Coordination Environment in Single‐Atom Catalysts for High‐Performance Electrocatalytic CO2 Reduction

The electrochemical reduction of carbon dioxide (EC CO2RR) is a promising technology to achieve a carbon‐neutral society. EC CO2RR can directly convert the greenhouse gas emitted from stacks into valuable fuels and chemical feedstocks for various industrial applications. Numerous metal‐based electrocatalysts have been researched to reduce overpotential and enhance the product selectivity of CO2RR. Recently, single‐atom catalysts (SACs) are attracting intensive attention due to their low‐cost, extremely high activities per loading amounts, and extensive stability of catalytic active sites due to the strong chemical interaction with coordination atoms. The coordination environments of SACs affect the electronic structure of active sites and change the energetics of the CO2RR pathways. Herein, the principles of EC CO2RR, including reaction mechanisms, figures of merits, and electrolysis systems, are first discussed. Then, the recent progress in the synthesis and characterization of SACs on various supports is accessed. Most importantly, the coordination environments of single‐metal atoms and their influence on CO2RR catalytic ability, product selectivity, and active site stabilization are focused. This review provides a milestone along the design of SACs from the perspective of optimizing atomic configuration surrounding the active sites for EC CO2RR.

DOI: 10.1002/sstr.202200236 The electrochemical reduction of carbon dioxide (EC CO 2 RR) is a promising technology to achieve a carbon-neutral society. EC CO 2 RR can directly convert the greenhouse gas emitted from stacks into valuable fuels and chemical feedstocks for various industrial applications. Numerous metal-based electrocatalysts have been researched to reduce overpotential and enhance the product selectivity of CO 2 RR. Recently, single-atom catalysts (SACs) are attracting intensive attention due to their low-cost, extremely high activities per loading amounts, and extensive stability of catalytic active sites due to the strong chemical interaction with coordination atoms. The coordination environments of SACs affect the electronic structure of active sites and change the energetics of the CO 2 RR pathways. Herein, the principles of EC CO 2 RR, including reaction mechanisms, figures of merits, and electrolysis systems, are first discussed. Then, the recent progress in the synthesis and characterization of SACs on various supports is accessed. Most importantly, the coordination environments of single-metal atoms and their influence on CO 2 RR catalytic ability, product selectivity, and active site stabilization are focused. This review provides a milestone along the design of SACs from the perspective of optimizing atomic configuration surrounding the active sites for EC CO 2 RR.
absorption fine structure (EXAFS), which can be used to check the bonding of the metal component.
In the case of carbon supports, the coordination of the single atom with the atoms doped on carbon affects the FE. Also, the reaction products from SACs using copper active sites change depending on the coordination of Cu atoms. However, many reviews on SACs focused on metal species, support type, or synthesis methods with less emphasis on the exact coordination of the single-metal atom. The coordination of carbon structurebased SACs is categorized into three types. The first involves a metal connection to four N atoms, usually comprised of an N-doped carbon substrate synthesized with metal precursors (M-N 4 -C). The second is a connection that either involves elements (vacancies included) other than N, or that involves the metal being connected to five atoms with one bond being out of the plane. Between the first and second sections, the effect of the bonding in single-metal atom sites can be compared. The third is dual-atom catalysts (DACs), and this section includes both SACs, which include two types of metals, and dual-atom site catalysts in which a pair of metal atoms are inserted into the carbon structure. The selection of metal elements in this type of coordination affects product selectivity, as elaborated in Section 4.3. [12] Afterward, other supports, namely metal and metal oxides, are investigated. In these cases, the coordination will be metal-metal or metal-O.
Various carbon supports like graphene, graphene oxides, porphyrins, and such are used for the synthesis of SACs, with synthetic methods varying from pyrolysis to wet chemistry, atomic layer deposition, and electrodeposition. Pyrolysis is often used in the synthesis of SACs for CO 2 RR. There have also been attempts to synthesize SACs by pyrolysis of metal-organic frameworks (MOFs) and metal precursors. MOFs are highly porous material that consists of metal atoms or metal clusters connected by organic linkers. They can be used as a precursor for synthesizing SACs; pyrolysis of zeolitic imidazolate framework-8 (ZIF-8) is a common example because of its N-doped carbon structure and the low evaporation temperature of zinc which facilitates exchange in metal sites. [13] In this article, we first introduce the mechanisms and parameters of CO 2 RR and how it is used in large-scale, real-life applications as compared to experimental results in various cells. Then the synthesizing and characterization methods of SACs are discussed. The main concept, which is the effect of the coordination of SACs on CO 2 RR, will then be investigated ( Figure 1). Finally, a brief summary of SACs and their use in CO 2 RR will be given with an overlook of its opportunities and challenges.

Products and Pathways of EC CO 2 RR
CO 2 is the most oxidized form of carbon. CO 2 molecules can be electrochemically reduced to various products, including CO, HCOOH, CH 4 , etc. Table 1 summarizes the formal redox potential and the number of required electrons and protons for each reaction. As shown in Table 1, CO 2 RR occurs via multiple proton-electron transfers. As the number of the required electrons increases, the reaction pathway gets complicated, and several mechanistic proposals are given for one product, with the mechanism for each product still under debate (Figure 2a). [14] Metalbased electrocatalysts for CO 2 RR have been extensively researched since Hori and coworkers classified metals according to their product selectivity. [15,16] Au, Ag, Pd, and Zn selectively produced CO; Sn, In, Hg, Cd, and Pb were prone to HCOOH production. The intrinsic tendency was attributed to their binding energies with the key intermediates, *COOH and *OCHO (Figure 2b,c). Cu has received much attention as it is the only metal that can produce C 2þ products, and a total of 16 or more products were detected through aqueous CO 2 electrolysis on the Cu cathode. [17] However, unlike bulk metal catalysts, the CO 2 RR product from SAC is usually limited to CO gas. [18] The gap between Formic acid 1.0 atm, 25°C, and additionally, pH 7; b,c) half-reactions of water splitting.
www.advancedsciencenews.com www.small-structures.com the two kinds of catalysts originate from the distinct electronic structure of SACs. Undercoordinated features and the effect of different coordination atoms change the adsorption energy and configuration of reaction intermediates and thus affect the reaction pathways. [19] The relationship between the metal used for SAC and the product is also very different from the tendency expressed in Hori's table. When hydrogen evolution reaction (HER)-selective metals such as Ni, Fe, Co, and Zn were atomically dispersed, HER suppression occurred, showing CO 2 RR selective properties. In a mechanistic study conducted by Rossmeisl et al., [20] the difference in selectivity was attributed to the energetics of HER in the single site porphyrin-like structure (M-N 4 -C). HER is widely known to occur by two mechanisms. Two mechanisms, Volmer-Tafel and Volmer-Heyrovsky, both have an initial proton-coupled electron transfer (PCET) step commonly referred to as the Volmer reaction. The Volmer reaction produces surface adsorbed *H. The second step is divided into the Tafel step, where dimerization occurs after the production of another *H on the surface, and the Heyrovsky step, where the H 2 molecule is produced at once through secondary PCET. [20] DFT calculations for Pt, Ni, Co, Au, and Cu bulk metals revealed that the Volmer-Heyrovsky mechanism has a much higher activation energy barrier than the Volmer-Tafel mechanism. Therefore the latter reaction is kinetically more favorable. [21] A Tafel reaction requires hydrogen adsorption to occur at each of two adjacent active sites. . Electrochemical reduction of carbon dioxide (EC CO 2 RR) on metal surfaces: a) CO 2 electroreduction pathways. Reproduced with permission. [30] Copyright 2021, Wiley-VCH GmbH. Volcano plots using the: b) *COOH binding energy and c) *OCHO binding energy as descriptors. Reproduced with permission. [120] Copyright 2017, American Chemical Society.
