Recent Progress in Thermal Conversion of CO2 via Single‐Atom Site Catalysis

CO2 emission has been an international issue of great concern. Utilizing of CO2, especially converting it to value‐added products, is widely investigated, among which the thermal conversion of CO2 has enormous potential for industry. Researches on single‐atom site catalysis (SAC) have become increasingly systematic during the last years. High performance and distinctive selectivity that SAC exhibits attribute to the isolated structure a large extent. To understand the structure–performance relationship of SAC in CO2 activation, issues including substrate, active components, coordination, chemical structure, etc. are of the essence not only in academia but also in industry. However, it is far away from the vision that the synthetic procedure and reaction pathway are deeply comprehended, thereupon precise single‐atom sites with specific structure are constructed and elementary reactions are regulated at will. Still a lot of efforts are needed to this field. Herein, CO2 reduction reactions are reviewed according to the products, and then catalysts are introduced by the substrate. The promoter, stability, synthesis/regeneration, characterization, and theory calculation issue related to SAC and CO2 activation are comprehensively summarized and discussed. Looking back the progress, challenge and outlook of single‐atom site catalysis are also proposed.


Introduction
The current CO 2 emission rate hovers about 35 gillion tons annually ( Figure 1a), with the predominant emissions from fossil fuels. [1] Despite that land and ocean each sink about one-quarter of the emission, there is still nearly one-half entered the atmosphere. Even more, this trend increases dramatically since 1950s. Reducing the greenhouse gas CO 2 emissions is a common goal of all mankind. Using fewer fossil fuels is a direct and significant method but impractical. Although the energy gap can be replaced by clean energy, the chemical industries are still highly dependent on petroleum products. Carbon capture and storage/ utilization (CCS/CCU) serves as important way for reducing the CO 2 emissions. [2] Beyond the storage mean by injecting CO 2 to deep underground, utilizing CO 2 as the potential feedstock of chemical industry, replacing the nonrenewable petroleum has profound implications.
Methods for of CO 2 utilization include biotechnological approaches via traditional crops or marine algae to produce fermentation products (ethanol, etc.), biofuels and bioplastics, and more constructive catalytical approaches. Among thermo-, [3] photo-, [4] and electrocatalysis, [5][6][7] the thermal catalytic approach has enormous industrial potential, through which CO 2 can be reduced to CO as a syngas component for Fischer-Tropsch (F-T) synthesis or CH 4 as composition of natural gas.
What is more, directly obtaining high value products, e.g., formic acid, methanol, and dimethyl ether, or C 2þ products is highly desired.
Hydrogen is the most feasible reductant to convert CO 2 to value-added industrial feedstock for CCU system. To quantitatively assess the actual impacts to the environment, the life cycle analysis must be considered. Some CCU systems may still emit more CO 2 than non-CCU system in extreme cases. [8] Currently, mass production of hydrogen still depends on fossil energy via water-gas reaction, water-gas shift reaction, or reforming. By using green hydrogen from renewable energy, CCU technique generally decreases the carbon footprint compared to the traditional petrochemical industries. However, the cost of green hydrogen produced from electrolysis and photolysis of water is still uncompetitive if no carbon tax adopted; meanwhile, the robustness and efficiency of catalysts still need to be improved for these reactions. It should be noticed that even the production of green hydrogen causes slight carbon footprint. Reactions with low energy efficiency will cause double disadvantages, both poor CO 2 emission reduction and unnecessary energy dissipation. Due to the complexity of elementary reactions and high activation energies, catalysts are of vital importance in catalytical approaches of CO 2 utilization. As the research has become increasingly systematic, single-atom site catalysis serves as a potential industrial catalyst. Here, we review the recent progress of the SAC for CO 2 activation (Figure 1b).
SAC was first proposed in 2011 [9] and the concept was quickly adopted and expanded to widened scope of materials in heterogeneous catalysis. A variety of materials are used as the substrate by now [10,11] not limited to carbon materials [12][13][14][15][16][17][18] and oxides, [19] such as transition-metal dichalcogenides (TMD), [20] carbides, [21,22] coordinated polymers, [23,24] MXenes, [25] etc., and the elements of supported atoms have spread all over the periodic table. [26][27][28] Related to the basis "isolated atom active sites, not closely connected with each other but interacted with the substrate," some concepts with subtle differences were also commonly used, such as single-atom site catalysis, single-atom catalyst (SAC), single-site heterogeneous catalyst (SSHC), atomically dispersed catalyst, and dual/tri-atom catalyst (DAC/TAC) for the specific condition with dimer/trimer of isolated atoms, which all have diverse or even unexpected catalytic properties compared to the pervious nanocatalysts. Although there are many different concepts, the underlying factors are common. The isolated atoms generate the discontinuous electron orbitals and the multiformity of the localized atomic site-substrate environments [29] behind the size effect, providing the tunability of electronic structure. These distinctions set SACs apart from the bulk, nanoparticles or clusters type of catalysts. Due to the variable valency [30] and the different coordination environment, [31][32][33][34][35][36][37] the SAC may generate more than one type of reactive sites. [38][39][40][41][42] Some CO 2 activation reactions such as alcohols and C 2þ production have complex elementary reactions and demand various active sites, which could be achieved by SAC or coexisting [43] catalysts via synergy effect by carefully designing the structure and precisely adjustment of the isolated atoms. In most cases, we use the common expression SAC to refer all above concepts, regardless of the emphasis to specific definitions. For coexisting materials, quasiatomic dispersed materials or suspected samples but not stated by the author, we will also discuss partially, at the same time indicate their status.

Reactions
CO and CH 4 are the most facile and simple product in CO 2 reduction (Figure 2), proved by thermodynamics data and high proportion of researches. Directly using hydrogen as reductant, the CO formation reaction and methanation reaction are commonly known as the reverse water-gas shift reaction (rWGS) and Sabatier reaction, respectively. Both CO and CH 4 are lowpriced, mainly obtained from fossil fuels, and their major application is gaseous fuel currently. Industrial production using rWGS or Sabatier reaction seems not a good idea unless forbidding fossil fuels, especially for the quite high energy dissipation of Sabatier reaction. Worse still, CH 4 is a far stronger greenhouse gas than CO 2 ; meanwhile, CO is toxic and hard to liquefy. [44] Handling the storage and transportation problems will limit its widespread use. CO and CH 4 are also obtained as byproducts in value-added CO 2 reduction reactions. CO 2 could assist the dehydrogenation reactions as a weak oxidizer, improve the selectivity and avoid the formation of coke, with reduction products of CO. Dry reforming of methane (DRM) uses the abundant natural gas to reduce CO 2 and produce the mixture of H 2 and CO, also known as the syngas. Apart from the fuel use, CH 4 could be used  in fuel cells and the syngas is important feedstock for Fischer-Tropsch (F-T) synthesis. Methanation prefers a low reaction temperature with the more complex elementary reactions, and rWGS is a simple and reversible reaction with the equilibrium constant (k) positively related to temperature. The CH 4 product exhibits strong size effect and weak relations to the types of substrates or metals. The prevailing view has been that single-atom sites prefer CO production via a carboxyl pathway and reject methane production. Even if CH 4 appears in the products, the reasons are explained as the active metal atoms aggregating to clusters or particles, which has multiple pathways towards CH 4 including starting from CO and intermediates via formate or bicarbonate. Indeed, most SACs show good selectivity in rWGS. But the CH 4 selectivity could be adjusted by the interaction between the substrates and metal sites. [45] Researches on CO 2 reduced to CO and CH 4 also provide insights of mechanism, which share similar intermediates and regulation strategies to other type of CO 2 activations. Inducing value-added products remains big challenge and is crucial for industry.
