stretching of O-C=O bonds in the intermediate products, respectively, such as aldehydes, carboxylic acids, and bidentate carbonates [85].

Electrocatalytic CO 2 Reduction
The electrocatalytic reduction of CO 2 is an appealing method for CO 2 utilization [86]. This process is characterized by the generation of a wide spectrum of products through different reaction pathways or a combination of the pathways, depending on factors, such as the nature of catalyst, nature of electrode, electrode potential, electrolyte composition, pH, temperature, and CO 2 feed concentration, the configuration of the cell, and presence of adsorbents [87][88][89][90]. The electrode potential regulates the production of electrons transported to the cathode, where they combine and reduce CO 2 to various products. Anode is where water oxidation reaction occurs. In the aqueous medium, if the electrons react with protons, hydrogen will be produced instead (hydrogen evolution). The products obtained from the electrochemical reaction can be made selective by changing the applied electrode potential and the catalyst. Certain products such as methanol and C 2+ hydrocarbons, difficult to produce by the thermocatalytic process at low reaction temperatures and pressures, can be easily realized by the electrocatalytic process at near or ambient temperature. Furthermore, electrochemical cells for the reduction of CO 2 are easier to scale up.
Since the CO 2 electroreduction reaction is governed by unfavorable thermodynamics and low Faradic efficiency, high overpotential (high energy) is always required to form a product. The kinetics are too slow to form an energetically reaction intermediate, which can be stabilized by selecting a proper catalyst that can lower the high overpotential requirements for the product formation and selectivity [91]. Meanwhile, some products require higher overpotential than others. For example, methanol and methane that need multi-electron for their formation require higher overpotential than CO or formate which one-or two-electron transfer is sufficient for their formation [92]. Moreover, the solubility of CO 2 in the aqueous electrolyte is poor, which hinders the free mass transfer of CO 2 from the feed gas to the catalyst surface, resulting in the overall low current density and productivity.
Many studies on this process have been carried out in both the liquid and gaseous phases using various catalysts and electrolytes [92]. The design of efficient CO 2 reduction electrocatalysts/electrolytes needs a proper comprehension of the mechanisms through which products are formed during the reaction. Among the reduction products, the formation of C 1 products (e.g., CO and HCOOH) on the bulk catalyst surface follows a relatively simple pathway. In contrast, the mechanism of CO 2 reduction to C 2+ products is a more complex process with many possible pathways involving both electrochemical and chemical steps [89,[92][93][94][95]].

Cu-Based Materials for Electrocatalytic Reduction of CO 2
Copper has a unique ability to promote the selective reduction of CO 2 to different products like formic acid, methane, methanol, ethanol, CO, and for C-C coupling toward C 2+ products with good Faradaic efficiencies when compared to other catalysts [94][95][96][97][98][99]. It has been reported that to date, only Cu catalyst can catalyze CO 2 reduction to substantial amounts of alcohols [96][97][98]. This may be associated with the adsorption energy of CO on Cu. Moderate or optimum adsorption energy is important because it allows the activation for further reduction and C-C bond formation, which is difficult to achieve with other metals that bind CO either too weakly or too strongly according to the Sabatier principle [8]. Cu has a CO adsorption strength close to the optimum, exhibits selectivity for various products from CO 2 reduction reaction. However, Cu catalysts often suffer from low conversion efficiency, poor selectivity, and unidentified intermediates [94]. To improve selectivity, the modification of the adsorption strength of CO on the catalytic sites is crucial [8]. Moreover, Cu catalysts perform at high overpotentials to attain substantial selectivity to C 2+ products [98,100].
The surface and its morphology plays a remarkable part in the selectivity of products. It was observed that the ratio of C 1 to C 2 /C 3 products was highly influenced by the surface morphology of Cu [101,102]. The open Cu (100) surface exhibited greater selectivity to products with C-C bonds than the close-packed Cu (111) facet [94]. Thus, identifying the possible mechanisms for the CO 2 reduction reaction remains a continuously interesting topic. Polycrystalline Cu is selective to CO, HCOOH, HCOO -, H 2 , and CH 4 at potentials more reductive than −0.8 V vs. RHE, while (110) and (111) steps nearby (100) terraces are selective toward C 2+ products [103]. This could influence the elementary steps that control the product distribution [94]. 2 Based on experimental data and density functional theory calculations, comprehensive mechanisms for CO 2 reduction on Cu electrocatalysts have been discussed. CO 2 is first chemisorbed, then reduces to adsorbed formate intermediates (HCOO * ), which eventually desorbs as formate ions. On the other hand, the * COOH intermediate, which is adsorbed via its C atom, would be converted to * CO intermediate. * CO is the key intermediate in the formation of hydrocarbons and alcohols [92]. Generally, it has been proposed and somewhat accepted that * CO is a key intermediate in the formation of hydrocarbons; however, a conclusive mechanistic view of Cu driven electroreduction of CO 2 to multi-carbon products is still far from reach due to the complexity of the involved reactions [98,104]. Understanding the detailed mechanisms for forming C 1 and C 2 has been pursued by several research groups in isolation and collaboratively.

Mechanism of Electrocatalytic Reduction of CO
Hori, one of the pioneer discoverers of Cu as a catalyst in the electrochemical CO 2 reduction to CO, proposed that, in the aqueous medium, CO 2 initially adsorbs on the surface of electrodes and forms carboxyl intermediate ( * COOH) via the formation of a CO 2 radical (CO 2 * ) from which CO is produced when CO 2 * is strongly adsorbed and stabilized; weakly adsorbed or free CO 2 * leads to the formation of HCOO * [93]. Recent studies based on DFT calculations proposed the formation of * COOH through a proton-electron transfer to CO 2 [92,105]. The electroreduction of CO 2 to methane follows the CO-mediated reaction pathway, involving the protonation of CO and the C-O bond scission in successive steps, formation of * CH species, and finally CH 4 by the proton-electron transfer process [106]. However, there is a disagreement about the intermediary CH 4 precursor. Hori [89] suggested a hypothetical intermediate * COH is the precursor to CH 4 , whereas, * CHO intermediate is found in more recent studies [94,106]. The CO 2 electrolytic reduction to methanol follows a similar reaction pathway to that of methane, involving the adsorption of formed CO and formation of * CH 2 OH species. This species then produces methanol via proton-electron transfers. Generally, in the formation of C 1 products, the conversion of CO 2 to CO precedes the adsorption of CO on the Cu surface which is considered as the first reaction step in many theoretical studies [94].