However, due to the absence of two similar close sites, SACs cannot undergo the Tafel reaction, forcing the Heyrovsky step burdened with a high-energy barrier. Therefore, HER suppression occurs with SAC ( Figure 3). [19,20,22] For similar reasons, the selectivity of C 2þ products in SAC is very low. The most crucial step for the formation of hydrocarbons is the step where two *COs meet, and dimerization occurs through C-C coupling. [23] Since SAC does not have adjacent active sites, dimerization is less likely to occur and therefore the FE of the multi-carbon product is low. Karapinar et al. reported Cu SAC with ethanol faradaic efficiency exceeding 43%. [24] Such an unusual result was explained by the reversible restructuration of the Cu metal site during CO 2 RR. Conversion of atomically dispersed Cu-N 4 coordination into Cu nanoparticles was shown by operando X-ray absorption spectroscopy (XAS) analysis. After electrolysis, the metallic copper phase disappeared, and Cu single-atom coordination was restored as the electrode was exposed to the air. The strong Cu II -chelating capacity of the N 4 sites was appointed as the reason for this experimental observation. In this case, the aggregation of metal atoms enabled the formation of the C 2 product. Without such reconstruction, *CO intermediates are physically blocked from dimerization.
Bidentate binding by two oxygen atoms of CO 2 molecule on the catalyst surface leads to the formation of formate/formic acid. [25][26][27] However, such CO 2 adsorption mode is improbable in single-atom coordination as two adjacent active sites with similar oxygen adsorption properties are required. Nevertheless, a few researchers succeeded in fabricating SAC capable of formate production. [28,29] Bidentate binding on one active site or monodentate oxygen coordination, as illustrated in Figure 2a, might account for the CO 2 conversion toward formic acid though such cases are rare. [30,31] Blocked C 2þ and formate production combined with HER suppression resulted in high CO faradaic efficiency of SACs matching that of Au-based electrocatalysts. [32,33]

Evaluation Parameters for EC CO 2 RR
There are several parameters used to evaluate the catalytic ability of electrocatalysts. In the case of water splitting, comparing the current density at a certain overpotential or, more generally, the overpotential at a certain current density through linear sweep Figure 3. EC CO 2 RR on SACs: a) Hori's table for SACs showing the relationship between metal species and product selectivity. Reproduced with permission. [18] Copyright 2020, Wiley-VCH GmbH. b) Different HER mechanisms on a metal surface and a SAC active site. Reproduced with permission. [20] Copyright 2017, Elsevier. Activation barriers for the HER on Pt (111) as a function of: c) the reaction energy and d) the electrode potential. Reproduced with permission. [21] Copyright 2010, American Chemical Society. voltammetry (LSV) would be the most typical way of evaluating the activity of electrocatalysts. Overpotential is the difference between the thermodynamic standard potential of a specific half-reaction and applied potential. [34] A smaller overpotential at a given current density or a larger current density at a given overpotential usually means the catalyst is more active for the reaction. However, in the case of CO 2 electroreduction, hydrogen evolution also occurs at the same electrode as a competing reaction. The measured current density cannot be directly related to CO 2 reduction only and faradaic efficiencies should be considered. Also, the standard reduction potential varies according to the products.
Faradaic efficiency shows the product selectivity of the CO 2 reduction reaction. Also known as Coulombic efficiency, faradaic efficiency is defined as the charge consumed for the formation of a certain product over the overall charge that has passed during the reaction. Other possible reactions include catalyst degradation, competing for reactions such as hydrogen evolution reaction, or CO 2 reduction reaction to a non-desired product. [35] Achieving high FE for valuable products is a major goal in designing CO 2 RR catalysts.
Another characteristic is the partial current density. The partial current density of a target product is equal to the product of the current density and Faradaic efficiency of the specific chemical. [35] Partial j can be converted to formation rate, which is useful in a practical assessment of the electrolytic system. Stability is one of the most important performance targets. Long-term (over 1000 h) stability is mandatory for the practical application of a CO 2 electrolyzer. However, many metal-based electrocatalysts suffer from structural instability within the reaction conditions or from poisoning by reaction intermediates. [36][37][38][39][40] CO poisoning could be alleviated by applying potential stepwise or employing bimetallic composition. The current density increased as the cell type changed to flow cell and MEA, and the increase in current density further promoted the instability of the catalytic material. To prevent morphological and chemical modification of the catalyst material under reaction conditions, a logical design is required at the catalyst design stage.
Energy efficiency (EE) can be calculated from the energy stored in the reaction products and the total energy consumed during electrolysis. High overall EE needs optimization of whole cell structure including electrolyte and oxygen evolution reaction (OER) catalyst, though many research papers demonstrate only EE of CO 2 RR half-reaction for simplicity. [41,42] Turnover frequency (TOF) refers to a residence time of a reactant on the catalyst surface to form a product molecule under reaction conditions. TOF determines the intrinsic activity unbiased by the amount of catalyst used, and concentration of active sites. [43] However, determining TOF is not always successful since the catalytic active sites may differ from the surface sites measured by chemisorption. [44]

CO 2 Electrolysis Systems
The type of cell in which the electrochemical reaction takes place has a great influence on the reaction. [45] Two independent halfreactions occur in the CO 2 electrolyzer. The reaction occurring at the cathode is the CO 2 RR and the reaction occurring at the anode is the OER. The compartments where the two reactions take place must be separated for the following reasons. First, CO 2 RR catalysts are vulnerable to pollution. CO 2 RR FE decreases when materials used in the anode, such as Ni, and Fe, dissolve in the electrolyte and adhere to the cathode surface under the influence of concentration gradient and electric fields. Second, preventing the cross-over of the products from each half-reactions is important. Once the gas products are mixed, they must go through a separation process of impure gases for use in fuel or chemical feedstocks. Products in the liquid state might diffuse to the counter electrode and cause re-oxidation. For these reasons, most CO 2 electrolyzers use polymer separation membranes.