Methanol and its dehydrated production dimethyl ether (DME) are the most cost-effective reaction with mature catalyst system so far. The methanol economy is proposed by Nobel Prize winner George A. Olah [46,47] and is widely recognized. Apart from energy applications such as hydrogen substrate, internal combustion engine fuel, and direct methanol fuel cell (DMFC), methanol has the potential to be the next-generation precursor in industry. Through methanol-to-olefins (MTO), CO 2 could be converted to ethylene and then polymerized to hydrocarbon products, which currently rely on nonrenewable petroleum or F-T synthesis. The mature Cu/ZnO/Al 2 O 3 methanol production catalyst using syngas also applies in hydrogenation of CO 2 with considerable conversion. The business demonstration factory has been operated in Iceland. But the rWGS by-reaction is significant and the more water from CO 2 hydrogenation induces the deactivation of the catalyst. The state of art SACs could be used in methanol production with better water resistance, high selectivity, but poor conversion. Defects on the interface of substrate with proper basicity enhance the adsorption of CO 2 , forming formate intermediate and dissociation of H 2 at active metal sites. Precious metal especially Pd and Pt exhibits good activity among variable substrates. In particular, In 2 O 3 substrate is favorable to methanol production, avoiding production of CH 4 . Comparing to selectivity or yield in experimental condition, space-time yield (STY) in industrial condition is more crucial, which also highly demands the stability of SAC. Methanol may innovate the intermediate system of current chemical industry. SACs could be the optimal candidate for catalyzing CO 2 hydrogenation to methanol or DME.
Formic acid is another high value C 1 product and could be thermally decomposed back to hydrogen without using external water, which make it a promising hydrogen carrier. Formic acid is easy to handle with low toxicity and flammability, and the relative high density ensures a volumetric hydrogen capacity of 53 g L À1 . [48] Although the rather low mass hydrogen capacity of 4.35 wt% does not meet Technical System Targets: Onboard Hydrogen Storage for Light-Duty Fuel Cell Vehicles, US Department of Energy, formic acid as hydrogen carrier still has potential application in heavy equipment, factories or strategic reserves field. Fuel cell using formic acid is also developing. [49] Due to the ΔG limitation, formic acid hardly forms in a direct manner for gas phase reaction. In solution, the CO 2 and H 2 are dissolved or reacted with the solvent, which significantly changed the chemical thermodynamics. Semihydrogenation products including formic acid and alcohols could be obtained through liquid phase reactions with high selectivity and quite mild condition. But product separation process could be the big problem and the base is always required for improving the activity, thus causing high cost and extra steps. Currently, several precious metal-based homogeneous catalysts are reported and applied in aqueous phase formic acid (formate) synthesis by CO 2 (carbonate) hydrogenation. SACs show similar turnover frequency (TOF) comparing to homogeneous catalysts with the advantages easy to separate and recycle. Some multistep routes are developed to avoid liquid phase reaction or the usage of base. Using novel solvents such as ionic liquid and supercritical CO 2 is also promising high activity and base-free formic acid formation reaction with potential for industry. Similar to methanol formation catalysts, substrates providing suitable basicity and enough basic sites accompanied with precious metal, such Ru and Pd, for hydrogen activation have better performance in aqueous formate synthesis. [50] Owing to the mild condition in formation of formic acid, continuous CO 2 hydrogenation synthesis via slurry bed also has industrial potential and low implementation difficulty, promoting its hydrogen carrier application.
Directly obtaining C 2þ products from CO 2 hydrogenation has enormous attraction in industry. C 2 products such as ethanol and ethene are formed by C-C coupling via insertion of formyl or species like CO*/CH*, which demands various active sites. SAC with designed adjacent atoms has been reported and shows good activities in ethanol formation via CO 2 hydrogenation. SAC could play important role in synthesis of liquid hydrocarbon and higher alcohol through tandem reaction. The single-atom sites are able to boost part of the elementary reactions. [51] In fine chemicals field, cyclic carbonates and carbamates could be obtained by CO 2 cyclization or addition reaction, which used to be catalyzed by homogeneous catalysts and could be replaced by SACs. Yet, still lack of the researches focused on these reactions.

Catalysts for CO 2 Activation
Dispersing the active component, usually the precious metal, to atom sites allows reducing the usage, thus cutting down the cost of catalyst by maximizing the atomic utilization. Most CO 2 reduction reactions activate under gas phase with hydrogen as reductant, which is compatible with existing plants but also produces considerable water and CO. These byproducts have strong migration ability to isolated atoms, leading to sintering and deactivation. The relative high temperature, strong reducing and migration atmospheres have been a big challenge to the stability of SACs. The substrates are the first to be considered, which play important role not only in anchoring the atoms, but also in providing specific chemical environment. Their diverse capabilities could regulate the catalytic activities. At present, the most reported SACs used in CO 2 thermocatalysis are supported on oxides, which is totally different from the electrocatalysis with majority of carbon-based materials. [52] Since most studies focus on gas phase reaction with react at high temperature, oxide substrates provide ultrahigh thermal stability, but with relative low surface area and limitation in coordination adjustment. The weaker bonding of atoms in carbon materials better suits the liquid phase reaction. Some new types of substrates are also adopted, such as sulfide, hydride, MOF, polymer, etc. exhibiting the diverse chemical environments and shapeselectivity through reticular chemistry. SACs for CO 2 activation are summarized according to their substrates.

In 2 O 3
Pure In 2 O 3 could effectively catalyze the hydrogenation of CO 2 to methanol with high selectivity even at room temperature but with quite low activities. The surface of In 2 O 3 provides oxygen vacancies, as well as the adjacent In ions dissociate the hydrogen molecules heterolytically and form formate species as the intermediate. [53] The adding of transition metals serving as the promoter could improve the activities due to the synergistic interactions. Additionally, In 2 O 3 -based catalysts show good water resistance and bring high durability in hydrogenation of CO 2 .
