Harnessing single-atom catalysts for CO 2 electroreduction: a review of recent advances

Electrochemical CO 2 reduction is an eﬀective pathway to convert CO 2 into valuable fuels and chemicals, which provides a potential alternative to fossil fuel resources and plays a notable role in mitigating environmental issues and energy crises. The feasibility of the CO 2 reduction reaction (CO 2 RR) hinges on the development of catalysts that feature high activity, selectivity, and stability. As a new research frontier, single-atom catalysts (SACs) have shown immense potential in the field of CO 2 reduction by virtue of their unique geometric/electronic structures, and have also provided new opportunities for atomic-level understanding of structure–function relationships. Therefore, this review aims to outline recent advances of SACs for CO 2 RR. We start by introducing the current research status and general synthesis strategies of SACs, and then shift our focus to analyzing the various regulation strategies and deciphering the structure–function relationships of SACs in the CO 2 RR. Finally, we pro-pose future directions and opportunities for CO 2 RR-oriented SACs, while also highlighting potential challenges that may be encountered along the way.


Introduction
2][3][4][5] Besides, as CO 2 accounts for over 96% of the Martian atmosphere, utilizing CO 2 resources for Mars exploration is also a new research direction for future human space programs. 6Thus, effective CO 2 conversion and utilization technologies are pressing and important grand research challenges in maintaining carbon neutrality balance and alleviating energy shortages.][19][20] Furthermore, the CO 2 RR can be driven by renewable (but intermittent) energy sources (such as wind, solar, and tidal) without extra CO 2 emission, and the reduction process can be easily performed by adjusting the potential range and current densities under ambient pressure and temperature. 17,21,22owever, the CO 2 RR suffers from two stable CQO double bonds (bond energy, 806 kJ mol À1 ) in the extremely inert linear CO 2 molecule, generally leading to a low energy efficiency with high overpotentials. 23,24Table 1 summarizes the standard potentials E 0 (with respect to the standard hydrogen electrode, SHE) of CO 2 RR processes and the competitive hydrogen evolution reaction (HER) in aqueous media. 17,18,25Specifically, a significant external driving force is needed to initiate the single-electron reduction of CO 2 to the key intermediate *CO 2 À during the CO 2 RR process (CO 2 + e À -*CO 2

À
), or to undergo the proton-coupled electron transfer (PCET) process and produce different hydrocarbons and oxygenates, including (but not limited to) CO, HCOOH, HCHO, CH 3 OH, CH 4 , C 2 H 4 , and C 2 H 5 OH. 26,27In addition, thermodynamically, the reduction potentials for the main reduction products mentioned above in the CO 2 RR typically fall within the range of AE0.2 V from that of the HER. 17However, the presence of a high barrier for CO 2 activation and the involvement of multiple electron transfer steps in the reaction pathways of the CO 2 RR result in sluggish kinetics and side reactions involving the HER.As a result, the CO 2 RR usually shows high overpotentials and poor selectivity.In this regard, developing efficient and highly selective electrocatalysts to promote CO 2 RR technology towards industrial applications is the key to addressing the aforementioned challenges. 20,246][47] Most notably, single-atom catalysts (SACs) have emerged as a new family of catalytic materials by virtue of their distinctive structural properties and exceptional catalytic performances. 48,49As a new class of frontier materials, the SACs feature maximized atomic utilization with their active sites isolated with each other but interacting with the support. 50n 2011, Qiao et al. reported a single-atom platinum catalyst supported on iron oxide nanocrystallites (Pt 1 /FeO x ), and the concept of a ''single-atom'' catalyst was proposed for the first time. 51Since then, the concept was quickly accepted by the scientific community and the scope of relevant materials has broadened significantly.Various materials have been explored as the support (such as carbon materials, 52,53 MXenes, 54 oxides, 55 and carbides 56 ), and the elements for the metal active centers have nearly covered all the non-radioactive metals in the periodic table. 57,58The CO 2 RR over SACs can be traced back to the 1970s, when molecular catalysts (cobalt/nickel phthalocyanines) were used as the active sites. 59With the development of synthesis technology, characterization technology (highangle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) and X-ray absorption spectroscopy (XAS), in particular), and theoretical computational chemistry, the establishment of the structure-function relationship between the structure of SACs and the performance of the CO 2 RR has made great progress.Specifically, SACs have adjustable coordination environments and structures, making it easier to obtain uniform catalytic active centers.9][50] For the CO 2 RR, the reaction process typically involves multiple PCET steps accompanied by complex intermediates, and SACs with well-defined structures have become an ideal research model.This could help to identify the reaction mechanism and to understand the structure-function relationship of SACs in the CO 2 RR, guiding us to develop new advanced electrocatalysts.For example, SACs featuring Ag, Mn, Fe, and Ni single-atom active sites often exhibit excellent CO selectivities. 27In the process of the rational design of CO 2 RR catalysts, researchers have found that the generation of non-CO products (such as C 2+ products) is very difficult and complex. 60Because of the isolation and simplicity of their active sites, SACs are not so competent in some cases, prompting researchers to develop dual-atom catalysts or multiatom catalysts (DACs or MACs) with more than one type of active sites. 61,62][65][66][67] Compared with other types of heterogeneous catalysts, SACs have unique advantages in geometric/electronic structures, which gives them great potential to become marvellous catalysts for the CO 2 RR.In this review, as shown in Fig. 1, we use the metaphor of an alchemical furnace as the main concept, with three supporting pillars representing the general synthesis strategies for SACs.The four different ''elixirs'' within the furnace chamber represent the diverse regulation strategies for SACs in CO 2 RR applications.Eventually, the reactant CO 2 undergoes transformations within this alchemical furnace to produce various products.At the end of this review, we present an outlook on the future directions, opportunities, and challenges in the field of SACs for the CO 2 RR.