The most basic form of CO 2 electrolytic cell using such a separation membrane is a batch-type cell (Figure 4a). Also known as H-cells, they are used on lab scales and used for screening electrocatalyst materials. [45] Gases that evolved from the compartments pass through gas chromatography (GC) instrument to be quatitatively analyzed. And the electrolytes are collected to be analyzed via high-performance liquid chromatography (HPLC) or nuclear magnetic resonance (NMR) analysis. Since the FEs of liquid-phase products are often low, the electrode area/cell volume ratio must be large, and the electrolysis time must be long enough to exceed the detection limit of the analyzer. [17] However, CO 2 depletion occurs faster when the cell volume is relatively small. Considering the poor reactive mass transfer due to the low solubility of CO 2 gas when using aqueous electrolytes, the limitations of H-cells appear to be distinct. For commercialization of CO 2 electrolysis, a current density of Figure 4. a) H-type electrochemical cell. Reproduced with permission. [121] Copyright 2016, Elsevier. b) Flow cell with a membrane electrode assembly. Reproduced with permission. [47] Copyright 2018, American Chemical Society. c) J-V curve of CO 2 RR with carbon source as aqueous phase and gaseous phase. Reproduced with permission. [48] Copyright 2016, American Chemical Society.
www.advancedsciencenews.com www.small-structures.com 200-400 mA cm À2 should be reached. [46] This value cannot be obtained in the moderate range of overpotentials from batch-type reactors and thus requires other types of reactors. Flow cells overcome the mass transfer limitation of H-cell and can generate a current density of 100 mA cm À2 or more. As the CO 2 -saturated electrolyte continues to circulate, it serves to supply reactants and rapidly extract products near the electrode surface. The most representative form is the membrane reactor, which has already been extensively researched and developed for fuel cells and water electrolysis cells. The membrane reactor consists of a polymer electrolyte membrane (PEM), a cathode/anode gas diffusion electrode (GDE), and a cathode/anode flow plate. [47] A membrane electrode assembly (MEA) can be used instead of a liquid electrolyte to shorten the distance between electrodes and reduce ohmic overpotential ( Figure 4b). Furthermore, humidified gas can be supplied to the reactant to increase CO 2 gas supply efficiency and increase CO 2 RR selectivity ( Figure 4c). [48,49] Recently, research was conducted revealing that the humidity of gas affects the product selectivity of C 1 and C 2 . [50] The type of membrane used in the MEA, as well as the reactant phase, determines the FE of competitive HER. Using a cation exchange membrane (CEM) creates an acidic environment around the cathode catalytic active site, where HER dominates over CO 2 RR. [49,51] Therefore, many studies use bipolar membranes (BPM) or anion-exchange membranes (AEM) to construct MEAs.

Synthesis of SACs
Many synthetic methods have been developed to fabricate SACs. Pyrolysis, wet chemistry method, atomic layer deposition (ALD), and electrodeposition are the representative synthetic processes. Appropriate fabrication methods along with the selection of precursors enhance the catalytic performance of catalysts. One of the strategies is to increase the density of single-atomic sites. Single atoms are anchored in defects with heteroatom bonding. Thus, the density of defective sites with C, N, O, and S coordination atoms should be sufficient. Nanoparticles or clusters formed during synthesis are eliminated to maintain homogeneity by acid leaching. To withstand such extreme conditions of the post-treatments, corrosion-resistive carbon-based materials like various MOFs, covalent organic frameworks (COFs), and carbon nanotubes (CNTs) are widely used to make SACs. Carbon templates are also preferred for their low-cost, good electrical conductivity, chemical stability, and high porosity. [34] Precisely designed carbon supports such as MOF conserve their highly porous structure and atomically dispersed metal sites even after high-temperature pyrolysis. [52] Heat treatments are efficient for regulating metal-N coordination and catalytic performance. [8,32,53] Various N-coordinated metal SACs have been explored by Strasser's group. Ni, Mn, Fe, Co, and Cu chloride solution with bipyridine-based coordinated polymers were pyrolyzed at 900°C for 2 h. The result was interconnected macropore walls with high accessibility on homogeneously distributed N-containing carbon lattices. Catalytic activity toward CO 2 RR showed a strong dependency on metal species. Ni-N-C and Fe-N-C exhibited the highest CO FE among transition metals experimented with. [54] Porosity can be further enhanced by employing MOF precursors. [55] Zhang et al. synthesized MOF-derived Bi-N 4 electrocatalysts. The porosity varied according to the temperature of the heat treatments. High porosities of the samples were observed by high-resolution transmission electron microscopy (HRTEM) and STEM. Brunauer-Emmett-Teller (BET) measurements confirmed the high surface area derived from the porous structure. [56] Zn-based ZIFs are widely used as sacrificial templates. Zn ions evaporate at temperatures above 800°C. Li's group utilized Zn-Co bimetallic ZIF to make Co-N X SACs. Zn atoms evaporated during pyrolysis, leaving Co coordinated with N ( Figure 5a). [57] A similar strategy was used to make Fe-Co-N 6 coordination. Starting from ZnCo-ZIF, Fe was doped into the evaporated zinc sites during the pyrolysis of ZIF and Fe salts. [58] However, the difficulty of detailed control during the pyrolysis limits its propagation. In many cases, the resultant samples are in powder forms, and immobilization of the catalysts onto conductive substrates should be considered also. As the adsorption ability usually depends on chemical interactions, drop casting is not always successful. If heat-resistant electrodes are used to support the catalyst in the first place, the inconvenience will be alleviated.
The wet chemistry method proceeds only in mild conditions and can ensure the inheritance of the precursor's structure and thus keep a high structure-activity relationship. [59] In wet chemical synthesis, the metal precursor is mixed with supports to gradually immerse and anchor on the defective sites of the supports ( Figure 5b). Supports are typically inactive toward CO 2 RR, and active atoms are immobilized to form heterogeneous catalysts. In this type of synthesis, it is important to keep the concentration of metal precursors low, as agglomeration easily takes place. [60,61] In Wang's research, the formation of Pt nanoparticles was suppressed only at mass loadings lower than 2 wt%. As Pt single atom/CeO 2 exhibited a 7.2 times higher reaction rate, the importance of suppressing clustering was again emphasized. [60] Lowering the temperature of precursor solution can also assist in synthesizing atomically dispersed sites. Huang et al. came up with the idea of increasing the nucleation energy barrier by keeping a low temperature. A water and alcohol mixture with a low freezing point was kept at À60°C, achieving a sluggish nucleation rate to suppress nuclei formation. [62] The dispersed cobalt electrocatalyst showed superior oxygen reduction reaction (ORR) performance. Molecular catalysts such as metalloporphyrin, phthalocyanine, and polypyridine are also used. [63][64][65] To immobilize molecular catalysts on carbon substrates, organic functional groups were introduced. [63] The ALD process is emerging as a facile synthesis method of SACs. ALD equipment allows relatively large and conformal deposition of films. The self-limiting mechanism of ALD deposition is the key to making single atoms. In each sequence, only one atomic layer of the metal precursor is formed on the surface of substrates. If the sequence number is kept as low as a few cycles, they cannot form films and keep atomically dispersed states. [66,67] The ALD process is beneficial for increasing the density of single atomic sites. To keep the metals from agglomerating, nucleation sites are better defined before the ALD process. Anchoring sites can be created in various ways, such as oxidation in acid or annealing in oxidative gases. Yan et al. first oxidized graphene nanosheets in concentrated sulfuric acid, potassium permanganate, and hydrogen peroxide (Figure 5c). Many oxygen-containing functional groups such as carboxyl group, carbonyl group, ether group, and hydroxyl group were attached to the benzene rings. The functional groups were limited to hydroxyl groups to anchor Pd metal atoms. The authors thermally induced deoxygenation of needless oxygen functional groups, precisely tuning the amount and type of anchor sites. [66] Electrodeposition is a well-known technique for coating metal films and even polymers. [68,69] Changing the applied voltage, current, time, and reaction condition of electrodeposition alters the morphology, electrochemical state, and thickness of the film. The diversity of the process brought attention to the electrodeposition of single atoms recently. Standard reduction potential (SRP) shows at which potential the metal ions reduce to metal or another oxidation state. SRP determines the minimum potential of the electrode for the metal ions to start depositing. However, this value was obtained under the assumption of a pre-existing metal substrate with the same metal species. So, with careful design of substrate, metal ion, and reaction condition, an atomic layer of the metal can be deposited at a more positive potential than SRP. This is a self-limiting procedure also called underpotential deposition. Shi et al. reasoned that the atomic layer can be upgraded to site-specific single-metal atoms if the substrate has isolated active sites (Figure 5d). [70] Chalcogen atoms (S, Se, Te) were pointed out as possible isolated sites to anchor and stabilize metal single atoms. MoS 2 with nonuniformly distributed sulfur atoms was used to support Pt atoms.