The electroreduction mechanisms of CO 2 into multicarbon products are more complex. Reaction pathways to C 2+ products, especially ethylene and C 2+ OH, have been widely studied, and several mechanisms have been proposed [94,[107][108][109]. One of the mechanisms, according to Koper and co-workers, follows the CO dimerization pathway, in which the rate-determining step is the coupling of two CO molecules mediated by electron transfer to form * C 2 O 4 . Subsequently, this dimer is transformed into C 2 O 4 and EtOH by proton-electron transfer [107]. Based on this, Cheng et al. [108] devised a more complete mechanism in which hydrogenation of an oxygen atom ( * COCO+H → * COCOH) follows CO dimerization. However, using the CEP model, Goodpaster et al. [109] found that the reduction of CO to * CHO followed by reaction with * CO to generate * COCHO was preferred to * CO dimerization and subsequent reduction at high potentials. A closely related mechanism was reported by Bell and co-workers [94]. The adsorbate CO is firstly hydrogenated to * CHO intermediate, and subsequently couples with another CO molecule to form * COCHO. In this pathway, the C-C bond is formed, and various products (e.g., ethylene, ethyl alcohol, and aldehyde) are produced via proton-electron transfers. Over Cu (100) surface, at -1 V with reference to the RHE, the first reaction step was the CO adsorption of the surface with a good binding strength due to charge donation from the surface to the empty π * orbital of CO. This was followed by the C-C bond formation via the following reactions to * COCHO, * CO+H → * CHO and * CHO+CO → * COCHO ( Figure 4) [94]. Lin et al. [95], by computational method, explored the mechanistic steps in ethanol formation on the Cu-Cu(I) ensembles and found the various hydrogenated intermediates to start from the hydrogenation step of the * CCO generated from the dehydration step of the observed * OCCOH intermediate. The pH influence on the reaction mechanism has also been demonstrated and discussed, and variation in the pH of the system could lead to the formation of a different product [109][110][111]. The fact that ethylene formation is pH-dependent on a RHE scale implies that the rate-determining step proceeds in without a proton [111]. Also, that C-C bond formation is the rate determining step via the * CO → * CHO → * COCHO pathway justifies the observed pH dependence [109]. For the combined pathway of CO dimerization followed by reduction to * COCOH ( * CO → * COCO → * COCOH), the C-C coupling step is the rate-determining step, too, because the reaction is also pH sensitive. This is also the case if the * CO → * COH → * COCOH sequence is considered [110]. Therefore, the combined evidence suggests the existence of pH-sensitive mechanisms for C-C bond formation on Cu.
Cheng et al. [108,112,113], using full solvent quantum mechanics molecular dynamics, studied the mechanism and identified the reaction intermediates during the operando electrocatalytic CO 2 reduction to CO. It was predicted that the reaction involves physisorbed CO 2 (l-CO 2 ), chemisorbed CO 2 (b-CO 2 ), formation of * COOH, and * CO at pH 7 [113]. The characteristic peaks listed in Table 1 are IR active but Raman inactive. The spectroscopy of CO 2 reduction to CO reported from experimental investigations exhibited debated peaks assignment; however, the b-CO 2 , * COOH, and * CO 3 2all exhibited comparable spectroscopy fingerprint signals at 1400 cm -1 , difficult to distinguish them [114][115][116][117]. The possible reaction intermediates and products of electrochemical reduction of CO 2 on Cu electrode were studied using in situ ATR-SEIRAS and isotopic labeling [118]. Results found that formate and carbonate, such as bidentate species, were formed at negative potential. The more negative potential shifted the selectivity towards CO and bicarbonate. The formation of CO became significant upon the appearance of the CO 2 dimer radical anion by disproportionation reaction. Adsorption of CO was restricted to -1.4 V, below which no significant adsorption was feasible; only residual bi-carbonate was present. One issue with this characterization technique is the transient nature of some species, which might escape measurement due to their short lifetimes.  [112,113].

Types of Cu-Based Catalysts
The most selective catalytic materials for this process are based on Cu. Several Cu-based catalysts have been investigated for the direct or indirect transformation of CO 2 to important chemicals and fuels [119]. In our previous review [10], we discussed Cu-based catalysts for the CO 2 reduction to methanol. Cu catalysts in the modified or supported form are very efficient for methanol synthesis in the thermocatalytic process. In this follow-up work on determining the structure and active sites of these catalysts, we briefly discuss different kinds of Cu catalysts applied in converting CO 2 by different processes discussed in the previous sections. The Cu-based catalysts can be roughly classified into four types: (i) Cu/CuO x nanoparticles, (ii) supported Cu catalysts, (iii) Cu bimetallic and multimetallic systems, and (iv) oxide-derived Cu (OD-Cu) catalysts [120].
Cu/CuO x NPs are synthesized mostly by chemical, photochemical, electrochemical, sonochemical and thermal synthesis methods [119,121,122]. Pure Cu NPs can be prepared by the chemical method on inert support that has no contribution to the intrinsic activity of the catalyst or interaction with the active Cu (e.g., SiO 2 ) [123]. Very small Cu NPs with a narrow size range often result when a higher Cu loading is placed on the active supports. Cu NPs and the corresponding bulk materials exhibit different activities because of their differences in size and shape. They typically perform CO 2 reduction by different processes; however, their activity is depended on the particular process adopted. The use of pure Cu NPs is limited by their intrinsic instability under atmospheric conditions and ability to oxidize easily. Cu 2 O NPs of different morphologies can similarly be prepared by the chemical method via the dissolution of Cu precursors at a low temperature in the presence of structure-directing agents [124]. Such oxidic catalysts have pronounced activity in photoreduction reactions. This type of catalyst is unstable in this reaction; however, using suitable supports, such as carbon materials, can remarkably improve the stability.
(ii) Supported Cu catalysts. Cu NPs sit on active supports exhibit superior properties compared to the bare Cu NPs, partly due to the increased stability of the particles by tuning their sensitivity to physical and chemical conditions such as oxygen, water, and other chemical entities. This, in particular, has enabled tailoring the properties of NPs in general toward specific applications. Several materials are useful as supports for Cu catalysts. They include carbon materials, metal oxides (e.g., ZrO 2 and Ga 2 O 3 ), polymer materials, silica, zeolites, etc. [99]. This support can combine with the active metal to form interfaces highly active as sites selective to a product. For instance, over the Cu/ZrO 2 catalyst for CO 2 conversion, in which ZrO 2 is the support, the formation of both CO and CH 3 OH occurred on the surface and Cu-ZrO 2 interface, respectively [123]. Cu NPs can also be supported on an oxide of Cu (CuO x ) to tune the selectivity of the product. For example, the deposition of metallic Cu NPs on Cu 2 O films was demonstrated to change the product selectivity from gaseous products to CH 3 OH on the Cu/Cu 2 O interface in the photocatalytic conversation of CO 2 to methanol in aqueous solution [128]. The Cu/Cu 2 O interface can be adjusted by loading different masses of Cu on the Cu 2 O layer and allowing it to evaporate. Cu 2 O supported on reduced graphene oxide, prepared by a one-step micro-waved assisted chemical method, showed enhanced activity for the photoreduction of CO 2 when compared with an optimized bare Cu 2 O or that supported on RuO x [60]. The improved catalytic activity is attributed to the efficient charge separation and transfer and the stability and protection offered by the reduced graphene oxide.