Fundamental parameters of the CO 2 RR
It is necessary to introduce the fundamental parameters before proceeding further with this review.At present, the standard metrics used to evaluate and compare the performance of CO 2 RR electrocatalytic materials mainly include the following: faradaic efficiency (FE), onset potential, current density, partial current density, stability, turnover frequency (TOF). 25,68.1.1Faradaic efficiency.FE refers to the percentage of the number of electrons consumed in the generation of the target product to the total number of electrons consumed, which can be calculated using the following formula: FE = anF/Q.Among them, a is the number of electron transfers that generate a specific product (for example, if CO 2 is reduced to carbon monoxide (CO), the number of electrons transferred in this process is 2, a = 2); n is the total amount of a specific product (in moles); F is the Faraday constant (96 485 C mol À1 ); Q represents the total amount of charge during the CO 2 RR process.By analysing the FE, the difference in selectivity of different catalysts for the CO 2 RR can be directly compared.
2.1.2Onset potential.Thermodynamically, CO 2 RR is a non-spontaneous process.The onset potential is the lowest externally applied potential that can initiate the CO 2 RR (with the generated target product detectable).Since there is a certain potential barrier to overcome in the process of the CO 2 RR, the onset potential is generally more negative than the standard reduction potential.Overpotential is the difference between the thermodynamic standard potential and the external potential of a specific CO 2 RR, typically used to activate inert CO 2 molecules to form *CO 2 À anionic radicals, driving material exchange and electron transfer in the CO 2 RR process.The onset potential is the parameter that can directly indicate the magnitude of the overpotential and assess the catalytic activity of the electrocatalyst.

Current density.
Current density refers to the current passing through a unit electrode area, which is the ratio of the current magnitude to the electrochemical active area, and can be used to estimate the reaction rate.The magnitude of current density is one of the indicators of whether a catalytic system can be commercialized.
2.1.4Partial current density.Partial current density expresses the effective current density that drives the formation of a specific product, which can be obtained by multiplying the FE for that specific product by the total current density.Commonly, the partial current density is not only dependent on the activity of the electrocatalyst itself, but also influenced by experimental conditions (electrodes, cell components, electrolyte, etc.).For example, the H-type cell configuration is limited by the solubility and diffusion of CO 2 in the aqueous electrolyte, whereas the flow cell configuration is equipped with gas diffusion electrode components (which can continuously provide CO 2 gas and electrolyte), thus achieving high current density. 69.1.5Stability.Stability is an important metric that indicates whether a catalyst can be commercially applied.Therefore, stability is no less important for a catalyst than its activity and selectivity.At present, the stability tests of catalysts in the laboratory usually take only a few dozen hours, which is far from enough for the commercialization of catalysts.The decline of stability is usually caused by the coupling of the deactivation of the catalytic active site, the destruction of the catalyst structure, and the continuous consumption of electrolytes.An ideal CO 2 RR catalyst should be designed to retain both current density and FE without increasing the overpotential during long-term continuous operation.
2.1.6Turnover frequency.For decades, TOF has been an accurate descriptor for the intrinsic activity of catalysts.While current density is important in determining the apparent activity, but it is affected by parasitic reactions centered on the catalyst and competitive reactions centered on the electrolyte.Fortunately, TOF provides a precise measure of the intrinsic activity without involving these electrocatalytic phenomena.TOF refers to the quantity of target products produced on one active site in unit time, and its calculation formula is as follows: TOF = [product (mol)]/[active sites (mol)]/time.Among these, there are multiple analytical methods available to quantify reactants or products; however, determining the exact number of active sites is often the most challenging task.In particular, for cluster/nanoparticle catalysts, even when the active sites are identified, each site may have a different chemical environment.In contrast, for SACs, the situation is much simpler.Currently, it is presumed that the active sites in SACs are isolated, and their chemical environments are almost uniform.Hence, it can be assumed that each individual metal atom acts as an active site. 25g. 1 Schematic diagram of the SACs for the CO 2 RR in this review (including the general synthesis strategies for SACs and the regulation strategies of SACs for the CO 2 RR).