Characterization of SACs
The characterization of SACs requires a detailed analysis of the atomic dispersion and bonding, as they are prone to aggregation and their CO 2 RR activity changes depending on the coordination. The most often used methods of characterization can be divided into two parts; electron microscopy and XAS. Electron microscopy allows us to visually characterize SACs and includes imaging techniques of scanning electron microscopy (SEM), transmission electron microscopy (TEM), scanning tunneling microscope (STM), and HAADF-STEM. XAS provides a nonvisual analysis of SACS and includes XANES and EXAFS. Electron microscopy works by hitting samples with an electron beam and gathering information from transmitted electrons, secondary electrons, backscattered electrons, or X-ray emissions, depending on the type of machine. In XAS, X-ray beams ranging in energy are shot onto the sample. The resulting absorption spectra are divided into two parts, with XANES taking the lower energy part, including the edge peak, and EXAFS comprised of the 30 eV to 1000 eV region after the edge peak, showing single scattering. The XANES spectra includes the pre-edge and edge regions. The edge region shows peaks that can be linked to oxidation states and electronegativity, and the small peaks in the pre-edge region are linked to oxidation states, geometry, and bonding characteristics.
The function of each characterization method in CO 2 RR of SACs can be understood through an article by Li et al., where Pt atoms were used as catalysts to synthesize Fe, Co, and Ni SACs ( Figure 6). [71] While SEM images are often provided in Figure 5. Synthesis methods of SACs. a) High-temperature pyrolysis. Reproduced with permission. [57] Copyright 2016, Wiley-VCH GmbH. b) Wet chemistry method. Reproduced with permission. [122] Copyright 2016, Wiley-VCH GmbH. c) Atomic layer deposition (ALD). Reproduced with permission. [66] Copyright 2016, American Chemical Society. d) Electrodeposition. Reproduced with permission. [70] Copyright 2016, Nature Publishing Group. SAC characterization data, the resolution is not enough to isolate single-metal atoms. This can be done through HAADF-STEM at high resolutions, as seen in Figure 6c. As the target metal has a high z contrast compared to the carbon support, we can visually check the samples to determine if the metals are dispersed atomically. The low-resolution image ( Figure 6a) is also given to demonstrate that nanoparticles are not formed. Also, the contrast between Pt and Co makes it possible to distinguish between the two, with the bright Pt atoms marked by white circles and the less bright Co atoms marked by red circles. Furthermore, energy dispersive X-ray (EDS) mapping (Figure 6b) shows the carbon support, N doping on the support, and a high, even loading of metal atoms. Although not presented in this paper, STM is used to gather visual information on samples before and after CO 2 adsorption. In situ STM can also be used for further analysis of the adsorption process and is usually seen in an analysis of the mechanism for CO 2 RR. Fourier transform (FT)-EXAFS of Co K edge in Figure 6d can be used to determine if the metal is atomically dispersed or if Co─Co bonds are formed by comparing peaks from the sample to peaks of Co foil. In this case, it is derived that ALD is without prior Pt atom loading. K-edge XANES spectra are shown in Figure 6e, where the Co 1 Pt 1 -N-C deviates from Co coil and Co 3 O 4 references. XANES can gather information about the local atomic structure, with the first derivatives giving the metal oxidation states. When coupled with simulated results of pyrrolic-N 4 and pyridinic-N 4 structures, the conformity shows which structure the SAC samples are in. Density functional theory (DFT) calculations can be used besides the two experimental characterization strategies. DFT is a computational method that uses quantum mechanics to calculate electron distribution and thus predict material properties. It can be used to model the electronic configuration of expected SAC structures (Figure 6f ) or draw Gibbs energy diagrams for the CO 2 RR, through which the rate-determining step or limiting step can be determined.  Table 2). Among them, the case where a metal atom on   Batch cell [75] Ni- an N-doped carbon structure is supported by four nitrogen atoms; M-N 4 -C, is the basic structure. This coordination with N lets the metal atom be stabilized against aggregation while the electronic structure of the metal, therefore the catalytic activity, is affected by the surrounding electronegative nitrogen atoms. In the case of Fe-N-C catalysts, the N atoms have been observed to generate CO molecules, operating as an active site alongside the Fe atom, which correlates with CO re-adsorption and thus methane formation. [72] Ideally, the M-N 4 -C catalyst has a symmetrical structure and can mainly be divided into two forms depending on the form of the N-doped carbon support: pyrrolic and pyridinic. Such symmetric structures have the advantage of being structurally stable, which is emphasized in SACs as high stability means less chance for aggregation of the single-metal atoms. SACs showed high selectivity of CO with metals that were not used in CO 2 RRs when it was in bulk form. Ni and Fe are examples, where the metal is often used as a catalyst in HER, which is a competing reaction with CO 2 RR, meaning it shows low CO 2 -to-CO conversion efficiency. In an experiment by W. Ju et al. [54] Ni-N-C catalysts synthesized through pyrolysis showed a high CO FE of 85% at À0.78 V RHE (Figure 7a-h). Using the same synthesis method, Fe-N-C, Co-N-C, Mn-N-C, and Cu-N-C are also obtained (Figure 7b), with the partial current reaching the values of Au nanowires in the case of Fe-N-C (Figure 7c). The FE of each for H 2 and CO, CH 4 is also evaluated as seen in Figure 7d-f, with Ni-N-C and Fe-N-C exhibiting high CO FE, Co-N-C showing high H 2 FE, and Fe-N-C and Mn-N-C producing a small percentage of CH 4 . Figure 7g,h shows the experimental CO formation TOF as well as the DFT analysis used to explain the CO production and HER. For example, the CO FE change at different potentials can be attributed to the weak binding to COOH* of Ni-N X and Cu-N X from Region 1, while Mn, Fe, and Co-X moieties show strong CO* binding from Region 3. Thus, Ni and Cu-N X are ideal for the production of CO, and Mn, Fe, Co-N X , Fe, and Mn-N X hold the CO* state long enough for CH 4 formation. Although this cannot be generalized due to the differences in the synthesis methods, it can be used to estimate the different CO 2 RR activities for each metal.