Shrotri's group compared the STY of methanol catalyzed by transition metals in group 8-10 doping in the In 2 O 3 matrix. Among these metals (Fe, Ru, Co, Rh, Ni, Pd, Pt), the Rh sample exhibited the highest STY. [54] The existence of Rh 3þ changed the chemical environment of In 2 O 3 . The partially reduced In 2 O 3 stabilized the Rh 3þ ions through the charge transfer and avoided sintering under industrially relevant conditions, indicated from the H 2 temperature programmed reduction (TPR) and X-ray photoelectron spectroscopy (XPS) spectra. Rh ions promoted oxygen vacancies for CO 2 adsorption, observed by the CO 2 temperature programmed desorption (TPD) of the used catalyst.
Recently, Li's group reported a coexisting catalysts with SAC and cluster form of Pt embedded in the lattice of In 2 O 3 for methanol synthesis with a selectivity up to 91.1%. [53] The XPS results indicated a relation between the chemical state and methanol selectivity. Single-atom site Pt species were in the cationic state and maintained a majority of þ2 valence state for the 0.58 wt% loading sample during the reaction, thus improving the methanol production. The selectivity was stable for more than 60 h with conversion decreasing slightly in the initial time for the stability test. In contrast, a 2.50 wt% loading control sample with Pt nanoparticles prepared by wet impregnation had a favor of rWGS and metallic Pt detected after reaction.
Pérez-Ramírez's group studied the promotion effect with Pd for In 2 O 3 in CO 2 hydrogenation. [55] Series samples with different Pd loading were prepared by coprecipitation and dry impregnation synthesis methods and compared (Figure 3a-c). According to the extended X-ray absorption fine structure (EXAFS) and time of flight-secondary ions mass spectrometry (TOF-SIMS), Pd was doped into In 2 O 3 matrix when prepared, and transformed into nanostructure with configurations of Pd 2-3 In 2 at the beginning of the reaction, thus improving the methanol selectivity, which fitted well with theoretical elucidation. The dispersion state appeared virtually unaltered for the coprecipitation sample after the long-time reaction, but the dry impregnation sample sintered significantly.
For other nonprecious metal dopants, Hensen's group reported a Ni-In synergy catalytic system prepared by flame spray pyrolysis with controlled Ni-In interactions via varying the ratio for CO 2 hydrogenation. [56] The low Ni loading promoted methanol formation and the Ni species were reduced to small clusters; meanwhile, partition of the Ni retained as isolated ion form with valence state of þ3 during the CO 2 hydrogenation reaction. A 5 wt% Ni loading sample processed the similar STY to Pd sample, outperforming the performance of Co and Cu system. Due to the higher rWGS activity and relatively low pressure, the optimized Ni-In catalyst encountered a slightly lower selectivity. The STY along with the selectivity fell with Ni loading increasing, and CH 4 appeared in the products when the content is above 75%.
Huang's group reported Ir 1 -In 2 O 3 materials for ethanol synthesis in gas-liquid reaction by a wet chemistry synthetic method. [57] As indicated in the scanning transmission electron microscopy (STEM-HAADF) and EXAFS (Figure 3d-f ), ultralow content of Ir (0.04 wt%) was loaded on partially reduced In 2 O 3 with abundant defects on the surface, forming single-atom structure without particles. The oxygen vacancies served as distinct sites. CO 2 was first reduced to the CH 3 O* active intermediate. Simultaneously, the isolated Ir and adjacent oxygen vacancies formed Lewis acid-base pair as catalytic centers and reduced CO 2 to CO* intermediates, evidenced by density functional theory (DFT) calculation and infrared spectra. Then the C-C coupling could be achieved after both steps. Higher temperature and metallic form of Ir from high loading led to the formation of methanol ( Figure 3g).
As shown in Table 1, compared with other substrates, In 2 O 3 is specialized alcohol production favoring. It is the intrinsic properties of In 2 O 3 originated from the adjacent oxygen vacancies structure. The doping of transition metal, especially precious metal, induced partially reduction of In 2 O 3 , thus increasing the density of defects and promoting CO 2 adsorption. The H species are adsorbed and dissociated at metal sites, improving the reaction rate of the elementary step. The synergy between SAC and In 2 O 3 promotes high efficiency and potential in industrial methanol production. The In 2 O 3 -based catalysts still face some problems. It is a trade-off between reaction temperature and methanol conversion; as the temperature improves, the conversion increases but the selectivity decreases significantly. The anchoring of atoms is quite weak as revealed by the aggregation after reaction. Another problem is that the abundance of In is relatively low in earth crust. It is even lower than the rareearth elements, and accordingly, the cost is rather high. Using In 2 O 3 as the promoter and supported on other compatible oxides may cut the cost.

CeO 2
CeO 2 is a reducible rare-earth oxide and important substrate in heterogeneous catalysis. It has good stability and abundant oxygen vacancies [58,59] due to the existence of Ce 3þ . CeO 2 could be thermally reconstructed and traps migrant SA. Strong metalsupport interaction (SMSI) is widely achieved between CeO 2 and transition metals and promotes strong tunability for chemical environment. Size effect is a key factor for nanocatalysis. Amal's group compared Pt SAC and clusters on CeO 2 for CO 2 reduction. [60] With ultralow loading, single-atom Pt was uniformly dispersed, inducing partially reduced Ce 3þ of the CeO 2 substrate. The Pt species of the SAC sample had an intermediate valence between 0 and 2 after catalysis, presenting strong SMSI. Such Pt SAC had good selectivity toward rWGS reaction and exhibited good thermal stability as high as 500°C.
Ma's group reported Ir/CeO 2 catalysts with tunable SMSI effects, and found that both SAC and small nanoparticles exhibited high selectivity toward CO while large nanoparticle favored methanation. [61] Behind the size effect, the chemical structures of the active metal were found pivotal to the selectivity evidenced by EXAFS, XPS, and EELS. The partially oxidized Ir species had weaker interaction with CO, resulting in inhibition of methanation.
Another SMSI tuning phenomenon was reported by Zhang's group; they investigated SMSI and H-spillover effects in CO 2 methanation catalyzed by Ru/CeO 2 in SAC, cluster, and NP forms. [62] Activation of Ru-CO intermediate and dehydration are the key elementary steps, which is controlled by SMSI and H-spillover, respectively. As indicated in Raman, XPS, and chemisorption analyses (Figure 4a-c), SMSI and H-spillover effects were competitive, and SAC sample had the strongest SMSI but no H-spillover among different size regimes, resulting in low CO activation and enhanced water removal. The cluster sample governed the best methanation activity with a good balance.