(iii) While the interaction between metal NPs and support may be restricted to the surface, some metals can combined with Cu to form multimetallic systems such as alloys [125,126] and intermetallics [127]. These systems differ from supporting materials because the foreign element can penetrate the lattices of the base metal, altering the electronic and geometric properties. Such materials often exhibit unusually high activity, physical and chemical properties, and thermal stability.
(iv) Oxide-derived Cu (OD-Cu), generally denoted as CuO x /Cu, is a group of Cu-based catalysts that can be obtained by a variety of methods, including oxidation of Cu foils or films, electrochemical reduction of copper oxides, and oxidation-reduction treatment of Cu [128][129][130][131]. For example, the synthesis of Cu 2 O-derived Cu was achieved through the electrochemical reduction of Cu 2 O on carbon paper substrate [132,133]. According to XRD data, the Cu 2 O films exhibited predominantly (111) facet but largely retained their morphology with rough surface and cracks at grain boundaries. The OD-Cu catalyst is often associated with changes in morphologies compared to its precursor metal/oxide, which benefits its performance [134]. Mostly utilized in the electrochemical reduction of CO 2 , the OD-Cu catalysts show an overall higher activity for producing C 2+ products such as ethylene and ethanol at lower overpotentials than Cu NPs and have high Faradaic efficiency (FE) for CO reduction [90,96,128,[134][135][136]. The former is due to the presence of residual oxygen, which is debated in the literature [137]. Although the OD-Cu catalyst exhibits better FE for reducing CO to alcohols, it is more difficult to effectively and directly reduce CO 2 to products than CO and formate. The formation of heavy carbon products would proceed through subsequent CO and/or formate reduction [138]. Zhuang et al. [139] synthesized and characterized an OD-Cu catalyst for the electroreduction of CO to C 3 alcohol fuels. The Cu 2 O NPs were first formed via the nucleation and growth of nanocrystals, followed by a mild acidic etching that resulted in the formation of Cu 2 O particles with open morphology [140]. The obtained Cu 2 O particles were deposited onto a carbon substrate and the final nano-hollow Cu catalyst was produced via an in situ CO electrochemical reduction. Another Cu material with comparable with but simpler chemistry that the OD-Cu is the polycrystalline Cu (129,130). This material is often used as working electrodes in the electroreduction process. It exhibited a good activity and selectivity for CO reduction at low potentials in KOH solution [141]. Generally, Cu electrocatalysts form highly reduced products from CO 2 and more selective to C 2+ due to their intermediate binding of CO [142].

The Nature and Structure of Active Sites
In any catalytic reaction, the catalyst employed should have appropriate activity, selectivity, stability, and be affordable, mainly depending on the kind of active sites it has. For heterogeneous catalysts, active sites could be referred to as atoms, ensemble of atoms, crystal faces or areas on the catalyst surface that directly catalyze a reaction [103,143,144]. In regard to CO 2 reduction, the active sites must be characterized by the ability to activate CO 2 and form bonds to yield product precursors or intermediates. Moreover, the active sites should not be too active for dissociating H 2 into atomic adsorbed H * [145].
Different Cu catalysts possess different active sites. Cu NPs present Cu species as active sites for hydrogenation of CO 2 to chemicals [146]. Recent experimental investigations have shown the Cu NPs are active in converting CO 2 and CO to alcohols [147,148]. In particular, the grain boundaries of the NPs resulted in excellent catalytic performance. Typically, the * OCCOH intermediate formation was feasible on the sites with an under-coordinated surface near a subsurface stacking fault according to the multiscale simulation method [147]. The under-coordinated sites are more active than the low index surface, such as facet, edge, or corner sites in CO 2 electroreduction [149].
Combination of Cu with another metal and/or metal oxides proves an invaluable strategy to enhance the effectiveness of the active sites [146]. The Cu NPs supported on active support materials can exhibit different active sites from the pure NPs or that supported on an inert support. For example, the active sites of Cu/ZnO-based catalysts explored in the CO 2 to methanol synthesis have been a subject of interesting discussion. It has been proposed that the Cu-Zn interface, Cu-Zn bimetallic surface, Cu-Zn alloy or Zn decorated in Cu are all active for CO 2 conversion. The possible active sites or structures of some Cu-based catalysts are discussed subsequently. How the interfacial contact between Cu and Characterization and the Nature of Active Sites ZnO is constituted is an important factor that determines the character of the active site [125]. The OD-Cu catalyst reduces to metallic Cu during electrochemical reduction of CO 2 [150]. Moreover, CuO x exists on the catalyst surface and responsible for the adsorption of * CO 2 and OCOspecies. The oxide species is crucial in activating CO 2 and in C-C coupling according to studies using ambient-pressure X-ray photoelectron spectroscopy (AP-XPS) and electron energy-loss spectroscopy (EELS) [150,151]. Each of these active sites plays defining role in the overall reduction process.
One of the fundamental hindrances in developing a rational design protocol for the CO 2 reduction catalysts, as observed in most heterogeneous catalysts, is the complexity of the resulting catalyst system that complicates the detailed characterization of their active sites [125]. Thus, a proper identification of the structure and composition of the active site is essential for comprehension and improvement of catalyst properties [152][153][154]. Generally, the real active sites of the Cu-based catalysts for CO 2 conversion remain a heated topic of discussion in the literature with catalytic activity attributed mainly to Cu 0 , Cu + or both species and Cu-metal/metal oxide interface [10,33,155,156]. For example, different types of active sites exist on the conventional (Cu/ZnO/Al 2 O 3 ) methanol synthesis catalyst: the Cu-Zn interface, metallic Cu sites, and ZnO exposed sites [157][158][159]. Moreover, the composition of the different chemical state of Cu, which most times are the active sites, can vary. As reported, the composition of Cu 2+ and Cu + during CO 2 transformation varied over a range depending on factors such as synthesis method, presence of promoter and matrix, and extent of reduction [160].