Critical Role of Coordination Environment in
The product for the CO 2 RR of M-N-C catalysts is usually CO or HCOOH. However, copper SACs show products beyond CO* depending on the applied potential and overall structure. As such products show potential for CO 2 RR to be used in producing fuels and reducing energy consumption, copper SACs are first investigated for their CO 2 RR mechanisms. Creissen and Fontecave [73] explained the mechanism for C 2þ products in Cu SACs by the reversible formation of Cu clusters. The CO 2 RR mechanism for C 2þ products is explained by C-C coupling between adsorbed CO molecules, the active sites for which are metal. However, the observed distance between Cu-N 4 -C moieties does not sustain such dimerization for two CO molecules adsorbed onto the metal sites. Therefore, it is suggested that the potential applied during the reaction creates clusters by reducing Cu into Cu 2þ ions (Figure 8a). This reaction is reversible, with the clusters turning back into Cu-N-C structures either upon exposure to air or when an oxidation voltage is applied. The formation of clusters increases the selectivity of C 2þ products, with the size of the cluster affecting the specific product. In the case of metals other than Cu, there is less reduction of SAC sites and even when reduction does happen, it is often not reversible. Which explains why C 2þ products are not usually observed for SACs other than Cu.
Karapinar et al. [24] synthesized Cu-N 4 -C catalysts by pyrolysis of ZIF-8, pen, and a Cu precursor to obtain ethanol as the main product, achieving a Faradaic yield of 55% at À1.2 V RHE (Figure 8b-e). As can be seen through the scheme of CO 2 RR in Figure 8d, Cu clusters are formed to produce ethanol, with EXAFS data showing Cu-Cu contribution. This is in an agreement with the mechanism suggested in the previous paragraph. Zhao et al. [74] synthesized Cu-N 4 -C catalysts using different fabrication methods and precursors to obtain acetone as the main product, with an FE of 36.7% at À0.36 V RHE (Figure 8f-s). This study, the formation of C 2þ products was not attributed to cluster formation. Instead, the formation of C 2þ products is contributed to C-C coupling happening within the single atom site, with N taking a role, as can be seen in the DFT calculationbased scheme of CO 2 RR in Figure 8s. In this scenario, the difference in the main product and selectivity for Cu-N 4 -C structures is attributed to the change in bond length and coordination; whether it is Cu-pyrrolic-N 4 or Cu-pyridinic-N 4 . DFT calculations of Gibbs free energy shown in Figure 8s acts to validate the lower FE of the material with higher pyridinic-N and lower pyrrolic-N (Figure 8q,r). The conclusion of this paper allows an understanding of why catalysts with the same M-N 4 -C structure show differences in overpotential or FE; by differences in the pyrrolic and pyridinic ratio of N. In the article, the interpretation of CO 2 RR to products beyond CO* other than cluster formation is suggested. Interpretations assuming the possibility of C-C coupling occurring on single sites, or the possibility that the distance between Cu-N 4 -C moieties is enough for dimerization to occur, are available. However, further characterization, like in situ IR or Raman spectroscopy, is needed to conclude that Cu cluster formation does not contribute to the C 2þ product formation mechanism. For this section, the nonsymmetrical coordination of metal single atoms, where the coordination number or type of coordination atom changes, is introduced. The increase or decrease in coordination number intuitively induces a balance break in the electron configuration of the single-metal atom, and the difference in electronegativity of heteroatom doping breaks the electron density distribution. This affects the catalytic activity and thus, we expect the CO 2 RR to differ from that of the standard M-N 4 -C structures. Wang et al. [8] synthesized Co SACs with different nitrogen coordination numbers of Co-N 4 , Co-N 3 , and Co-N 2 by varying the pyrolysis temperatures for the Co/Zn ZIF precursors ( Figure 9a) and compared their CO 2 RR characteristics. XRD and EXAFS spectra shown in Figure 9b,c suggest atomic dispersion and the lowest N coordination of Co-N 2 . CO 2 RR results in Figure 9d,e show that Co-N 2 has the highest CO FE among the three in the potential range À0.85 to À0.45 V RHE , as well as having the highest current density. The Gibbs free energy is calculated (Figure 9f ) to show a less endergonic formation of CO 2 .À* for atomically dispersed (Figure 9g,h) Co-N 2 , attributing to the higher FE of CO. This can be explained by electronegativity difference, for C is less electronegative compared to N, and this will affect the affinity of the catalyst to CO 2 RR reactants and intermediates. A similar experiment comparing Ni-N 2-C, Ni-N 3 -C, and Ni-N 4 -C also showed that Ni-N 2 -C has a higher current density and CO FE. [75] Thus far, the coordination number around the metal atom has been 4. This is not always the case for SACs. The M-N 5 coordination is observed by Zhang et al., [76] with the Fe-N 5 -C structure showing lower overpotential (0.35 V), significantly higher current density, and CO FE (97% at À0.46 V RHE ) than Fe-N 4 -C (Figure 9i-l). The Fe-N 5 -C structure synthesized through pyrolysis of hemine, melamine, and graphene is drawn with an extra out-ofplane N atom connecting to the metal atom (Figure 9l) confirmed by calculations of XANES matching experimental data (Figure 9i,j). There is a similar case, where there are five coordinated atoms to Fe in the article by Wang et al. [77] where the coordination is Fe-N 4 O (Figure 10k-p). Precursors containing oxygen were used to synthesize Fe-N/O-doped carbon and the axial connection of O to Fe is similar to the case of Fe-N 5 , as can be seen in Figure 10l. The ratedetermining step in CO formation is COOH*, unlike that of Fe-N 4 . The experimental comparison was not directly made with Fe-N 4 , but a high FE of 96% is shown, comparable to the 97% in a previous work done on Fe-N 4 , and the structure also showed a relatively high current density of À5.6 mA cm À2 . Figure 8. a) Schematic of Cu n cluster formation and deformation upon the application of potentials, Cu-Cu distances in nanoparticles, and Cu-N-C structures. Reproduced with permission. [73] Copyright 2021, Nature Publishing Group. b) FT-EXAFS for Cu K edge and c) HAADF-STEM image of Cu 0.5 NC. d) Schematic for CO 2 RR on Cu 0.5 NC upon applied potential e) Faradaic yields depending on flow rate, potential, and electrolyte cation for CO 2 RR on Cu 0.5 NC. Reproduced with permission. [24] Copyright 2019, Wiley-VCH GmbH. Cu-SA/NPC images of: f ) SEM, g,h) TEM at different resolutions, i,j,k) HAADF-STEM with corresponding l,m,n) EDS mapping. o) CO 2 RR production rates and p) Faradaic efficiency of Cu-SA/NPC. XPS N 1s spectra for: q) Cu-SA/NPC and r) Cu-SA/NPC Ar . s) DFT calculated free energy diagrams for CO 2 RR to CH 3 COCH 3 at potential À0.36 V for Cu-pyridinic-N 4 and Cu-pyrrolic-N 4 , and t) reaction intermediate structures on Cu-pyrrolic-N 4 (gray: catalyst C, black: adsorbate C, blue: N, white: H, orange: Cu, red: O). Reproduced with permission. [74] Copyright 2020, Nature Publishing Group. As was seen in the case of Fe-N 4 O, the coordination of metal atoms to oxygen is another standpoint. Cai et al. [78] investigated the CO 2 RR on Cu-N 2 O 2 -C catalysts (Figure 10a-j). These metal single atoms have in-plane coordination with two N and two O (Figure 10a) and showed a high CH 4 FE of 78% and a comparison with Cu-N 4 moieties given through the free energy diagram of CH 4 formation and H 2 formation (Figure 10b,c). This high FE is attributed to the charge density change (Figure 10d,e) that differentiates the limiting potential of CH 4 from the limiting potential of other possible products (Figure 10j). The case where the atom is not surrounded by another atom and there are unsaturated metal species, in other words, when there are vacancies in metal coordination, is studied by Mou et al. [79] These vacancies can be obtained via a change in the pyrolysis temperature of the Ni-graphene oxide melamine foam. The uncoordinated Ni-N x sites make the free energy of the rate-determining step for CO 2 RR lower, thus inducing high catalytic activity to produce CO.