Supported Ru catalysts usually produce CH 4 in CO 2 hydrogenation. Interestingly, Cargnello's group observed the restructuring of Ru nanoparticles supported on CeO 2 into SAC during CO 2 reduction, and the CO 2 conversion switched from methanation to rWGS. [63] The redispersion of Ru NPs was induced by oxidation atmosphere in rather low temperature (as low as 210°C). The oxidative pretreatment changed the catalyst structure and  [55] Copyright 2019, The Authors. Published by Springer Nature. d) Schematic illustration of the fabrication procedures of Ir 1 -In 2 O 3 . e) Corresponding microscopy analysis of Ir 1 -In 2 O 3 .Scale bar: 5 nm. f ) EXAFS spectra in r-space with different Ir loadings. g) Selectivity, STY, and TOF Ir values over different Ir-In 2 O 3 during CO 2 hydrogenation at 200°C for 5 h. Yields and selectivity over Ir 1 -In 2 O 3 : effect of temperature and reaction time. Reproduced with permission. [57] Copyright 2020, American Chemical Society.
proofed by in situ XANES and ex situ DRIFTS. The reduction of CO 2 presented preferential CO formation with single-site nature and high coordination number of the Ru active center.
Zhang's group prepared Ru/SnO x dispersed on CeO 2 using [Ru@Sn 9 ] 6À zintl clusters. [64] According to the XRD and XPS results, the chemical environment of the Ru center remained largely unchanged after loading and Sn species were oxidized. The water washed sample exhibited good rWGS properties with >95% CO selectivity and by increasing the water pressure, the reaction switched to %100% methanation with different reactive sites.
Hu's group reported a synergy of Ni and Ru SAC for DRM reaction (Figure 4d-f ). [65] Ru 1 and Ni 1 sites activated the C═O bonds in CO 2 and the first C─H bond in CH 4 , respectively. Their synergistic effect was confirmed by the lowest apparent activation barrier and DFT calculation. Comparing with the Ni or Ru monocomponent catalyst, synergetic dual-metal catalyst exhibited highest CO and H 2 TOF and good coke resistance.
Yan's group compared the formic acid formation of supported Pd catalysts on CeO 2 and ZnO and these possessed different structural sensitivity to bicarbonate hydrogenation (Figure 4g,h). [50] CeO 2 readily activated CO 2 with the high density of basic sites on the surface but was limited by the hydrogenation step. By contrast, formation of the carbonate intermediates was the rate determining step for ZnO samples. So the nature of the Pd species in CeO 2 influenced the activation energy but not in ZnO. Limited by the low hydrogen splitting ability of SAC, metallic Pd and substrates with enough basic sites (such as TiO 2 ) performed better in formic acid formation.
Recently, Liu's group synthesized Pd atom dimers supported on oxygen vacancy-riched CeO 2 . [66] By strictly controlling the pH, Pd complexes were obtained and loaded on CeO 2 nanorods. The dimer structure (DAC) is proofed by the coordination structure from the EXAFS fitting and DFT simulated model (Figure 4i-l). The high homogeneity of reactive centers with unique Pd 2 O 4 structure enabled CO 2 to CO dissociation and subsequent one-step C-C coupling. The C 2þ formation was appropriately inhibited, thus achieving a selectivity of ethanol up to 99.2%. Table 2 provides an overview on the CO 2 activation performances of CeO 2 -based SACs. As a reducible oxide substrate, oxygen vacancies on surface of the CeO 2 provide plenty of anchoring sites for SAC. The atoms could stabilize under high temperature owing to the strong SMSI effects and thermal reconstruction. The oxophilicity and Ce 3þ /Ce 4þ redox cycle of CeO 2 ensure CO 2 adsorption, forming the intermediate. As shown by these researches, chemical structures of CeO 2 -supported SAC are regulated by different synthetic procedure and thermal treatment. It even induces the switchable selectivity from rWGS to Sabatier reaction. CeO 2 has high potential applicated in industrial catalyst as substrates and its properties meet the demand for CO 2 reduction in both liquid phase and gas phase reaction.

TiO 2
TiO 2 has two common phases, anatase and rutile. And it is commercially available with various trade names, containing pure phase or the mixture of both phases. TiO 2 could be partially    reduced in hydrogen-rich atmosphere, producing abundant oxygen vacancies, which also induces thermal reconstruction and strong SMSI effect as the reducible CeO 2 .
Christopher's group studied the size effect of Rh supported on TiO 2 substrate, drawing quantitative relationships. [67] The diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) technique was used to determine the atom site fractions (Figure 5a-c), which decreased along with the increasing of Rh loading. The Rh sites were dynamic and nanoparticles disintegrated during the reaction. Various CO 2 :H 2 ratios were tested, exhibiting entirely different selectivity owing to the competing parallel reaction pathways (Figure 5d).
Combining theoretical modeling and experiment was also adopted in Ir/TiO 2 system by Zhang's group. [45] Coordination number is the key factor for size effect and tuned the rWGS and methanation during the CO 2 hydrogenation. The C-O scission of the intermediates was the rate limiting step of the methanation, influenced by dissociation barrier of metal carbonyls, which was verified by the CO 2 hydrogenation activities of TiO 2 -supported Pt and Au nanoparticles catalysts. Through coordination environment tuning, Ir SAC showed almost 100% rWGS selectivity.
Szanyi's group reported redispersion of Pt on TiO 2 substrate for rWGS. [68] As mentioned in CeO 2 part, chemical environment was the key factor to catalytic properties. A redox cycle was employed for Pt/TiO 2 and transferred the highly O-coordinated Pt species to less coordinated ones with high accessibility (Figure 5e). Nearly threefold rWGS activity was achieved comparing to the fresh SAC. The SAC could be reactivated through reoxidation but large particles were too stable to redispersed in the redox cycle.
Serp's group found regulation of charge transfer important in CO 2 hydrogenation. [69] Transition metal Ni and Ru were stabilized on defective carbon nanotube (CNT) or reduced TiO 2 in blue color. The defects on both substrates were the key for avoiding clusters or nanoparticles. Ni samples showed good rWGS selectivity for both substrates and Ru samples processed an interesting correlation between the selectivity and Ru 3p binding energy of XPS, indicating the tuning effect of electron density. It should be noticed that trace Na was retained in the blue TiO 2 , and the activities of CNT sample improved significantly with Na modification.