The complexity of the nature of active sites will be increased due to the synergistic interaction between the support and metallic Cu particles [161]. Synergy can result in the formation of new active sites [36,[162][163][164]. For example, the addition of ZrO 2 to Cu/ZnO catalyst promoted a new Cu phase formation as observed by H 2 -TPR experiments [36]. In addition to Cu + that formed upon reduction of the catalyst, a peak that emerged at about 423 K indicated the formation of a new CuO phase, which is thought to result from the dissolution of zirconium ions in the copper oxide phase. From both experimental and theoretical investigations of ZnCu and ZnO/Cu model systems, surface oxidation of ZnCu occurred under the reaction conditions such that Zn on the surface transformed into ZnO, allowing ZnCu to attain the activity of ZnO/Cu with the same Zn coverage [164]. It demonstrates the importance of the synergistic interaction of Cu and ZnO at the interface that facilitates methanol synthesis.
The synthesis of methanol by Cu NPs can be promoted by ZnO. It has been demonstrated from both experimental and theoretical calculations that Zn atoms migration in the Cu surface promotes methanol synthesis [163]. The methanol synthesis activity was found to be highly dependent on Zn coverage, which was determined to relate to the reaction conditions and size of Cu and ZnO NPs ( Figure 6). Thus, the size-dependent activities of NPs reveal that a synthesis strategy that can regulation the size of NPs can be adopted to design synergetic functionality in binary NP systems. [163].

Figure 6. (a) Modeling of Zn coverages of Cu NPs and (b) relative methanol turnover numbers of the Cu/ZnO/Al2O3 catalyst as a function of ZnO and Cu particle sizes
The chemical and structural changes of Cu catalysts may affect their stability and selectivity toward certain reaction products while identifying the catalyst active sites provides insights into the mechanisms of the catalytic processes. Generally, Cu-based catalysts present three major active sites for CO 2 reduction reactions, besides other specific sites reported in the literature. The results of detailed theoretical and experimental investigations on these active sites using state-of-the-art characterization techniques are discussed. [152,[165][166][167].

Functions of Various Cu Species at different Oxidation States (i) Metallic Cu
Metallic Cu (Cu 0 ) plays a fundamental role in the adsorption and activation of H 2 , which spills over to the support, providing H atoms to form intermediates. Metallic Cu sites adsorbed and dissociated H 2 into adsorbed H * atoms for reaction with adsorbed CO 2 at adjacent In 2 O 3 sites on a Cu-In-Zr-O catalyst [162]. Cu 0 was found to be the active site on single crystal Cu (100), Cu (110), and polycrystalline Cu films with exposed Cu (111) facets [161,168,169]. From the HRTEM image of reduced Cu-In catalysts on zirconia that clearly marked the formation of three different phases with lattice corresponding to the (111), (111) and (011) planes of Cu 0 , CuO and ZrO 2 , the subsequent mechanistic study identified Cu 0 was active for methanol synthesis [170]. Cu 2 O with different orientations and thicknesses was deployed in electrocatalysis. The result showed catalyst to be active after reduction to Cu 0 and the selectivity was dependent on the thickness of the oxide [171]. Cyclic voltammetry and electrochemical mass spectrometry evidenced CO 2 reduction only upon Cu 2 O reduction to metallic Cu. As observed with in situ Raman spectroscopy, Cu 2 O-film reduced to Cu metal at low potentials in aqueous CO 2 -saturated 0.1 M KHCO 3 [172]. Mandal et al. [173] who combined the DFT study and the in situ Raman with real-time product detection reported a similar observation. It showed that Cu 2 O reduction typically precedes CO 2 reduction because the former was energetically more favorable. After reducing Cu 2 O to Cu, adsorbed CO, the intermediate in CO 2 reduction, was detected. Copper-complex materials (copper (II) Phthalocyanine, CuPc, HKUST-1 and [Cu(cyclam)]Cl 2 )), according to in situ and operando XAS under the working conditions, structurally reconstituted and formed ~2 nm metallic Cu nanoclusters, which catalyzed the conversion of CO 2 to CH 4 during electrocatalytic CO 2 reduction. This was confirmed by DFT calculations which concluded that the restructuring behavior was attributed to the reversible nature of the metal ion-ligand coordination in the structure of copper complex and the generated small size copper clusters under the reaction conditions [174]. An in situ XPS experiment revealed that Cu 0 existed on the surface of CuZnZrGaY catalyst before and after reaction. There was no evidence for the existence of higher oxidation state species as confirmed by the Cu (LMM) Auger spectrum of the catalyst [33]. The CO band frequency attributed to the metallic Cu surface and to oxidized Cu δ+ charged states for some Cu-based catalysts in Table 2 varies due to the presence of different matrices, which act as supports. ZnO as support uniquely influences by shifting the adsorption to lower frequencies, attributed to some sort of interaction with Cu. Moreover, such bands are less intense in comparison with those without the presence of ZnO [175]. The presence of subsurface oxygen was found to modify the electronic structure of Cu and enhance the adsorption of CO, resulting in the existence of Cu in the zero oxidation state according to an investigation using in situ ambient pressure XPS [137]. This means that the presence of sub-surface oxygen influences the activity of Cu.

(ii) Cu-interfacial Sites
The copper-metal/metal oxide (Cu-MO x ) interface has been characterized as active sites for the CO 2 conversion to chemicals [123,146,186,[189][190][191][192][193]. Theoretical studies on Cu/ZrO 2 catalysts showed high activity for intermediates formation at the Cu-ZrO 2 interface [123,[189][190][191]. Liu and co-workers [189,190] found CO 2 to methanol transformation to be more stable and effective on the Cu/ZrO 2 interface where both the Cu and Zr atoms formed bonds with the reactions species [190]. Polierer et al. [191] found that key intermediates: HCOO, H 2 COO, H 2 COOH, and H 3 CO were linearly correlated with the oxygen's adsorption energy. Behrens et al. [157] found the active sites of a Cu/ZnO catalyst to consist of Cu steps decorated with Zn atoms and stabilized by a series of well-defined bulk defects and surface species. According to the DFT study, Zn-decorated Cu site was more favorable for activating CO 2 to form intermediates and their subsequent hydrogenation to methanol because of its much lower energy barrier of key intermediate step than the pure Cu site [186]. These studies show that Zn and Zr modified the reaction routes in these catalysts compared to the surface of pure Cu. Zn dissolved Cu particles was found as an active sites for methanol production [176,194]. Laudenschleger et al. [177] recently found the active site on the conventional CO to methanol catalyst (Cu/ZnO/Al 2 O 3 ) consisted of non-metallic Zn species with an associated positive charge nature (Cu 0 -Zn δ+ ) under industrial reaction conditions. This active site was investigated by probing the surface of the catalyst with both molecules of "N" reacting species (ammonia, NH 3 ) and the one that blocks the N species (trimethylamine, TMA), applying an in-house built high-pressure pulse method. The catalyst composition, such as the Cu: ZnO ratio, Cu and ZnO contact, and synthesis route, affected which type of surface sites would be exposed for reaction. These factors could be optimized to maximize the Cu-Zn interface sites [177]. XRD and EDS provided important evidence for the migration of ZnO x from ZnO to Cu particles to form interfacial Cu-Zn alloy upon reduction with H 2 [167]. The role of ZnO x is to stabilize the Cu species on the surface, although evidence is only provided for the stabilization as Cu + [167,176,195,196]. Such surface species of Cu stabilized by ZnO x was found to promote methanol synthesis during the hydrogenation of CO 2 [152]. Figure 7 shows the nature of the active sites on the Cu/ZnO catalyst as reported by Tisseraud et al. [152], where the active ZnO x covers the Cu core. The Cu/ZnO/Al 2 O 3 catalyst in the reduced state investigated with HRTEM showed an agglomeration of Cu, ZnO, and Al 2 O 3 NPs in Characterization and the Nature of Active Sites proximate contact [163]. The Cu lattice fringes present extended to the surfaces of the NPs without obvious changes in their spacing (d = 2.1 Å) and structure, an observation that indicates the Cu surface in direct contact with the gas environment and covered with a submonolayer of Zn NPs.