There are also other elements that can coordinate with the single-metal atoms in the place of N or O, such as S, P, and Cl The example of Cl being incorporated into the carbon support is seen in the article by Zhang et al. [80] where the CO 2 RR of Mn-N 4 Cl is demonstrated using manganese chloride as a precursor. The overall current density is lower than that of Mn-N 4 , but the CO FE and partial current density are much improved. Without Cl incorporation, the free energy diagram for the production of H 2 shows a low reaction barrier, making it easy for HER to happen instead of CO 2 RR. Wang et al. [81] fabricated Bi-N 3 S sites through polymerization of PDA on Bi 2 S 3 , followed by annealing. The electronegativity difference between N (3.04), S (2.58), and Figure 9. a) Schematic of the synthesis and structure of Co-N 4 and Co-N 2 . Characterization of Co-N 2 , Co-N 3 , and Co-N 4 in order of: b) X-ray diffraction (XRD), c) extended X-ray absorption fine structure (EXAFS) spectra, d) X-ray absorption near edge structure (XANES), and e) XPS. f ) Comparison of Co-N 2 and Co-N 4 in DFT calculated Gibbs free energy for intermediates in the CO 2 RR to CO. g,h) HAADF-STEM images for Co-N 2 , red circles representing Co atoms. Reproduced with permission. [8] Copyright 2018, Wiley-VCH GmbH. Experimental XANES curves compared with DFT simulated curves of: i) FeN 4 and j) FeN 5 . k) XPS spectra for N 1s in FeN 5 catalyst. l) Schematic for the synthesis of FeN 4 and FeN 5 . Reproduced with permission. [76] Copyright 2019, Wiley-VCH GmbH. Bi (2.02) atoms makes S act as an electron donor, shifting the energy level of Bi. [82] This change induces higher CO FE of a maximum of 98.3% at À0.8 V RHE , and a higher partial current density compared to Bi-N 3 S catalysts to Bi-N 4 catalysts. While various combinations in the coordination are possible, atoms having high electronegativity are most often used in the place of N, as can be seen in the research specified above. The synthesis of atoms coordinated with different metals can be optimized for maximum CO FE. The asymmetrical atomic interface is polar, thus enhancing the CO 2 R performance by facilitating the transfer of electrons. [83] With the addition of axial elements, the high adsorption of intermediate caused by highly electronegative N 4 is reduced. [84,85] Similarly, the substitution, subtraction, and addition of surrounding elements [8,75] in the N 4 plane relax the strong adsorption of intermediates to SACs with M-N 4 structures. Thus, highly doped carbon-based SACs with unsaturated coordination show high efficiency in CO 2 RR.

Dual Metal Atom Coordination on N-Doped Carbon
The characteristic of DAC is that there is more than one single-metal atom dispersed on the support. One form of DACs is binuclear dual-atom pairs, where the metal atoms are distributed in pairs, showing either M, M 0 -N-C coordination or M 2 -N-C coordination in which the same metal is merely paired. Another form of DACs is the mononuclear heterogeneous single atoms, with the coordination being M-N-C and M 0 -N-C for N-doped carbon supports. The interaction between metal atoms can show synergistic effects led by the optimization of electronic structures that enhances catalytic activity. [86] In the M, M 0 -N-C coordination or M 2 -N-C coordination DACs, where metal atoms are adjacent and show stronger synergistic effects. [87] The dual-atom pair's importance in CO 2 RR can be related to the mechanism of metal cluster formation decribed . Reproduced with permission. [77] Copyright 2020, Elsevier. in Section 4.1. With dual-atom sites, the active metal sites are no longer too far for CO dimerization to occur as the two metal atoms are adjacent to each other, suggesting products other than CO forming without the need for cluster formation. While the exact mechanism for performance increase in DACs is still under debate, this could be accounted for by the change in catalytic activity. Another point in DACs is that the metal atom loading is higher than that of SACs, which is also a factor that can lead to higher FE. [87] Binuclear homogeneous dual-atom pairs are observed in a study by Wang et al. [88] where Fe 2 -N-C DACs are compared with Fe 1 -N-C SACs (Figure 11a-g)). Both catalysts are synthesized via pyrolysis of ZIF-8 along with a Fe precursor, with the resulting material differing according to atmospheric conditions during pyrolysis; Ar, 5% H 2 , 10% H 2 , 20% H 2 flow, each forming Fe 1 -N 4 -C, Fe 2 -N 6 -C-o with Fe of ortho position, Fe 2 -N 6 -C-p with Fe of para position, and Fe 1 -N 3 -C samples (Figure 11a, d-g). The CO 2 RR results show that the Fe 2 -N 6 -C-o catalyst has the highest CO FE. Gibbs free energy diagrams for the CO product are drawn for each sample, as well as the limiting potential of CO 2 RR with regard to HER. Another example of this type is examined in a study by Zhang et al., [89] where Pd DACs are synthesized to show a high CO FE of 98.2% with the reason attributed to electron transfer between adjacent metal atoms.
Mononuclear heterogeneous single atoms can be seen in the research by A. Guan et al. [88] where DACs with the coordination Cu-N 2 @Cu-N 2 , called by the name "Cu-N-C-800" to indicate its pyrolysis temperature, were synthesized (Figure 11h-l). While "Cu-N-C-900" comprised of SACs shows CH 4 to be the major product of CO 2 RR, Cu-N-C-800 produces C 2 H 4 , a C 2þ product, with a higher FE than CH 4 of up to 24.8% FE at potential À1.4 V RHE . The reason for this is said to be that the high Cu concentration makes C-C coupling easier, which in turn makes C 2 H 4 formation easier.