Piccolo's group evaluated several Mo-modified commercial TiO 2 for methanol synthesis. [30] Rutile-type samples (trade name: RNR, RL11A) had better methanol selectivity than anatase (trade name: PC500, DT51D) or their mixture (trade name: P25, P90), which was related to the nature of supported Mo (Figure 5f-h). Decreasing oxidation state was observed from anatase to rutile and most Mo ions were reduced to þ4 or þ5 valence. The Mo coverage also influenced the activities with Mo oxo species favoring methanol formation. RNR-supported 3 wt% Mo has the highest methanol STY at 35 g MeOH ·kg cat À1 h À1 among all tested catalysts. Similar to CeO 2 , chemical structures of TiO 2 are highly related to the activities in CO 2 hydrogenation ( Table 3). Besides the regulations from size effect, oxygen vacancies and SMSI effects, the phase of TiO 2 substrates also show important role in the production selectivity. TiO 2 has weak acid sites on its surface, thus the Na modification is found effective improving CO 2 adsorption. TiO 2 has good stability and SAC on it could be redispersed by oxidative calcination, which regenerates the catalysts and benefits in industrial application. Table 4 provides an overview on the CO 2 activation performances based on oxides except In 2 O 3 , CeO 2 , and TiO 2 . Other reducible oxides are also used in CO 2 reduction researches. Llorca's group synthesized Pd/Fe 3 O 4 catalyst with different size regimes and found the SAC sample processed good selectivity to value-added product ethanol. [70] Due to the reaction condition at rather low temperature and even atmospheric pressure, the CO 2 conversion is quite low. Pd single atom sintered and deactivated if the temperature above 350°C, exhibiting a CO favored activity as cluster and nanoparticles samples. Increasing the pressure also decreased the selectivity to ethanol, and produced more methanol, propanol, or CO. Pd/Al 2 O 3 SAC was prepared by the same procedure with CO as main product in CO 2 hydrogenation.

Other Oxides
Oxide-supported SACs are widely used in CO 2 activation not limited to reducible oxides. Inert oxides are found not inert in complex catalyst systems, and serve as critical components. [71] The inert oxides still have the ability to activate CO 2 and influence the reaction by chemical environment adjustment. Szanyi's group compared two inert substrates, Al 2 O 3 and CNT supported Pd for CO 2 hydrogenation (Figure 6a). [72] Two loadings (0.5 and 10 wt%) of Pd/Al 2 O 3 were synthesized and determined as clusters and isolated atoms, respectively. Pd is able to dissociate H 2 , as a result the CO 2 reduction activity was slightly higher with more Pd but the SAC sample produced more CO among the tested temperature. The small fraction of CH 4 may be attributed by partial sintering of the 0.5 wt% Pd, deduced by the authors. But unlike the oxide substrate, CNT could not active CO 2 , thus causing very low activity (onset temperature of %400°C). Indeed, the La 2 O 3 was required as promoter for activation of CO 2 in hydrogenation and the onset temperature dropped down to %200°C, which was almost same as Pd/Al 2 O 3 . A good rWGS selectivity was processed for Pd/La 2 O 3 /CNT, compared to Pd/Al 2 O 3 . Condition was different for Ru/Al 2 O 3 system in their subsequent progress (Figure 6b,c). [71] Ru/Al 2 O 3 with various Ru loading was synthesized using similar method. The 0.1 wt% low loading had mostly atomically dispersion of Ru species, which exhibited high rWGS selectivity at onset temperature of %300°C, which is higher than both Pd/Al 2 O 3 samples of %200°C. Interestingly, selectivity to methanation increased Reproduced with permission. [67] Copyright 2015, American Chemical Society. e) Schematic representation of the evolution of Pt/TiO 2 during the redox cycle. Reproduced under the terms of the Creative Commons CC BY license. [68] Copyright 2021, The Authors. Published by American Chemical Society. f ) Effect of TiO 2 nature on product yields and methanol selectivity. g) Near-ambient-pressure XPS analysis and h) H 2 -TPR profiles for 3 wt% Mo samples. Reproduced with permission. [30] Copyright 2021, Royal Society of Chemistry. significantly in long-term test along with the sintering of Ru, indicating the size effect. But the atomic dispersion of Pd sample could stabilize in oxidation and reduction pretreatment under 500°C. CO was determined not an intermediate of the methanation of CO 2 hydrogenation, and the activation energy of which is lower than that of CO hydrogenation. As for other inert oxides, Liu's group synthesized ZnO nanowire-supported Pd SAC and evaluated several reactions. [73] In CO 2 hydrogenation, the SAC showed rWGS activity and no methanol formation detected. Another Pd/ZnO SAC showed formic acid selectivity in liquid phase bicarbonate hydrogenation, as discussed in the CeO 2 section. [50] Shrotri's group prepared Co SAC doped in ZrO 2 with up to 15 atom% loading for rWGS reaction. [74] The Co ions maintained divalent revealed by EXAFS. Due to the imbalance of charge, plenty of oxygen vacancies formed and promoted CO 2 adsorption as the Co loading increased, resulting in high CO selectivity under wide experimental condition via formate intermediate (Figure 6d-g). Increasing the Co loading to 50% caused the formation of Co 3 O 4 and predominant of methanation activity.
Frei's group replaced 1-10 atom% Ni in MgO matrix by a solid solution approach with isolated dispersion and tendency locating on the surface. [75] The Ni ions favored low coordination number, as proofed by XRF and XPS results and predicted by DFT, which decreased the CO 2 adsorption energy. The selectivity was restricted to rWGS for Ni SAC and as a matter of fact, 2e À redox was applied, while multielectron reaction to methanol or methane required clusters. After Ni sintering at 350°C, CH 4 generated as a byproduct, accompanied with higher CO formation rate in CO 2 hydrogenation, even the temperature was adjusted back.
Liu's group combined theoretical modeling and experiment, tuning the DRM activity in Ni/MgO catalyst. [76] The atomic dispersed sample performed low DRM activity and rWGS occurred as by-reaction. The Ni 4 site could be strongly anchored on MgO with strong SMSI according to the calculation, thus preventing sintering and coke of the active sites during the reaction. The 10 wt% sample had similar size distribution as the 5 wt% one and exhibited high conversion and the H 2 /CO ratio closed to 1.