Reaction mechanisms of the CO 2 RR
The CO 2 RR is a rather complex process involving multiple PCET steps, with different reaction pathways associated with 2-18 electron transfers, resulting in a wide distribution of products, including C 1 products (such as CO, HCOOH, CH 3 OH, HCHO, and CH 4 ), and C 2+ products (such as C 2 H 4 , C 2 H 5 OH, CH 3 COOH, and CH 3 COCH 3 ).Great efforts have been made both experimentally and theoretically to elucidate the CO 2 RR mechanism, but there is still a lack of consensus on some reaction steps, and their reaction mechanisms are still not fully understood.More in situ characterization techniques, even more advanced coupled multi in situ characterization techniques, as well as theoretical calculations are needed to further decipher the CO 2 RR reaction mechanism.Here, we summarize the possible reaction pathways of C 1 products and the main C 2 products based on the current prevailing understanding of CO 2 RR mechanisms, so as to design high-activity, highselectivity, and high-stability SAC catalysts rationally.
2.2.1 Pathways for C 1 products.CO is the most common product of the CO 2 RR, and a large portion of SACs (such as Fe, Co, Ni, and Mn) exhibit high CO selectivity (Fig. 2). 70Although researchers have made significant efforts to analyze the pathways of CO generation, a clear picture of the reaction mechanism remains controversial.1][72][73][74] The resulting *COOH intermediate would undergo the next PCET process to convert into *CO, and the weakly bound *CO would eventually desorb from the catalyst surface to produce CO. Unlike other C 1 products, the generation of HCOOH or HCOO À has a reaction pathway of its own.The generation of HCOOH or HCOO À may be achieved through an intermediate, which is believed to be mainly through the binding of oxygen atoms to the catalyst surface, and the generation of this intermediate may have two different pathways: (1) CO 2 À radicals bind to the catalyst surface and react with adjacent water molecules or protons to generate HCOO À or HCOOH; (2) CO 2 is directly inserted into metal-hydrogen or metal-hydroxyl bonds to form intermediates, which further yields HCOOH or HCOO À . 70,772.2.2 Pathways for C 2 products.The reaction mechanisms of C 2 products are even more complex.Here we summarize the plausible reaction mechanisms of the two main C 2 products: C 2 H 4 and C 2 H 5 OH (Fig. 3).*CO is considered a key intermediate for the formation of C 2 products.It can directly form the *C 2 HO 2 intermediate through symmetric dimerization followed by further hydrogenation.Alternatively, it can first undergo hydrogenation to produce *CHO or *COH, which then asymmetrically couple with *CO to yield the *C 2 HO 2 intermediate. 13,71,78Among the various catalysts developed so far, Cu-based catalysts are the only ones widely recognized capable of catalyzing the CO 2 RR to C 2+ products because of their moderate binding energy with *CO. 79,80After the C-C coupling, multiple PCET steps ensue to obtain the key *C 2 H 3 O intermediate that determines the formation of C 2 H 4 and C 2 H 5 OH products.However, in most catalyst systems, the selectivity for C 2 H 4 is higher than that for C 2 H 5 OH because the energy barrier from the *C 2 H 3 O intermediate to the C 2 H 4 product is lower than to the C 2 H 5 OH product. 81Although Cu is considered to have unique advantages in the selectivity of C 2+ products, there are few reports on Cu SACs for producing C 2+ products, as it is difficult to promote C-C coupling or *CO dimerization on isolated Cu single atoms.[84]

Synthesis strategies of SACs
The term ''SACs'' was coined on the basis of the characteristics of their active centers; however, in reality, individual atoms cannot exist in isolation but instead rely on a supporting material. 48,49Owing to their high surface energy, isolated metal atoms have a tendency to aggregate into thermodynamically more stable metal clusters.Hence, the interaction between the support and single atoms is crucial in preventing the aggregation of individual atoms.Since the concept of ''single-atomcatalysts'' emerged in 2011, 51 various strategies have been proposed to synthesize SACs with high metal loadings and high CO 2 RR performances.In this section, we mainly discuss the challenges and current strategies for increasing the metal loading and tuning the electronic structure of SACs, which are categorized into ''heteroatom coordination strategy'', ''spatial confinement strategy'' and ''defect design strategy''.In addition, some other advanced synthesis strategies of SACs are also introduced.