A computational study by N. Karmodak et al. [90] summarized the DFT calculated formation energy and gives the activity volcano plots for SACs of various metals and DACs paired with Fe (Figure 11m-o). The metal atom arrangements are organized by vacancy sites as seen in Figure 11m. The formation energies for Fe-M dimers (M ¼ first-row transition metals) for the different vacancy sites show that stable dual atom sites are as stable as the corresponding single atom sites, which means they can be simultaneously synthesized. Thus, controlled MOF structures are used to synthesize the specific DACs. The activity volcano plots for metal (211) surfaces of Figure 11n,o show which reaction is the rate-limiting step by which side of the black line the material is on. The Gibbs free energy must be at a balance between the reaction of CO 2 (g) ! CO 2 * and CO 2 ! * COOH* for best TOF and SACs, to show a linear relationship. The early transition metal dopants exhibit high binding strengths, making CO* desorption the limiting step while the CO 2 * adsorption or COOH* formation acts as the rate-determining step in later transition metal dopants. The plot for DACs showed larger deviances from the SAC's linearity. With screening from these two plots, it is concluded that Cr-N-C, Fe-N-C, and Co-N-C will show high TOF for CO and for DACs, the favorable Fe-M dimer differs according to the doping sites, but candidates Fe-Co, Fe-Mn, and Fe-Ni are found. As DACs provide a larger pool of candidates than SACs, such screening using DFT calculations is often used to predict which combination will show good CO 2 RR. Experimental examples of binuclear heterogeneous dual-atom pairs are found in the work by Li et al. [91] and Liang et al. [92] The former introduces (Fe-Ni)N 6 synthesized through carbonizing and annealing, using the precursors ZIF-8 nanocrystals, Ni(NO 3 ) 2 and Fe(NO 3 ) 2 . This combination of metals which already show good CO conversion in SAC form, gives CO FE of 96.1% at À0.7 V RHE , with the FE similar to that of the Fe SAC and Ni SAC while the partial current density for CO exceeds both. The latter introduces dual atom sites of Zn and La. In this case, the objective was not to obtain higher CO FE but to produce syngas mixtures.
DACs are either in the form of metal pairs or in SA moieties of different metals that are simply mixed. The area presents a broader range of possible catalysts, especially if the elements surrounding the metal atoms are also varied to provide diverse combinations. As the interaction between different metals resulting in synergistic effects for CO 2 RR has been seen, this area presents an important hallmark for future studies on SACs. The key point of future study would be on experimentally achieving high FE for C 2þ products using DACs and other multi-component SACs, as is examined in theoretical studies and some experimental studies; for example, one study which synthesized Cu 1.4 Ni DACs with methane as the major product and Cu DACs showing a maximum FE of 91% for C 2þ products at À1.66 V RHE , [93] and another study which synthesized Cu few-atom catalysts comprised of diatomic and triatomic sites to achieve FE of 53.8% for acetate. [94] 4.4. Single-Metal Atom Coordination on Defective Oxides: Even though carbon-based SACs are considered promising materials, a series of metal oxides-supported SACs have been continuously developed since 2011. [95][96][97][98][99][100] Oxides with low dimensional and porous structures are good choices of substrates due to their enhanced electrical conductivity and mass transport. [101] To strongly bind metal atoms on the substrates, the introduction of surface oxygen vacancies (V O ) is crucial. Qi et al. combined first-principle calculations and an artificial intelligence approach to forecasting the stability and activity of various metal single atoms on several metal oxide substrates with V O . Combining 29 transition metals and 8 defective metal oxides led to the analysis of 232 catalytic systems, but among them, only 28 systems remained stable during CO 2 RR. (Figure 12a,b) The stability was attributed to three factors, namely the d-band center of metal oxides, the coordination number of the adsorbed species, and the electronegativity of transition metal single atoms. [102] V O -ZrO 2 -supported SACs were predicted to be promising materials for CO 2 RR, as confirmed by experiments of several other research groups. [96,97] Dostagir et al. demonstrated inherently defective ZrO 2 decorated with Co single atoms. CO 2 gas was selectively hydrogenated into CO even though the CO 2 adsorption mode was bidentate oxygen binding, typically favorable to produce formic acid/formate. Operando diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) analysis revealed the in situ formation of formate and subsequent decomposition to CO under the reaction condition.
The isolated Co atoms existed near oxygen vacant sites. [96] Samson et al. obtained Cu SAC/V O -ZrO 2 by wet impregnation method. The existence of oxygen vacancies stabilized the originally metastable phase of ZrO 2 and Cu cation. The active center for methanol synthesis was also elucidated. According to the IR analysis, increasing the content of the t-ZrO 2 phase Reproduced with permission. [123] Copyright 2022, American Chemical Society. h) Schematic of the synthesis of Cu-N-C-800, Cu-N-C-900. i) XANES region of XANES first derivative curves for Cu2O, CuO, Cu-N-C-800, and Cu-N-C-900. j) Faradaic efficiencies, partial current densities, and ratios for CH 4 and C 2 H 4 products of CO 2 RR for Cu-N-C samples prepared at the given temperatures. k) LSV of N-C, Cu-C-900, and Cu-N-C-900. l) Gibbs free energy diagrams for different Cu-N X coordination on CO 2 RR in C-C coupling. Reproduced with permission. [88] Copyright 2020, American Chemical Society. m) Stable doping sites for single and double atoms on a graphene sheet denoted according to vacancy sites; DV: divacancy, TV: trivacancy, QV: quadvacancy, PV: pentavacancy, HV: hexavacancy. (grey: C, blue: N, orange: metal) CO 2 RR active volcano plots with descriptors in binding energies of CO 2 * and COOH* for: n) SACs and o) DACs with transition metals and varying coordination. Reproduced with permission. [90] Copyright 2022, American Chemical Society.
www.advancedsciencenews.com www.small-structures.com stimulated catalyst acidity, and catalyst acidity had a linear correlation with catalytic activity. [97] Lewis acid sites in metal oxides were suggested to promote CO 2 conversion toward CH 4 . The metal atoms in oxides act as Lewis acid sites and undergo Lewis acid-base interactions with the O in CO 2 . [103][104][105][106] This type of interaction facilitates the breaking of C─O bonds, thereby activating CO 2 molecules and promoting the production of methane. [107][108][109][110] Chen et al. explored the effect of Lewis acid sites on the electronic structure of Cu single atoms. Theoretical calculation of Al 2 O 3 and Cr 2 O 3supported SAC showed that stronger Lewis acidity improves methanation on Cu SA. According to the DFT calculations, HCOO* formation was more favorable than COOH* in the Al 2 O 3 substrate, showing an energetically advantageous pathway toward CH 4 than CH 3 OH, CO, and HCOOH. On the contrary, the methanation pathway was much more unfavorable in Cr 2 O 3 than other competing reactions regarding the high formation energy of H 2 COO*. Theoretical results were confirmed by measuring the catalytic ability. The Cu single atoms on Al 2 O 3 achieved CH 4 faradaic efficiency of 62% and a current density of 153 mA cm À2 at À1.2 V RHE , while Cu SAs on Cr 2 O 3 mainly produced hydrogen. [95] Coordinating oxygen vacancies not only stabilize single atom sites but also helps in the CO 2 dissociation to CO*. [98] Cu with V O -TiO 2 helped in the adsorption of activated CO 2 . However, similar DFT and experimental results were not reproduced in TiO 2 or Cu-dispersed stoichiometric TiO 2 . Also, only a specific metal, copper, was effective.