Cu-Zn interface was found to be active for both CO 2 adsorption and activation. Also, this surface was actively involved in the production, adsorption and desorption processes of the reactive H 2 CO * species for methanol synthesis [197]. Investigated active sites in other Cu-based catalysts also play relatively similar or slightly different roles. For example, on the La/Cu-SBA-15 catalyst, the Cu-LaO x interface promoted the adsorption of CO 2 and key intermediates species. The formation of Cu-In alloy on the surface of Cu-In bimetallic catalysts helps to suppress H 2 formation on Cu during CO 2 electrocatalysis.
Bimetallic surfaces have also been identified as the active sites for many Cu-based catalysts, as listed in Table 3. The Cu-Zn bimetallic surface is of several orders of magnitude higher than that of the single Cu surface or the pure ZnO catalysts, resulting from the Cu-Zn synergistic effect [186,198]. As generally held, Cu is the active site for the Cu/ZnO catalyst, while ZnO modifies Cu and enhances its activity. It has been made evident from recent studies that Zn alloy is hardly stable on Cu (111) and Cu (100) at high temperatures (above 500 K) at which practical CO 2 hydrogenation is carried out. In this case, the Zn-Cu interface formation takes precedence and is beneficial for binding on the catalyst surface and converting CO 2 into intermediate formate species stable on the catalyst surface at high temperatures [196]. Generally, the role of Zn is to improve the performance of Cu catalysts via one or more of the following ways: (a) act as a promoter to promote the structure and modify Cu; (b) enhance the activity, and (c) provide a reservoir for atomic hydrogen storage and use [157,199]. Metallic Cu NPs on Cu 2 O films changed the CO 2 electroreduction product distribution from gaseous to liquid (mainly CH 3 OH) on pure Cu 2 O. The product change stemmed from the modification of the Cu 2 O surface and the formation of Cu/Cu 2 O interfaces, which provided sites for binding of H * and CO * intermediates in methanol production [187]. The Cu/Cu 2 O used as the cathode in the photochemical cell exhibited a good FE for methanol production. CuO was fully reduced to metallic Cu in the reduced Cu-ZnO catalyst [165]. The Cu crystallite size was used to estimate the metallic Cu surface area, although some researchers apply the chemisorption method using N 2 O as a probe molecule to calculate the surface area [200]. The Cu surface area (S Cu ) was inversely proportional to the Zn content, but this did not correlate with catalytic activity. It was argued that if the metallic Cu was the active site, TOF versus S Cu should be constant [201]. Opposing, the TOF increased with a decrease in the S Cu , revealing that Cu metal was not the active sites [165,202]. It has been established that since the metallic Cu surface could not explain the catalyst activity, synergism between the catalyst components (Cu and ZnO) should be relied upon to explain the active site formation. With increasing Zn concentration, the interaction between Cu and ZnO caused the lattice parameter of Cu to expand, indicating an increase in Cu lattice strain and the creation of the Cu-Zn alloy [39,176]. Kanai et al. [176] observed the formation of Cu-Zn brass-type alloy in Cu-ZnO catalysts reduced at between 250 and 550°C due to the expansion of the lattice Cu when the ZnO content increased. There was a corresponding decrease in the lattices constant of ZnO with decreasing Zn content. The Zn content in the alloy can be determined from the Cu lattice constant. Generally, the content of Zn in Cu-Zn alloy is high at high reduction temperatures [176].

Figure 7. (a) Model of the methanol synthesis core-shell catalyst, (b and c) TEM image of Cu@ZnOx catalyst; spots on (c) show presence of Cu and Zn, (d) model of the selective methanol synthesis nano-core-shell catalyst, (e and f) TEM image of Cu@ZnOx/ZnO catalyst, (g) Model representation of active sites for
methanol and CO production from CO2 hydrogenation over a co-precipitated Cu-ZnO catalyst, pathways to methanol synthesis, RWGS, and methanol decomposition are indicated in red, blue, and yellow arrows, respectively [152].

(iii) Cu 2 O (Cu + )
Studies have shown that Cu + functions in adsorption and stabilization of the intermediates in CO 2 reduction reactions [151,195,203,204]. TPR, and XPS and Raman spectroscopies revealed that Cu exists as Cu + species that enable adsorption of CO 2 and serve as active sites for activation of CO 2 molecules on CuO-CeO 2 catalyst in the electroreduction of CO 2 to ethylene [195]. In OD-Cu catalysts for CO 2 electroreduction, Cu + sites are good CO-binding sites. On Cu/TiO 2 system, it has been suggested that the active sites for CO 2 photoreduction and methanol production are mainly Cu + species [69]. Using in situ XANES, Cu + was detected after CO 2 electroreduction to C 2+ products commenced on OD-Cu catalyst [151,204]. The residual Cu + was crucial to the selectivity of C 2 products. It was found that methanol synthesis on oxidized Cu (100) was higher than that on clean Cu (100) [205], indicating that the Cu ion was the active site. From this study, the mechanism for CO 2 hydrogenation to methanol involves: a. The adsorption of CO 2 on support. b. Generation of intermediate species.