3.1.1Heteroatom coordination strategy.The rational design of coordination sites with the aim of adsorbing and binding metal precursors or metal atoms to prevent their agglomeration and migration has been a basic design principle for the synthesis of SACs.Owing to the strong interaction between metal atoms and coordinating atoms, metal single atoms can be uniformly and firmly anchored onto the supports.The construction of highly active and stable SACs using the coordination strategy requires the selection of suitable reactants (e.g., metal precursors and coordination polymorphs) and appropriate experimental conditions.Several non-metallic elements with lone pairs of electrons are good coordination sites because of their affinity for metals, such as N, O and S. 20,21,48,85 Many studies have already shown that the physicochemical state of SACs can be optimized by adjusting the coordination number, coordination structure and coordination environment of the metal single-atom centers, thus further improving their CO 2 RR performance.The coordination number of the central metal atoms can modulate the electronic structure of the catalyst and thus alter the catalytic performance.Wang et al. reported two Co-based catalysts with different coordination numbers, denoted as Co-N 4 and Co-N 2 .The catalysts were prepared by adjusting the pyrolysis temperature to control the extent of volatilization (Fig. 4a). 86Zhang et al. devised a postsynthetic metal substitution (PSMS) strategy in combination with the characteristic that Zn-based nitrogen-doped carbon is easily converted into a Zn-N 3 structure at high temperature. 87he uniformly dispersed Ni-N 3 -C with a metal loading of 0.85 wt.% was obtained, which had a lower *COOH free energy than Ni-N 4 -C.The coordination structure of SACs can also affect the CO 2 RR performance.By replacing part of the Zn 2+ sites on nitrogen-doped carbon nanotubes with Ni 2+ , Hou et al. developed Ni SACs for the efficient conversion of CO 2 to CO. 88 By direct pyrolysis of metal-organic frameworks (MOFs) assembled with Fe and Ni-doped ZnO nanoparticles, Jiao et al. precisely constructed a novel Fe 1 -Ni 1 -N-C catalyst, with neighboring Fe and Ni single-atom pairs decorated on a nitrogen-doped carbon support (Fig. 4b). 89The unique architecture of the carriers facilitates the design of the local coordination environment for SAC active sites.][92] MOFs have great potential for application in the CO 2 RR because of their well-defined metal nodes, tunable organic ligands and ordered pore structure. 93,94The preparation of SACs using MOFs can follow two strategies.First, catalytic organometallic complexes or inorganic metal species are immobilized into MOF pores.Second, the single metal sites are intrinsically created into metal nodes and metallolinkers within frameworks during and before the synthesis process.With MOF carbonization and by firmly embedding single atoms into the carbon skeleton through robust metal-heteroatom (N, O, or S) coordination bonds, SACs with exceptional features and applications can be obtained. 95n recent years, a variety of metal SACs have been successfully synthesized using the coordination design strategy.However, the relationship between the coordination structure of SACs and the performance of CO 2 RR is not sufficiently clear.To design CO 2 RR electrocatalysts with high activity and selectivity, the structure-performance relationships need to be further investigated.
3.1.2Spatial confinement strategy.Spatial confinement has been developed as an effective strategy for constructing SACs with precisely defined microenvironments.7][98][99] Li et al. developed a general host-guest strategy for preparing various metal SACs on nitrogen-doped carbon (M/CN, M = Pt, Ir, Pd, Ru, Mo, Ga, Cu, Ni, Mn) (Fig. 4c). 100 The activity of Ir 1 /CN is 16 times and 19 times that of Pd/C and Pt/C, respectively.In addition, Ir 1 /CN showed high tolerance to CO poisoning.Zhao et al. prepared a catalyst containing single Ni sites using MOFs for efficient electrocatalytic reduction of CO 2 . 101This synthesis is based on ion exchange between Zn nodes and Ni ions adsorbed in the MOF cavity.Jiang et al. studied ZIF-8, ZIF-108, ZIF-7, and SIM-1 with the same SOD (sodium salt) topology and different CO 2 RR organic ligands in an aqueous solution (Fig. 4d). 102In situ XAS measurements and density functional theory (DFT) calculations show that the imidazolate ligand coordinated with the Zn(II) center in ZIFs is the active site of the CO 2 RR.DFT calculations suggest that ZIF-8 is the most ideal catalyst for catalyzing the CO 2 RR.ZIF-8 is currently the most extensively used porous MOF precursor for the preparation of SACs, as its derivative porous N-doped carbon has abundant N sites that can anchor individual atoms.Li et al. prepared nickel-metal catalysts of different sizes: from single atoms to more than 100 nm. 103The suspension obtained after mixing Ni(NO 3 ) 2 with ZIF-8 was centrifuged, and the powder thus obtained was calcined at 1000 1C for 2 h; the Ni precursor was converted into Ni single atoms, and ZIF-8 into N-doped carbon support.For Ni metals with different sizes, the CO 2 RR performance is also different.Ni SAC has an excellent FE for CO.Adli et al. used ZIF-8 to confine the FeN 4 site for the CO 2 RR. 104For FeN 4 sites, the porosity, catalytic surface area, and carbon carrier graphitization were optimized by designing particle size and Fe doping, thereby changing the morphology of the catalyst.The intrinsic activity is controlled using different pyrolysis temperatures, which affects the length and local strain of the Fe-N bonds.DFT calculations suggest that there is a close relationship between the enhancement in intrinsic activity of the CO 2 RR and the optimized local strain of Fe-N bonds.Hai et al. identified GAFRUDs, CAJQEL, and cg400449c as potential catalysts for electrocatalytic CO 2 RR through multistep highthroughput screening of the Computational Ready Experiment (CoRE) MOF database, including structural plausibility checks, pore size screening, adsorption capacity prediction, open metal site identification, CO 2 molecule activation capacity, and reaction path calculations. 105In addition, on the basis of DFT calculation, they propose that the polarity of the coordination bond between the metal atom and the coordination atom in the ligand would significantly affect the activation of CO 2 molecules, which could provide effective guidance for the rational design of MOF-based CO 2 RR electrocatalysts.Furthermore, they proposed that the polarity of the coordination bonds between the metal atoms and the coordinating atoms in ligands has a significant impact on the activation of CO 2 molecules, and the selectivity for HCOOH mainly depends on the adsorption energy difference between *HCOO and *COOH, which could provide guidelines for the rational design of MOF-based electrocatalysts for the CO 2 RR.