Lauterbach's group used Co/SiO 2 as models to probe CO 2 hydrogenation. [77] A uniform monolayer with tetrahedral coordination was formed according to XPS and EXAFS and maintained stabilized up to 600°C. The Co SAC had high rWGS selectivity below 550°C but quite low activity. Clusters and large particles were formed by varying the reduction temperature, which tuned the chemical structure of Co species and significantly improved the CO 2 reduction rate. Ternary oxides are also used as catalyst substrates, such as spinel (AB 2 O 4 ) and perovskite (ABO 3 ). Researches about ternary oxide pyrochlore (A 2 B 2 O 7 ) become active in recent years. Yao's group evaluated a Rh-substituted pyrochlore in DRM. [78] By citric acid-assisted calcination synthetic procedure, Ru ions formed solid solution in the La 2 B 2 O 7 oxide and dispersed as single atoms or clusters substituting to the Zr with pyrochlore structure or Ti with perovskite structure at B site, respectively. In DRM reaction, Rh SAC/La 2 Zr 2 O 7 led to rapid carbon, resulting from the fast CH 4 dissociation. In contrast, Rh cluster/ La 2 Ti 2 O 7 had better performance in CO 2 activation, owing to the oxygen vacancies accelerating electron transport, which contributed to the excellent activity and long-term stability of DRM. High entropy oxide (HEO) is a type emerging material, [79] which provides new possibility of metastable substrates of catalysts among enormous possible compositions. Dai's group used mechanochemical method to prepare rocksalt-structured J14 HEO with (NiMgCuZnCo)O composite. [80] Owing to the new entropic tuning strategy, HEO with up to 5 wt% Pt or Ru doping exhibited good activity to rWGS and thermal stability even after 700°C calcination (Figure 6h). Limited to our understanding, the mechanism of high entropic tuning is still hard to explain and unpredictable by far, but such materials endow a new field for heterogeneous catalysis including CO 2 activation.

Carbon-Based Materials
Carbon-based SACs are widely used in electrocatalysis, especially for N-doped carbon. N-contained groups provide strong anchoring ability to the metal among the periodic table. [81][82][83] But there are rarely reports on carbon-supported SACs for the CO 2 thermal reduction comparing with the oxide substrates. Some preliminary results are listed here ( Table 5). CNTs are often used as reference to show the effect of SMSI, such as La 2 O 3 -promoted Pd/CNT reported by Szanyi [72] and defective CNT-supported Ni or Ru reported by Serp's group, [69] already discussed in TiO 2 and Other oxides part, respectively. Indeed, the CNT substrate is inert for CO 2 adsorption and activation. The SACs with transition metal supported on CNT only exhibit poor activity if no promoter existed in rWGS reactions. Ji's group used a mechanochemistry method (Figure 7a) preparing DACs on N-doped carbon (NC). [84] The metal and interatomic distances were controllable by mixing nitrogen-doped porous carbon and various binuclear organometallic complexes. The synthetic procedure could be extended to FePd and FeNi DACs. The Ni 2 DAC had a distance around 3.7 Å (Figure 7b), and keep mostly unchanged after catalysis despite partial Ni species sintered to cluster. The DAC structure remarkably improved the conversion in CO 2 hydrogenation and maintained good rWGS selectivity comparing with Ni SAC sample. Figure 6. a) Schematic representation of CO 2 reduction activities of different supported Pd catalysts. Reproduced with permission. [72] Copyright 2013, American Chemical Society. b) TOFs of CO 2 , CO, and CH 4 as a function of time-on-stream over Ru/Al 2 O 3 catalyst and c) corresponding microscopy analysis c1-2) before and c3-4) after CO 2 reduction. Reproduced with permission. [71] Copyright 2013, American Chemical Society. d) Operando DRIFTS experiment over the 10 wt% CoZrO x catalyst and corresponding e) formation rate of CO. f ) Peak positions of formate species and g) fitting for CO evolution in formic acid TPD over CoZrO x catalyst. Reproduced with permission. [74] Copyright 2021, American Chemical Society. h) Schematic representation of mechanochemical synthesis of high stability Pt/HEO for rWGS. Reproduced with permission. [80] Copyright 2019, American Chemical Society. Kim's group synthesized Co-Fe single-atom alloy (SAA) derived from N-doped carbon-supported Co SAC for CO 2 F-T synthesis (Figure 7c-g). [51] Fe salt was impregnated on Co SAC/NC and SAA was induced in the following calcination. DFT results indicated the suppressing effect of methanation for FeCo alloy.
Stepwise synthetic procedure could stabilize SAA and inhabit migration of Co. Although the strong anchoring from pyridinic groups of the substrates, the SAA catalyst prepared via coimpregnation still dealloyed after the CO 2 hydrogenation reaction.
Carbon-based materials seem inadequate in thermocatalysis for its inertness toward CO 2 , unlike the oxides in which defects and lattice oxygen could assist the reactions. Another important factor is stability. The SACs anchored by M─N or M─C bonds trend to mirage and the carbon supports may be instable in the catalytic condition due to the Boudouard reaction with onset temperature of %400°C. [69] Despite these disadvantages, some theoretical researches [85,86] also prove that in relative low temperature, carbon-supported SAC has the potential for the reduction of CO 2 into high-value production.

Single-Atom Alloys
Even small fraction of doping atoms in alloy could significantly change its property. [87] SAA is an expanded concept of SAC and also has potential application in CO 2 activation (Table 5), as revealed by theoretical researches [88][89][90] and experiment from Sasmz's group. They reported Pt-Ni SAA and CeO 2 cores confined in nanotubular SiO 2 structure for DRM. [91] Ce species had strong interaction with both components in Ni-Pt SAA and may enter the lattice evidenced by the HRTEM and TPR. The oxygen vacancies improved the Ni dispersion. No Pt─Pt bonds were detected in 0.25 wt% Pt concentration. The confined structure is similar to yolk-shell and exhibited excellent resistance to carbon deposition for over 120 h. Mechanism researches further support the properties of SAA. Pd SAC on Cu surface was found having lowest dissociation barrier and promoting hydrogenspillover. [92] Single Pt atom on Cu(111) promotes rWGS reaction, evidenced by in situ XPS and the produced CO molecules also prevent the diffusion of isolated Pt atoms to subsurface. [93] SAA is thermodynamically stable under the reducing atmosphere; the single-atom sites may move but keep isolated during catalysis. Studying SAA catalysts not only helps us to understand the dynamics of nanoparticle type catalysts toward CO 2 activation, but also expects to develop industrial catalysts.

Porous Materials
Porous materials have unique advantages in catalysis. Its high surface area is benefit for anchoring SAC and its pores and channels could tune the products via shape-selective catalysis. Zeolites are inorganic material and widely used in industrial catalysis.
De Jong's group developed a 13X zeolite confined Ni catalyst for CO 2 methanation and CeO 2 was added as promoter. [94] CeO 2 assisted in the dispersion of Ni species but the isolated atoms were not specified according to the energy dispersive X-ray spectroscopy (EDX). TPR indicated strong H-spillover effect of Ni and reduction of CeO 2 (Figure 8a). The 2.5 wt% CeO 2 as promoter decreased both Brønsted/Lewis acid sites and basic sites compared to other additive amount or fresh 13Â (Figure 8b), achieving high methanation selectivity and long-term stability. A balance of acidity and basicity should prevail for adsorption and activation of CO 2 . Ordomsky's group encapsulated Ru in MCM-4 for formic acid synthesis. [95] Isolated Ru 3þ sites were incorporated to the channel wall with coordination of CTAB, forming solid micelle. A proposed mechanism involving that the H 2 molecule heterolytically splitted in Ru sites and formed formate intermediate was suggested by DFT modeling. With the aid of tertiary amine, the catalyst reached high TOF ¼ %143 h À1 in water free condition or high formate concentration of 4 M in aqueous solution of CO 2 hydrogenation.