c. The dissociative adsorption of H 2 and reaction with intermediate to form methoxy species on the Cu surface. The CO bond frequency is highly sensitive to the charge on the adsorption site of Cu catalysts [175,206]. Adsorption of CO was observed on reduced Cu/ZnO/SiO 2 and Cu/SiO 2 catalysts [206]. In both catalysts, Cu + species were not stabilized by the ZnO component. A high CO adsorption was recorded on Cu + sites while the adsorption on Cu 0 sites was moderate. The Cu/ZnO/Al 2 O 3 catalyst showed CO adsorption band at frequencies (2065-2094 cm −1 ) [38,175,207,208]. Upon surface coverage with adsorbate such as formed formate during CO 2 reduction to methanol, electron transfer (withdrawing effect of formate) takes effect, leading to an upward shift in the CO stretching frequency. The molecular CO can form carbonyls (Cu + -CO) with IR bands in the range 2160-2080 cm -1 upon strong adsorption onto Cu + sites [167,209]. The formation of stable carbonyls with Cu 2+ ions at ambient temperature has not been reported [210]. In a study by Kanai et al. [167], the CO adsorption sites were assigned to Cu + -O-Zn species. When the Cu/ZnO catalysts were exposed to the reactant gas atmosphere (CO 2 -H 2 ), Zn oxidized by CO 2 to Cu + -O-Zn, which usually contained ZnO x moieties that have been identified as active sites for methanol synthesis [152].
A Cu/Pt-TiO 2 co-catalysts prepared by the stepwise photodeposition technique of Pt NPs and then Cu onto TiO 2 revealed the formation of a core-shell catalyst with Cu covering the Pt core according to the HRTEM imaging. Both the irradiation time and the Cu content were found to affect the catalyst structure. The lattice fringes with an interplanar spacing of 0.211 nm revealed by the HRTEM analysis suggested the formation of Cu 2 O on the shell, that is, Cu existed in the Cu + state [203]. Further characterization using XRD revealed diffraction peaks belonging to the TiO 2 . The XPS binding energy of Cu 2p 3/2 for the catalysts at about 932.5 eV suggested the presence of Cu 0 and Cu + . With the Cu (LLL) Auger spectrum, only Cu + was observed, which corresponded to the HRTEM result. However, it should be noted that Cu 0 is easily oxidized to Cu + under experimental conditions or on exposure to air [211,212].
The in situ synchrotron powder diffraction (SPD) experiments were conducted to determine the active sites and understand their dynamic changes during electrocatalytic reduction of CO 2 over Cu/graphene and Cu/graphene-Ar [121]. During a 90 min CO 2 reduction, the results showed characteristic peaks of Cu 2 O (002), Cu 2 O (222), Cu (111), Cu (002) and Cu (002) in the diffraction patterns of Cu/graphene and Cu/graphene-Ar. It was observed that graphenesupported Cu showed a feature of CuO phase that directly transformed to the Cu (111) phase. Both Cu 2 O and Cu phases stabilized on the Cu/graphene-Ar. This study identifies Cu 2 O as the active Cu species in electrochemical CO 2 reduction to hydrocarbons because CuO exhibited poor adsorption for CO 2 molecules, and hence, less active for CO 2 reduction. Also, CuO phase was directly reduced to metallic Cu rather than Cu 2 O. Hence, the superior CO 2 reduction performance of Cu/graphene-Ar was attributed to its ability to form a higher content of surface Cu 2 O species during its synthesis and maintain it under relevant reaction conditions and the easy transformation of CuO to metallic Cu [121].

(iv) Mixed/Other Cu Species
The mixture of Cu + and Cu 0 can be beneficial for the adsorption and splitting of H 2 . XPS measurements revealed Cu/TiO 2 and Cu/Pt-TiO 2 catalysts contained Cu either in the chemical state of 0 or +1 [211]. Previous data showed that catalytic products might change as a result of catalytic surfaces with varying compositions of Cu 0 and Cu + ; surface with higher content of Cu 0 led to CO, while that with mixed Cu 0 and Cu + resulted in C 2 products such as ethanol [95]. The mixed oxidation states of Cu can improve reactivity and selectivity toward C 2+ products by way of enhancing CO 2 adsorption on one site and C-C coupling on another site [213]. For example, the C-C coupling reaction of two * CHO species can occur on the Cu 0 -Cu δ+ atomic interface with a low kinetic barrier of 0.57 eV in the formation of ethanol [213]. According to Tseng et al. [214], a relationship exists between the chemical states of Cu and methanol production on Cu/TiO 2 catalysts which increased in the order of Cu 2+ > Cu + > Cu 0 through formate and methoxy-intermediates on the Cu sites, facilitating the electron transfer from the catalyst surface [215,216]. Compared with bulk crystalline Cu 2 O with Cu-O bond lengths of 1.86 Å, the Cu-O bond lengths of 1.92-1.95 Å was calculated around the CO 2 and H 2 O adsorption sites, characteristic of the DFT calculated Cu-O of 1.95 Å in bulk crystalline CuO [124]. XANES and EXAFS are reliable techniques for characterizing the electronic states and local structure of Cu and ZnO in Cu/ZnO-based catalysts (e.g., CuO/ZnO, CuO/ZnO/Al 2 O 3 , and CuO/ZnO/Al 2 O 3 /ZrO 2 ) by measuring the Cu K-edge and Zn K-edge regions [217][218][219]. The Cu K-edge XANES spectra of CuO/ZnO/Al 2 O 3 /ZrO 2 mixed oxide catalysts exhibited XANES spectra similar to that of CuO at 8984 eV, suggesting the presence of Cu 2+ species [217]. Under working conditions, the nature of the active catalyst remained unchanged despite an obvious change in methanol yield. This observation needs further investigations to rule out the possibility of missing data. The suboxidic species have also been detected experimentally [95,[220][221][222] and EXAFS characterization confirmed an average Cu-O coordination number of 1.1 for suboxidic Cu [222]. Because of reaction intermediates and surface polarization generated by the applied electric potential, most Cu-based catalysts reconstruct under reaction conditions [223,224]. Species specified by their coordination and charges occur upon reconstruction, regardless of the starting structure: Cu 0 and oxidized Cu (Cu δ+ and Cu + ) [103,213,220]. The Cu δ+ on Cu 0 surface is critical to the activity and selectivity of electrochemical reduction of CO 2 [220].
During CO electroreduction on OD-Cu catalysts, high CO reduction activity was linked with surface sites that bound CO more strongly than low-index and stepped Cu facets, attributed to the metastable sites from the disordered surfaces at the grain boundary and defect terminations stabilized by the interconnected nanocrystalline network [141]. By reducing mixtures of 13 CO and 12 CO 2 , Lum et al. [225] showed that OD-Cu catalysts possess product-specific sites for C-C coupled products. Three different types of active sites were identified, each for a different product. In a follow-up study combining theoretical and experimental techniques, planar-square and convex-square sites were identified as the active sites for ethylene production, while step-square sites were responsible for generating C 2+ alcohols [226].