In summary, the spatial confinement strategy could significantly suppress the migration and aggregation of metal atoms within the porous carrier, while the synergy between the metal single-atom sites and the shape-selective carrier improves the CO 2 RR performance.The key issue in using this strategy is the selection of suitable metal precursors and support materials with uniform cage/pore sizes to ensure metal species confinement.
3.1.3Defect design strategy.Isolated single metal atoms can be stably anchored on supports through the design of surface defects, which have been developed as an effective method for synthesizing SACs.Various vacancies and coordination-unsaturated sites are often used not only to trap and stabilize metal single atoms, but also to alter the surrounding coordination environment. 57,58Typically, oxygen vacancies in metal oxides have proved ideal sites to anchor metal single atoms.For instance, Cu single-atom sites coordinated with three nearby oxygen vacancies on CeO 2 nanorods have been successfully prepared using a wet impregnation method (Fig. 4e). 106Multiple oxygen vacancy-bound Cu sites yield a highly effective catalytic center for CO 2 adsorption and activation.Nitrogen vacancies have also been developed to anchor metal single atoms through defect engineering in nitrides.Nitrogen vacancies induced by microwave plasma have been reported to trap Ni atoms and build highly defective Ni-pyridinic N 2 moieties for electroreduction of CO 2 to CO.The unsaturated Ni SAC has a lower energy barrier for the formation of intermediates COOH* and CO*, and exhibits high intrinsic activity towards the CO 2 RR. 107Carbon-based materials are often used as support materials for SACs.Carbon supports are activated with abundant defects and different functional groups on the surface to stabilize metal ions.Xie et al. synthesized kilogram-scale single-atom Sn d+ on N-doped graphene through a rapid freeze-drying-calcination method.The mixture solution of graphene and metal salts was quickly frozen to disperse Sn ions onto graphene oxide.The freeze-dried samples were calcined in a mixed Ar-NH 3 atmosphere, which is beneficial for N doping into the graphene to stabilize Sn atoms. 108raphdiyne (GDY) has also been used as the support to construct SACs. 109,110The formed Cu-C 4 moiety has a hybridization orbital between the Cu atom and the graphite alkyne carbon to avoid aggregation of Cu single atoms.In addition, the GDY support could be tailored by introducing electronwithdrawing and -donating groups (-F, -H and -OMe) to synthesize GDY derivatives, and their electronic structures are precisely perturbed by the neighboring functional groups. 109hus, the charge transfer between GDY and the atom sites could be regulated to optimize the product selectivity of the CO 2 RR.Defective metal sites such as metal vacancies and coordination-unsaturated sites can also serve as ''traps'' to anchor isolated metal atoms.Chen et al. report that Pd 10 Te 3 alloy nanowires can form rich Te defects on the surface after etching by NaOH solutions, which is conducive to the formation of stable and unique Cu 0 1 -Cu x+ 1 atom-pair catalysts. 111oreover, a site-specific underpotential electrochemical deposition (UPD) strategy has been developed to synthesize single-atom Cu deposited on defective sulfide-derived Ag nanowires.The defective Ag catalyst and the neighboring Cu atom are stabilized through interfacial interactions. 112n general, the construction of atomically dispersed metal catalysts through defect engineering has proven to be an effective strategy.The type and quantity of defects significantly affect the stability and electronic structure of single atom active centers. 20Therefore, facile and effective methods of defect construction still need to be developed.In addition, it is of great significance to investigate the formation mechanism of isolated catalytic centers at defect sites for the design of SAC catalyst configuration and the synthesis of high-loading SACs.
3.1.4Other strategies.Other than the strategies discussed above, there have been several more novel synthesis strategies proposed for preparing SACs.As a simple but efficient strategy, electrochemical deposition has been widely used to prepare SACs for the CO 2 RR.Ross et al. employed an underpotential deposition method to tune a nanostructured Au electrocatalyst using Cu atoms (Fig. 4f). 113This method provides a quantitative and systematic means, by which the ratio of surface Cu can be finely controlled.Xuan et al. employed an electroplating method to deposit single atoms of precious metals on various 2D materials.With this method, the doping sites were precisely located at Mo-and S-vacancies, and the surface atomic loadings for Pt, Au, and Cu reached up to 1.1, 7.0, and 14%, respectively. 114Wang et al. demonstrated a voltage-gauged electrofiltration method to fabricate Fe SAC from a bulk Fe plate. 115The bulk Fe plate worked as the counter electrode, and were converted into Fe 2+ under a positive potential.With the assistance of graphene oxide film, the diffusion rate of Fe 2+ is greatly reduced, resulting in an ultra-low concentration of Fe 2+ around the working electrode, which successfully prevented the growth of nuclei and the aggregation of metal atoms.Therefore, Fe single atoms were uniformly anchored onto the prepared nitrogen-doped carbon.