Reticular chemistry endows the new field of organometallic or organic materials with structure similar to zeolite. MOF is an organometallic complex with micropores. By tuning the secondary building unit (SBU) and ligand, MOFs with different structures and properties are widely reported. SAC could also be loaded on MOFs, [96,97] but the experimental studies about CO 2 activation are still limited. Zeng's group used Cr-based MIL-101 anchoring isolated Pt at ultralow loading by slow syringe pumping. [98] Pt 2þ ions were coordinated with oxygen atoms in the SBU (Figure 8c-e). The SAC had good methanol selectivity via formate intermediate and formic formed as the main byproduct. By simply increasing the amount of Pt and reductant, Pt nanoparticles were obtained, which processed totally different reaction pathway and CO was formed via carboxyl intermediate.
Lin's group synthesized Zr-MOFs with Zr 12 SBUs, in which 11 Cu 1þ ions and 3 alkaline metal ions were loaded on the blinding site of Zr-SBU. [99] The two adjacent Cu sites with short distance of 2.7 Å formed DAC, processing C-C coupling activation (Figure 8f-h). With the basicity of alkaline metal increased, the selectivity to ethanol increased and the Cs sample had >99% selectivity and highest TON among Li, Na, K, and Cs. Apart from the SBU, the ligand could also coordinate metal ions. In another research they anchored Cu and Zn by ligand and SBU in a UiO-67 derivation, respectively, and in situ reduced to small nanoparticles. Such SAC-derived catalyst had almost 100% methanol selectivity. [100] Porous organic polymer (POP) is an amorphous material with pores formed by covalent bonds.
Zhang's group used an aminopyridine POP to fabricate Ir SAC for liquid phase CO 2 hydrogenation with an ultrahigh TON ¼ 25 135 (%0.66 wt% loading). [101] The POP has plenty of C═O and N─H groups, accompanied with strong chemical interaction with noble metal and formed amorphous structure with mesopores of 7.6 nm. The Ir species were partially reduced by NaBH 4 and presented the less positive oxidation state. Ir was Figure 7. a) Schematic representation of large-scale synthesis of DACs by ball-milling approach. b) Microscopy analysis and corresponding intensity profiles of marked areas. EXAFS spectra in r-space of Ni SAC and DAC supported on NC. Reproduced with permission. [84] Copyright 2020, Royal Society of Chemistry. c) Schematic representation of CoFe SAA formation from Co/NC SAC. d) Microscopy analysis of as-prepared and spent SAA catalysts after reaction for 70 h. e) CO 2 conversion, product selectivity and f ) C 5þ product yield of different SAA catalysts. g) Product distribution. Reproduced with permission. [51] Copyright 2021, American Chemical Society.
www.advancedsciencenews.com www.small-structures.com fully reduced if using active carbon or C 3 N 4 supported via same synthetic procedure and formed metallic nanoparticles. The Ir SAC had similar coordination environment to the pincer ligand Ir complexes (Figure 8i,j), exhibiting similar catalytic behavior and high activity as homogeneous catalysis, however combined with good stability and recyclability. All the above porous materials present relative high loading of SACs and good activities ( Table 6), owing to their high surface areas. Although porous materials usually present shape-selectivity in catalysis, which doesn't take effort in CO 2 hydrogenation and partially relates to the small size of the production molecules. But the uniform pore structure provided by porous materials like zeolites, MOFs, and COFs could construct identical sites in each pore units. Porous materials confined SACs not only have great potential in industrial CO 2 activation, but also are ideal models for mechanism researches.

Other Substrates
Some nonoxide compounds are used as SAC substrates. Zhou's group reported an Al-doped MgH 2 methanation catalyst via reactive ball-milling. [102] The isolated Al atoms weakened the Mg-H and provide lattice H for CO 2 hydrogenation through formate intermediate. Comparing with the undoped MgH 2 with the same treatment, the conversion was slightly improved, with fewer byproduct of C-C coupling formed.
Ajayan's group used porous hexagonal boron nitride (h-BN) to support isolated Ru via vacuum filtration. [103] The Ru complex was trapped by hydroxyl groups on mesopores of h-BN and formed bonds with B, N, and O atoms after annealing. An intermediate valence of Ru was detected in Ru/BN, which slightly increased after calcination due to electron-withdrawing effect in Ru─O bond. Comparing to the Ru species with high valence,  [94] Copyright 2021, The Authors. Published by Elsevier. c) Structural model and d) microscopy analysis of Pt 1 @MIL. e) XPS spectra and XANES spectra for MIL-based catalysts after the treatment with H 2 and CO 2 . Reproduced under the terms of the Creative Commons CC BY license. [98] Copyright 2021, The Authors. Published by Springer Nature. f ) Schematic representation of three sites in Zr 12 SBU for Cu loading. g) EXAFS spectrum in r-space for Cs promoted sample. h) Proposed mechanism of methanol and ethanol formation over Zr 12 -bpdc-Cu catalysts. Reproduced with permission. [99] Copyright 2019, Springer Nature. i) Microscopy analysis of AP-POP and fresh/used Ir/AP-POP. j) Proposed mechanism of formate formation over H-bonding Ir SAC. Reproduced with permission. [101] Copyright 2019, Elsevier.
www.advancedsciencenews.com www.small-structures.com the Ru SAC exhibited better activity and selectivity toward methanation. Huang's group reported C 3 N 4 -supported Cu SACs with tailored coordination structures, which led to two distinct reduction products (Figure 9a-c). [104] The Cu-N 4 and Cu-N 3 structures were prepared by simple pyrolysis methods and identified by XPS and EXAFS, with mean Cu valences of þ1.05 and þ1.64, respectively. Such coordination structures processed different reaction pathways (Figure 9d), formate intermediate for Cu-N 4 with methanol productivity, and carboxyl intermediate for Cu-N 3 with rWGS activity in liquid phase base-free CO 2 hydrogenation.