Catalyst Structural Characterization by Advanced and Integrated in Situ Techniques (i) A Simple Comparison of Various Characterization Techniques
One of the major challenges impeding the rational design of heterogeneous catalysts for CO 2 transformation is the structural complexity, impairing our comprehension of the reaction mechanism and catalyst design efforts [158,197]. Researchers commonly deploy various characterization techniques to determine the structure-activity relationships of the Cu-based catalysts, including theoretical, microscopic (imaging), spectroscopic, and chemisorption methods [33, 124, 158, 188-190, 209, 217, 227]. These tools can offer key information of the catalyst structure, such as the active sites and the intermediates species of CO 2 reduction reactions. Recently, the accuracy of the information on the catalyst structure has improved with the development and deployment of advanced characterization and theoretical methods. The emergence of in situ/operando characterization techniques is one of the exciting developments in catalysis research in the last several decades that have revolutionized catalysis and facilitated measurements under controlled environments and practical catalyst working conditions. These techniques include X-ray diffraction/scattering (XRD), X-ray photoelectron spectroscopy (XPS), X-ray absorption spectroscopy (XAS), transmission electron microscopy (TEM), and other methods developed and deployed in unraveling the structure-activity relationships, reaction mechanisms and for understanding the reaction evolutional and dynamic changes in real-time [227][228][229][230]. With these developments, it is becoming easier to investigate the catalytic surfaces under reaction conditions. For example, in situ STM can image ceria islands of about one layer thick, exhibiting rough surfaces with a CeO 2 (111) termination in the CeO 2 /Cu 2 O/Cu catalyst for methanol production from CO 2 in real-time process [166]. In situ surface-enhanced infrared absorption spectroscopic (SEIRAS) can selectively probe species adsorbed on the metal surface to a depth of <10 nm [140].
It takes a very short time (a few minutes) for the metal oxidation state to reach a steady-state under the CO 2 reduction condition; thus, a time-resolved determination of the oxidation state of metal centers, especially at a time-scale of a few seconds, should be pursued [95]. For example, CuO x precursor would attain a steady-state condition in a few seconds according to the data from in situ Raman spectroscopy in electrocatalysis due to the simultaneous occurrence of electrochemical reduction and spontaneous oxidation [231]. Therefore, analysis of such catalysts under real-time is recommended. The application of some in situ techniques is limited due to special operation requirements. For example, in situ ambient pressure X-ray photoelectron spectroscopy (APXPS) requires ultra-high vacuum (UHV) condition to function properly. However, using the synchrotron radiation for APXPS can overcome this limitation [229]. The invention and application of APXPS represent a key advancement in the in situ XPS technique. Other techniques have witnessed advances too, such as the TEM in which the development of environmental TEM (ETEM) has enabled the in situ characterization of structural evolution of single-atom catalyst [227]. It is now possible to obtained sub-Angstrom's resolution imaging and analysis by using image or probe aberration-corrected TEM instruments. In situ TEM can reveal growth transformation and function in nanomaterials [232]. Further advances should rely on developing new in situ techniques combining multiple complementary characterization techniques that simultaneously compare results. Computational methods have also been utilized for analyzing the catalyst structure, including DFT, kinetic Monte Carlo simulation, and microkinetics modeling [145,186,191]. DFT, the most widely used approach, uses electron density functions to determine ground-state energy and properties derived from it. It has the advantage of low computational cost and being able to calculate larger systems like metal NPs and periodic surfaces. Moreover, results from theoretical studies are used as a validation of the experiments, thus can compensate for experimental deficits. DFT calculations have been applied to elucidate the roles of Cu and ZnO species at the Cu-ZnO interfacial sites, and the participation of the Cu-Zn sites over metallic Cu (211) surface and Cu-Zn alloy in methanol synthesis [157,186]. Combining the theoretical studies in real-time with state-of-the-art in situ characterization techniques is an innovation the field of Characterization and the Nature of Active Sites research is urgently needing. To help understand the different functions of various characterization techniques as applied to Cu-based catalysts for CO 2 reduction, we summarize and compare these techniques in Table 4.  Figure 8. Operando characterization of a single Cu2O particle photocatalyst (a) Setup showing the gas-flow nanoreactor, (b) Schematic of the electron beam and X-ray directions for TEM and SFXM imaging on a catalyst particle, (c) TEM of a catalyst particle [124].

(ii) Catalyst Structural Characterization by Integrated Methods
The characterization of single particles of a catalyst is possible with the use of advanced characterization techniques such as environmental transmission electron microscopy (ETEM) [233] and scanning fluorescence X-ray microscopy (SFXM) [124]. Photocatalytic Cu particles have been investigated with these techniques during the photoreduction of CO 2 [124,234]. Wu et al. [124] studied the local crystal structure of a single Cu 2 O particle used as catalyst for the photochemical reduction of CO 2 under typical working conditions. The catalyst was exposed to a gaseous mixture of CO 2 and steam, and irradiation source as arranged in Figure 8. The edges of the catalyst particle were found to correspond to the (220) planes by single-particle electron diffraction (SPED) (Figure 8c). The spatial location of the particle was determined using low-resolution scans employing Cu fluorescence signals from Cu 2 O particles. A specific facet of the particle can be focused by the X-ray beam parallel to desired facets while scanning of the incident X-ray energy, allowing facet-dependent spectroscopic information on the Cu active sites to be obtained. A white-line peak at 8981.0 eV was found in the X-ray fluorescence spectra recorded at the Cu K edge on the (110) facet of a Cu 2 O particle, indicating a Cu + oxidation state associated with this facet. The spectral peak observed on the (100) facet at 8981.5 eV indicates the existence of both Cu + and Cu 2+ charged states. Furthermore, information on the change in the oxidation state can equally be obtained. For example, the peak shift of between 1.0 and 1.5 eV towards higher energy from low energy is an indication of the change from +1 to +2 oxidation state of Cu or vice versa depending on the exposed conditions. With the aid of this technique, it was revealed that the (110) facet Cu 2 O photocatalyst particle was the active sites for CO 2 reduction to methanol during photocatalysis while Cu 2 O (100) facet was photocatalytically non-reactive. The ETEM images of Cu NPs on ZnO under different gas mixtures revealed the Cu NPs to be flat-like and spherical under reducing and oxidizing conditions, respectively [233]. That is, the particles changed shape (reconstructed) with changes in gas composition. High-resolution X-ray powder diffraction (HRXRD) and EPR were further applied to characterize the photocatalytically active sites of the Cu 2 O particles ensembles. The results revealed the obvious structural changes over the course of the reaction, which was monitored with the operando high-energy XRD. Data from this measurement showed that in the absence of light, the co-adsorption of CO 2 /H 2 O (g) increased the crystal lattice constant, confirming the electron density withdrawal effect of CO 2 on Cu active sites which caused an oxidation state transition from Cu + to Cu 2+ [124]. Photocatalytic reduction of CO 2 to liquid fuels by a photoinduction method was utilized to form Cu single atoms on the UiO-66-NH 2 support (Cu SAs/UiO-66-NH 2 ) [235]. The catalyst local structure was measured with XAFS which found the absorption edge positions located between that of Cu and CuO, suggesting a positively charged Cu state between 0 and +2 [235]. It was found that isolated Cu atoms were fixed at the atomic level in the UiO-66-NH 2 matrix and were coordinated to two N atoms with atoms distance of 1.97 Å (Cu−N). Confirming using the Fourier-transformed (FT) EXAFS, one main peak at 1.50 Å in the FT-EXAFS spectrum was observed, corresponding to the first coordination shell of Cu−N. A similar result was observed with the wavelet transformation (WT) plot of Cu SAs/UiO-66-NH 2 , which showed the WT maximum at 3.8 Å −1 , ascribed to Cu−N bonding. The peak at 2.3 Å, corresponding to Cu−Cu, was lacking in the case of Cu SAs/UiO-66-NH 2 . From both measurements, there was no observation attributed to Cu−Cu for Cu SAs/UiO-66-NH 2 . The catalyst achieved a photocatalytic conversion of CO 2 to CH 3 OH and C 2 H 5 OH with an evolution rate of 5.33 and 4.22 µmol h −1 g −1 , respectively.