Chemical vapor deposition is a novel method for synthesizing single atoms, Qu et al. developed a gas-migration/NH 3mediated strategy to convert bulk materials (Cu, Co or Ni) directly into single atoms. 116At high temperatures, ammonia molecules were used as a medium to drag metal atoms out of bulk metal to form gas-phase complexes, then the complex molecules were captured by defects on the nitrogen-rich carbon support, forming isolated metal sites.Considering that the corrosiveness of NH 3 would pose stringent requirements on the equipment, a new gas-migration strategy that employed volatile metal oxides as a single atom source was demonstrated. 117At high temperatures, metal oxides were evaporated to generate gaseous species, which were transported onto nitrogen-doped carbon and anchored by defects to form isolated metal atom/NC catalysts.Recently, a novel one-step laserplanting strategy was employed to create a set of high-density SACs on various supports, including carbon black, graphene quantum dots, metals, and oxides (Fig. 4g). 118The laser pulses created defects on the support and simultaneously decomposed the precursors into metal atoms, which were subsequently immobilized on the as-produced defects via electronic interaction.Using this strategy, a high loading of single atoms of 41.8 wt.% has been achieved.
In addition to the above-mentioned synthesis methods, a few other special strategies have been proposed including nanoparticles becoming single atoms via solid diffusion, [119][120][121] solid-state atom diffusion, 122,123 in situ thermal atomization methods, 119 electrospinning polymer nanofibers, 124 and crosslinking and self-assembly of graphene quantum dots. 125

Characterization techniques of SACs
Upon providing an overview of the synthesis strategies for SACs, determining the geometric structure and electronic environment of SACs is crucial for investigating their structure-function relationship in CO 2 RR applications.Primarily, ascertaining the geometric structure enables the provision of compelling evidence to discern between SACs and nanoparticle catalysts, thus illuminating the intricate coordination configuration inherent in SACs.Concomitantly, delving into the electronic environment facilitates the acquisition of critical data concerning the electron orbitals and oxidation states of catalytic active centers.While a plethora of techniques exist for material characterization, such as X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), Raman, and others, their applicability may fall short in fulfilling the exigencies of atomic-resolution structural characterization of SACs.In contrast, characterization techniques tailored expeditiously for resolving the geometric/electronic structures of SACs predominantly involve AC-HAADF-STEM, XAS, and Mo ¨ssbauer spectroscopy.Herein, we will not delve into the details of each characterization technique, but rather focus on discussing the distinct characterization techniques tailored for SACs.
3.2.1 AC-HAADF-STEM.Transmission electron microscopy (TEM) stands as one of the most ubiquitous techniques employed for visualizing material morphology.Nevertheless, when dealing with SACs boasting atomic-level resolution, a more sophisticated characterization methodology is in demand to discern delicate morphologies with precision.In recent years, owing to the thriving advancements witnessed in SACs, AC-HAADF-STEM has emerged as an indispensable characterization technique for atomically dispersed active sites in SACs.AC-HAADF-STEM is commonly employed to characterize metal atoms loaded on solid supports.Due to the stronger electrostatic interaction between the metal atomic nucleus and the incident electron beam, atoms with higher atomic numbers (Z) can scatter more electrons at higher angles.Consequently, metal atoms possessing higher Z values manifest as heightened contrasts relative to their supporting substrates, thus facilitating clearer delineation of their atomic configurations. 64This phenomenon is particularly pronounced in SACs on carbon-based substrates, given the stark contrast in Z between metal single atoms and the carbon support.A seminal exemplification in this regard is the work by Chung et al., who reported a Fe SAC supported on carbon and directly observed the Fe-N 4 configuration for the first time by using AC-HAADF-STEM combined with electron energy-loss spectroscopy (EELS) characterization techniques (Fig. 5a and b). 126This study provided crucial guidance for subsequent structural analysis of M-N-C catalysts at the atomic level and their investigation across various catalytic applications.
3.2.2XAS.AC-HAADF-STEM is capable of visualizing single metal atoms loaded on solid supports, particularly carbon supports, by utilizing their different Z contrasts.However, it does not provide information regarding the electronic states and coordination configurations of the metal active sites.In contrast, XAS (including X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS)) is a powerful technique for characterizing the electronic states, coordination and geometric structures of metal active sites. 64ANES is sensitive in capturing information related to average valence states and electron configurations.The near-edge position in XANES corresponds to the oxidation state of the target element, where higher oxidation states require greater energy to excite core electrons.On the other hand, EXAFS provides crucial details about the local geometry and coordination environment of the single atomic sites.By fitting the results from EXAFS, important parameters such as bond length and coordination numbers can be inferred.Additionally, the bonding and coordination configuration of the metal centers can also be reasonably deduced through the analysis of EXAFS data.We previously reported a method for customizing metal active sites, wherein we pre-synthesized mononuclear and binuclear Ag complexes as metal precursors.These precursors were then dispersed on graphene (G) to successfully synthesize Ag1-G and Ag2-G catalysts. 127Combining the customized synthesis strategy with XAS results, and conducting a detailed analysis of other test results along with fitting of EXAFS data, we have confirmed that the coordination configurations of the active centers in Ag1-G and Ag2-G catalysts are Ag-N 4 and AgN 3 -AgN 3 structures, respectively (Fig. 5c-e).