In some cases, the metal single atom itself induces oxygen vacancies to appear in the oxide substrate. Lin et al. measured the degree of V O generation in FeO x using H 2 temperatureprogrammed reduction (TPR). FeO x with Ir single atoms required the least amount of H 2 gas for reduction, verifying the reduction ability of Ir SACs and generation of oxygen vacancies in the FeO x support. Moreover, according to a calculation based on the particle size distributions and total activities, 70% of catalytic activity in Ir/FeO x was estimated to originate from single atoms. The activity was 1 order of magnitude higher than the Ir cluster or nanoparticles despite lower Ir mass loading. [111] Recently CeO 2 is drawing attention as an oxide substrate for metal decoration due to its strong metal-support interactions. [112] Strong metal-support interaction leads to the highly dispersed form of metal loading, even to an atomic level. [113] Graciani et al. first explored the use of Cu/CeO 2 for CO 2 reduction and obtained methanol from the process. The copper-ceria interface induced special reaction pathways for the CO 2 to CH 3 OH conversion. [114] Zheng's group succeeded in controlling the size of the loaded copper metal to an atomic scale, obtaining a high selectivity of CH 4 product on the Cu/CeO 2 nanorod catalyst. Structural characterization indicated high dispersion of Cu ion with the association of three oxygen vacancies. (Figure 12c) Figure 12. a) Single metal and oxide candidates for screening. b) Calculated ΔE anch s for 232 systems. Reproduced with permission. [102] Copyright 2022, American Chemical Society. c) Structure models of CeO 2 (110) with different numbers of V O s. d) Electrochemical CO 2 reduction performances. Reproduced with permission. [115] Copyright 2018, American Chemical Society. e) In situ environmental transmission electron microscopy (ETEM) images showing the evolution of Ag NPs to SAs. f ) XAFS spectra of Ag 1 /MnO 2. Reproduced with permission. [117] Copyright 2020, Wiley-VCH GmbH. Multiple vacancies around Cu ensured the stability of the structure and the high partial current density of methane ( Figure 12d). [115] The different product preferences of Cu/ CeO 2 electrocatalysts of the two research groups might have originated from the morphologies of Cu/CeO 2 catalysts. Zeng's group discovered that the nanorod Cu/CeO 2 exhibited the highest TOF for methane. Cu was doped into CeO 2 nanorods, nanocubes, nanoparticles, and nanospheres for comparison. Among them, nanorods possessed the largest electrochemical surface area (ECSA), higher density of oxygen vacancies, and better capability for adsorption and activation of CO 2 molecules. Abundant defects confirmed by the XRD pattern and Raman spectrum facilitate the contact between active Cu sites and the reactant. [100] Ceria with nickel single-atom was also employed for C─H activation. Ni SACs could be stabilized by oxygen vacancies and the concentration of vacancies was controlled by the introduction of smaller size cations (Mg, Co, Zn). These secondary cations substituted Ce 4þ cation and V O concentration increased from 22% to 31%. The same strategy can be applied in future research on oxide-supported SACs for EC CO 2 RR. [116] Wang and coworkers found a novel path toward stabilized SACs on an oxide. They found that large Ag nanoparticles caused the surface reconstruction of MnO 2 as the temperature rose, colliding with the substrate and gradually decreasing in the size. The in situ environmental transmission electron microscopy (ETEM), XAFS, and XRD results confirmed the transformation of Ag nanoparticles to single-atoms. (Figure 12e,f ) The best atomic structure of Ag-O 4 was established after combining the results of EXADS, XRD, and DFT calculations, and the synthesized Ag SAs showed a high FE of 95.7% for CO. [117] 5. Conclusion and Future Challenges EC CO 2 RR is an area of research regarding the current induced transformation of greenhouse gas into CO or other high-value organic materials. Recently, SACs are investigated as a new, novel catalyst for CO 2 RR. Prior works on SACs have assigned the accessibility of the metal active site to be a major factor for high CO 2 RR activity. While reviews with individual focus on the type of metal, the support, and the method of synthesis have been published, these factors can all be summarized into the difference in coordination. Therefore, this review summarizes the CO 2 RR of SACs from a coordination point of view.
The most commonly seen type of SAC is M-N-C, with the coordination of a metal atom surrounded by four nitrogen atoms on a carbon support. While these M-N-C catalysts show FE above previous metal-based bulk catalysts, doping of carbon with other elements induces further improvement. When nitrogen surrounding the metal atom is substituted with other atoms such as O, S, and Cl, it causes a change in the electron configuration of the single-metal atom because of the difference in electronegativities between them and N. The asymmetric SACs or DACs use the change in electronegativity to optimize the adsorption energy of intermediates and thus enhance CO 2 RR efficiency. Furthermore, there are DACs, where different metals or metal atom pairs are atomically dispersed within the support. This provides interaction between metals, especially for metal atom pairs where the single-metal atom is directly coordinated with the other single-metal atom. The FE is further increased by distortion in electron configuration, and a change in the main product is observed depending on the type of metal paired. There are also other kinds of support for SACs. Metal oxides show high electrical conductivity and mass transport due to their low dimensional and porous structures. The most commonly used metal oxide support for EC CO 2 RR applications is CeO 2 . Here, coordination with the support is facilitated by the bonding of metal atoms to surface oxygen vacancies. This bonding in the vacancy sites affects the stability and catalytic activity of these SACs.
The use of SACs in CO 2 RR holds possibilities for catalyst activity, product selectivity, and low cost. Further experimental work should especially be done on SACs with multi-component active sites, such as DACs or SACs with various surrounding atoms, as they show synergistic effects. [118] Especially, DACs hold potential in forming C 2þ products due to the adjacent metal sites being available for C-C coupling. However, the majority of articles are still on CO products and there are mostly theoretical papers for the possible DAC structures with C 2þ products. An example is a DFT calculation article published in 2022, where concerted *CO trimerization to form C 3 product, and further branching into long-chain oxygenates was suggested. [119] There are still many challenges left in applying SACs for CO 2 RR. 1) Synthesis of stable and reactive diatomic metal pairs that do not aggregate while also confining the coordination to one type as opposed to current studies showing a percentage of single atoms in the DAC samples.
2) Detailed analysis of the reaction intermediate states through in situ/operando characterization of the CO 2 RR in SACs to figure out the exact mechanism. 3) Developing machine learning techniques to better the DFT-based filtering of complex multicomponent SAC candidates for CO 2 RR. 4) Verifying the CO 2 RR capabilities of the various screened candidates. 5) Increasing the concentration of SACs for higher catalytic activity.