Zeng's group found the neighboring effect of SAC for altering the reaction pathway of liquid phase CO 2 hydrogenation in Pt/MoS 2 system. [20] The ratio of isolated/neighboring/patch of Pt monomers was counted based on HAADF-STEM images. Due to the layered structure of MoS 2 , the apparently adjacent atoms could represent the actual neighboring condition in 3D space. The neighboring Pt atoms decreased the activation energy and achieved higher catalytic activity in methanol production comparing with isolated Pt atoms (Figure 9e-g). Unexpectedly, reaction pathway was different according to in situ DRIFT and DFT modeling. Interaction of Pt atoms sequentially transformed CO 2 into formic acid and then methanol while isolated Pt atom favored the CO 2 to methanol path through COOH* to C(OH) 2 * transformation as intermediates (Figure 9h-i). If the feed ratio of H 2 /CO 2 gas was switched from 3/1 to 1/3, the main product switches to formic acid at the same time.
Tsang's group developed a complex assisted deposition method for preparation of various transition metals supported on TMDs. [105] Taking Fe/MoS 2 as example, the monolayered MoS 2 exfoliated by n-BuLi intercalation had plenty of sulfur vacancies, allowing the SMSI with Fe ions. Only Fe-S scattering was appeared in wavelet transformation spectra of EXAFS for 3-10 wt% loading sample, exhibiting high rWGS activity and selectivity, meanwhile maintaining good stability over 500 h.
Yamashita's group presented isolated Ru anchored on base substrate layered double hydroxides (LDH) for liquid phase CO 2 hydrogenation. [106] Ru formed hydroxide species indicated by EXAFS and its electronic state was tuned by electronic metal-support interaction (EMSI), showing a correlation between Ru 3p binding energy and TON based on Ru. The reaction activity was also related to CO 2 adsorption capacity, which originated from the basicity of LDHs and could be easily tuned by varying the metal components and ratio.
The nonoxide substrates could anchor the metal atoms with enough strength and high loading; meanwhile, the more coordination structures other than M─O bonds extend the chemistry environments for SACs. Nonoxide compounds are beneficial supplement besides oxides and carbon materials, but to a large extent, their properties in thermocatalysis are not clear and undiscovered. Still big efforts need to be made for the structure-performance relationship and industrial application. Table 7 provides an overview on the CO 2 activation performances of SACs based on these less common substrates. The hydrogen activation ability from hydride-supported SAC provides new thoughts for CO 2 hydrogenation. And coincidently, the rest SACs are all 2D layered materials and exhibit good properties in www.advancedsciencenews.com www.small-structures.com CO 2 activation, which represent an ideal platform for SAC research, owing to the high surface areas and defect-rich interface [107] for anchoring SA. As these materials are not conventional substrates in industry, the stability and robustness still need inspection. Further studies on the mechanism of these substrates are necessary and instructive.

Conclusion and Perspective
SACs have been one of the most promising material concepts at the moment, and adopted in variety fields among heterogeneous catalysis. Mechanism studies have confirmed the relation between the high activity and single-atom sites. Despite constructive progress on SAC in CO 2 activation to date, there are still many challenges to be solved. Most studies evaluated specific reactions lack of industrial prospect and thankfully, SACs show high potential in a variety of CO 2 activation reactions not limited to rWGS and Sabatier reaction. More studies should focus on value-added products.
Methanol economy draws the future of chemical industry and olefins or even starch have been obtained from methanol. Moreover, directly synthesizing of C 2þ productions such as ethene, alcohols, and dihydroxyacetone for building blocks of fine chemicals or even carbohydrate could reduce the synthetic steps and improve the atomic efficiencies.
It is challenging to regulate the selectivity of these products during CO 2 activation. SACs provide the opportunity in rationally designing and precisely controlling the active sites in atomic level by tuning the metal-support interactions or coordination environment. More tuning methods could be adopted for SACs. Hybridization and surface modification of various substrates combine the advantages. Additionally, basic sites play a particularly key role in CO 2 activation. Even trace alkali metal ions act as promoter for improving the activity and are considered effective in stability of some metal oxide-supported SACs. In liquid phase reaction, SACs cooperated with organic ligands could also enhance the conversion and selectivity, which could take the place of homogeneous catalysis, with similar chemical structure of the organometallic complex. Developing Proposed mechanism of addition of H* to COOH* over isolated and patch of Pt SACs. Reproduced with permission. [20] Copyright 2018, Springer Nature. promoters for SACs is worth studying and has promising industrial application. Stability remains the crucial issue for SACs. A majority of CO 2 activation reactions process under relative high temperature and high pressure, accompanied with strong reducible atmosphere. Most metal oxide substrates show good durability or beneficially partial reduction under such condition. The metal-support interactions play a decisive role in anchoring the SA, but the side products water or CO may still affect them. The long-term stability of SACs is rarely tested and still far away from the demand of industry. On the other side, the SAC and traditional catalysts are not distinct from each other. Even partial of activities in some traditional catalysts are originated from SAC without being recognized as active sites in the past. [108] Strong migration tendency becomes effective under the reaction condition, which promotes not only the aggregation but also redispersion of active metal dynamically. For example, Pt can be trapped as isolated atoms in reducible substrates via covalent metal oxide bonds. [109,110] If the aggregation is hard to overcome, developing renewable catalysts by thermal reconstruction is acceptable and should be well considered.
Large-scale and universal routes for fabricating SACs need to be developed. So far, strategies like mechanochemical approach (ball-milling) [111,112] and pyrolyzing coordinated polymer [113] have been reported, but most synthetic procedures are not suitable for multisubstrates or unpractically applied in batch production. Meanwhile, processing high loading of SACs can be another challenge toward industry.
The precise structure and catalytic mechanism of SACs largely remain undiscovered yet. The observation of isolated atoms is highly depended by aberration-corrected TEM and X-ray absorption spectroscopy (XAS) technique, which provide apparently localized image or average spectra, [114] respectively. The position and coordination structure of SAC could be acceptably determined and simulated by DFT, yet still unable to distinguish information, such as the types of sites, diversity of coordination bonds, and dynamic condition. More characterizations including Fourier transform infrared spectroscopy (FTIR), electron spin resonance (EPR), and Mössbauer spectra are helpful for structural analysis and should be properly adopted. Structureperformance relationship is of great importance for all catalysts, but it is poorly understood in the field of SACs. More advanced in situ/operando, high-resolution and transient characterizations should be developed and applied to monitor the transformation of SAs and intermediates of catalysis. Moreover, comparing to the time-consuming DFT, the high-throughput machine learning could assist in prediction and screening the SACs, [90] which will hugely reduce the cost to screen a suitable catalyst. The in-depth understanding of structure-performance relationship will promote rational design new types of SACs. CO 2 hydrogenation to value-added liquid products and other lucrative CO 2 activation over SACs are still far away from the satisfaction of industry. Despite this progress, we still believe that the emerging efforts on different directions of SAC fields will finally solve the problems of industrialization and help us understanding the structure-performance relationship originated from the active sites of small isolated atoms.