3.2.3Mo ¨ssbauer spectroscopy.Mo ¨ssbauer spectroscopy is an effective technique used to probe the coordination structures and spin states of various Mo ¨ssbauer-active metal centers (such as Fe, Sn, Sb, etc.) in SACs. 128Under ambient temperature testing conditions, normally the nuclear transitions of 57 Fe and 119 Sn have a higher probability of exhibiting the Mo ¨ssbauer effect.Currently, research utilizing Mo ¨ssbauer spectroscopy to investigate the electronic states of SACs is predominantly focused on the elements Fe and Sn.Jin et al. conducted a study where they designed Fe-N 4 catalysts with varying densities of single-atom active sites, aiming to investigate the influence of inter-site distances on the properties of Fe-N 4 . 129Through Mo ¨ssbauer spectroscopy testing, the results demonstrated that at inter-site distances of 0.5 and 1.9 nm in Fe-N 4 , the deconvoluted transmission profiles provided evidence that D1 and D2 doublets dominate most Fe species for the two samples (Fig. 5f).Notably, D1 was assigned to the low-spin Fe(II)-N 4 species acting as a highly active catalytic site, whereas D2 was associated with a medium-spin Fe(II)-N 4 species.Chen et al. reported a copper oxide-supported atomically dispersed Sn catalyst for the CO 2 RR. 130In their study, they employed 119 Sn Mo ¨ssbauer spectroscopy as a probe technique to track the catalyst's metastable species.The Mo ¨ssbauer spectroscopy results revealed that the atomically dispersed Sn 4+ sites on the CuO substrate underwent a transformation from Sn 4+ -O 4 -Cu 2+ to a metastable state of Sn 4+ -O 3 -Cu + during the CO2RR (Fig. 5g)

Coordination environment regulation
The coordination environment of SACs can directly influence the electronic and geometric structure of the metal active site, thereby altering the adsorption/desorption processes of key intermediates in the CO 2 RR.Additionally, different coordination environments can result in distinct microenvironments for the CO 2 RR, such as interfacial water and local pH, playing a significant regulatory role in the reaction process.In this section, we primarily focus on the discussion of two types of coordination regulations: in-plane and out-of-plane coordination.
4.1.1In-plane coordination.Regulation of the in-plane coordination environment typically involves the control of atoms in the first shell of the metal active site, which allows for a direct modification of the oxidation state or electronic structure of the metal centers. 131As is well known, Fe SACs with FeN 4 sites exhibit a lower CO 2 RR reduction onset potential and can efficiently electrochemically reduce CO 2 to CO at low potentials.However, this high activity and selectivity rapidly decrease as the potential increases, and high CO selectivity can only be achieved within a narrow voltage range.The main reason for this is the adsorption equilibrium issue of *COOH and *CO key intermediates on Fe SACs (with a higher formation energy barrier for *COOH and stronger adsorption of *CO).To this end, Zhao et al. replaced N with O atoms and synthesized Fe SACs with a unique FeN 2 O 2 coordination structure using a strategy of thermal decomposition of oxygen-containing MOF precursors (Fig. 6a). 132Compared to N, O atoms have higher electronegativity and show tremendous potential in regulating the reaction kinetics.The synthesized FeN 2 O 2 catalyst showed over 95% selectivity within the voltage range of À0.4 to À0.8 V. Theoretical calculations indicated that, with O coordination, the Fe sites exhibited a lower formation energy for *COOH and facilitated *CO desorption.Recently, Zhang et al. reported for the first time the Ga-N 3 S-PC SASC with a flexible structural feature achieved through coordination with S and P. 133 This unique structure enables the Ga active sites to exhibit excellent CO 2 RR performance with a selectivity of over 90% towards CO.Theoretical simulations suggest that the Ga-N 3 S-PC sites can undergo structural transformations and oscillations during the reaction process, optimizing the adsorption energy of *COOH to adapt to the CO 2 RR process.Compared to the Ga-N 4 structure, Ga-N 3 S-PC exhibits higher activity and stability.
H 2 O is a molecule that directly participates in the reaction, generating H through dissociation for the PCET process.Therefore, it is crucial to regulate the interfacial water for efficient CO 2 RR. 134In this regard, besides directly adjusting the electronic structure of the metal center by optimizing the first coordination sphere, controlling the interfacial microenvironment through non-coordinating heteroatom doping also has significant importance in the development of high-performance SACs.For example, in order to promote the synergistic adsorption and activation of CO 2 and H 2 O, Liu et al. constructed a SASC with FeN 4 /graphitic N atomic interface sites through a H 2 -assisted pyrolysis strategy, which achieved a CO faradaic efficiency of over 90% within the range of À0.3 to À0.8 V. 135 Operando attenuated total reflection surface enhanced infrared absorption spectroscopy (ATR-SEIRAS) and the first-principle calculations indicate that the presence of graphitic N sites facilitates the co-adsorption of H 2 O, thereby promoting the activation of CO 2 and its conversion into the key intermediate *COOH (Fig. 6b).In another work, Sun et al. achieved efficient electrochemical CO 2 to CO conversion on Mo SACs through the incorporation of Se into the carbon support. 136The study demonstrates that the introduction of neighbouring Se (6.0 Å) can effectively improve the reaction thermodynamics of the away from Mo SACs (48.5 Å) can improve the electrode/ electrolyte interface by repelling water molecules to suppress the hydrogen evolution reaction (HER), thereby enhancing the selectivity of the CO 2 RR (Fig. 6c).