Mistry et al. [151] investigated the surface species and structural changes in the oxide layer during CO 2 electroreduction by applying operando XANES and EXAFS. The results in Figure 9 revealed the prominent shoulder at ~8,982 eV on the XANES spectrum of the plasma-oxidized Cu, a feature of Cu 2 O. The existence of metallic Cu and Cu oxides in the catalyst can be clearly observed in the EXAFS spectrum (Figure 9 b and c). Upon the partial reduction of the film, XAFS spectra showed a dominantly metallic Cu underlayer signal. This technique can show without doubt the presence of the different Cu species viz-Cu 2 O, CuO, and metallic Cu, although the oxidic species were evident only for the oxidized catalysts. While reduction of the oxide phases was also visualized, reconstruction was such that the Cu 2 O remained highly stable, and thus it was concluded that Cu + , which was supplied from oxide surface and subsurface layers during the reaction, was the active reaction phase.

Conclusion and Future Perspective
This report reviews the Cu-based catalysts for the CO 2 valorization, the key information about the nature and possible structures of the active sites in them. Also, it compares the techniques developed for their characterizations. The obtained information about the catalyst active sites can guide the catalyst development and help understand the real catalysis on the catalyst surface.
Capturing and transforming CO 2 into useful products is an efficient and economical way to reduce carbon footprint and achieve near-neutral CO 2 emission. Promoted and/or supported Cu catalysts is efficient for thermochemical CO 2 conversion to products like methanol and methane. On the other hand, these catalysts are less efficient when light or electricity are applied for CO 2 conversion. CuO x oxides are more active catalysts for the photochemical process. Cu supported on CuO x or its inverse referred to as oxide-derived catalyst (OD-Cu) shows superior activity for conversion CO 2 in the electrochemical process. The performances of the different kinds of Cu catalysts depend on the quality and quantity of the active site in them.
Active sites play essential roles in accurately determining the reaction mechanisms and general properties of heterogeneous catalytic reactions. The activity, selectivity, and stability of a catalyst are all influenced by the kind and nature of active sites. For CO 2 conversion, the Cu active sites include but are not limited to metallic Cu, Cu 2 O, and Cu-metal/Cu-metal oxide interfaces. Underexplored structures of the Cu surface, such as suboxidic Cu surface, can drive CO 2 reduction. The mixed active sites can even drive further the selectivity toward different products. Therefore, engineering of such active sites on the same surface apparently can improve the selectivity of a specific product.
The dynamic behaviors during catalytic processes make the identification and investigation of the active sites still challenging, often because of the reconstruction of the surface by reaction intermediates and polarization caused by reaction conditions. However, significant progress has been achieved in the last two decades, thanks to state-of-the-art characterization techniques and advanced computation methods. These include the following strategies developed and deployed. (1) Development of individual in situ and operando techniques that are highly sensitive, accurate, and precise for characterizing the structure and active sites under ideal working conditions can further reveal the real active sites of Cu catalyst systems. For example, X-ray microscopy can map the active states of nanocatalysts under typical relevant reaction conditions. (2) Several advanced techniques, e.g., real-time product detection, theoretical calculations, and operando surface spectroscopy, can be simultaneously deployed. Raman spectroscopy, for instance, can be combined with selected-ion flow tube mass spectrometry (SIFT-MS); this strategy can in real-time examine the catalyst surface and detect the multicomponent products generated in situ during a reaction. Furthermore, SIFT-MS can simultaneously quantify multi-component products with finite vapor pressure in a split of seconds (typically 0.1 to 10 sec). This technique allows to capture reaction dynamics early enough and eliminate issues associated with the analysis of gaseous products of CO 2 reaction with conventional gas chromatography that needs longer analysis time and waiting till the end of the reaction. Furthermore, with these measurements, the dynamic structures of the active sites can be understood further, which can be used to guide the rational design of novel catalysts and modify the existing ones for optimum performance.
Developing novel high-performing catalysts for CO 2 conversion is needed as part of controlling CO 2 emission efforts. Although many catalysts have been developed, they are not good enough for the targeted catalytic conversion reactions. Tailoring active sites on a catalyst could drive the selectivity to a given product. Maximizing the sites selective to the target product and minimizing those to the undesirable product can be an important strategy to achieve this aim. For example, the design of Cu particles covered with ZnO x of thickness greater than the diffusion sphere of H 2 to inhibit the RWGS reaction is suggested. Core-shell systems are well-suited structures that can fulfill these specifications, such as a Cu@MO x catalyst. Additionally, the size of the core-shell particles must be controlled to reach a high dispersion and high surface area of the exposed ZnO x active phase. This division could introduce function-specific roles such that only reaction for the formation of desirable product is promoted over undesirable product formation. It remains to be shown whether practical strategies exist for further increasing the step/defect density in Cu NPs. In general, the future Cu catalysts for CO 2 conversion should be well-engineered at the surface towards a specific reaction product to eliminate the need to control the reaction conditions to improve catalyst activity.