4.1.2Out-of-plane coordination.The introduction of axial coordinating atoms can further enrich the coordination environment of the active site in SACs.When they form special configurations by coordinating and bonding with the metal atom center, they can regulate the spin state and electronic properties such as the d-band center of the metal sites. 137This, in turn, allows for modulation of the binding energy between the active site and reactants/intermediates, ultimately achieving control over the intrinsic catalytic activity of the catalyst.By employing a rapid pyrolysis and controlled activation strategy, a FeN 4 -O 1 site with axial O coordination was successfully synthesized by Chen and co-workers (Fig. 6d). 138The results indicate that the introduction of axial O can induce enhanced electron localization at the Fe site, thereby promoting the desorption of *CO and increasing the energy barrier for competitive HERs (Fig. 6e).The synthesized FeN 4 -O 1 SACs demonstrate nearly 100% CO selectivity within a wide potential window (310 mV).Using a similar approach, Pan et al. successfully constructed SACs with a unique Co-N 5 coordination structure.Thanks to its lower *COOH formation energy and facilitated *CO desorption process, this catalyst exhibits an excellent electrochemical conversion efficiency of CO 2 to CO. 139 From the above results, it can be observed that high selectivity in electrochemical CO 2 -to-CO conversion has been achieved through the coordination structure and electronic modulation of SACs.However, there are still limitations in terms of current density and catalyst durability.Recently, Huang et al. made a breakthrough in achieving high selectivity, high current density, and sustained catalyst durability through axial coordination engineering. 140Inspired by natural metalloenzymes, they designed and constructed a single-atom nanoenzyme with a Ni-N 5 -C structure featuring axial N coordination (Fig. 6f).Compared to the traditional planar Ni-N 4 coordination structure, the squarepyramidal NiN 5 site's d z 2 and d xz/yz orbital energy levels were increased and decreased, respectively (Fig. 6g).This facilitates the adsorption and activation of CO 2 molecules and the desorption of CO, greatly enhancing the reaction kinetics.Therefore, the synthesized catalyst achieved a 99.6% CO faradaic efficiency (turnover frequency of 69.7 s À1 ) at an ultra-high current density of 1.23 A cm À2 , and exhibited exceptional stability for 100 hours, surpassing significantly many other reported SACs.In addition to the coordination of N and O, elements such as S, Cl, Br, and I have also demonstrated promising potential in axial coordination engineering. 141,142Overall, the proposal and development of axial coordination regulation will contribute to further investigations of the mechanism of single-atom catalysts, as well as expanding the strategies and methods for the regulation of SACs. 143,144

Diatomic sites regulation
In recent years, DACs have gained extensive attention as extensions of SACs. 145Compared with SACs, DACs possess more complex and flexible active sites, enabling them to achieve better catalytic performance.This is particularly important for the CO 2 RR involving various reaction intermediates.When all intermediates are adsorbed on a single active site, it becomes challenging to achieve optimal adsorption states for each intermediate.By constructing bimetallic active sites, the adsorption energy of different intermediates can be optimized, thereby breaking this linear relationship and achieving greater catalytic performance.Furthermore, the construction of diatomic sites has potential for achieving synergistic adsorption and activation of both CO 2 and H 2 O, thereby further enhancing reaction kinetics.In this section, we will categorize DACs into homonuclear diatomic sites and heteronuclear diatomic sites for discussion.
4.2.1 Homonuclear diatomic sites.Fine-tuning interactions in the second coordination shell of enzymes or homogeneous catalysts is crucial for their functionality.Now, this concept has been applied to heterogeneous catalysts, utilizing copper atom pairs for the selective electrochemical reduction of CO 2 .Jiao et al. reported a copper atom pair catalyst (denoted as Cu-APC) loaded on Pd 10 Te 3 nanowires, where the active sites were characterized as Cu 0 1 -Cu x+ 1 units through AC-HAADF-STEM, XAS and theoretical calculations. 111This unique atom pair structure efficiently promotes the activation and conversion of CO 2 through a diatomic activating bimolecular mechanism, where the Cu x+ 1 site acts as a Lewis acid for adsorbing water molecules, and Cu 0 1 acts as a Lewis base for adsorbing and activating CO 2 .This synergistic proton-coupled electron transfer leads to a significant enhancement in the catalytic activity of Cu-APC catalysts (Fig. 7a).The concept of pairing atoms to achieve better performance opens up new opportunities for the design of SACs, where two are better than one when atomic synergies come into play.Similarly, in a recent study, Hao et al. successfully synthesized Ni dual-atom site catalysts through the in situ transformation of Ni nanoparticles. 96In situ XAS and theoretical calculations revealed that the Ni dual-atom centers can facilitate the adsorption of hydroxyl and form an electron-rich microenvironment at the Ni 2 sites during the CO 2 RR process, and the resulting catalytic microenvironment promotes the formation of *COOH and the desorption of *CO.As a result, the synthesized catalyst achieved a 499% CO faradaic efficiency at an ultra-high current density of 1 A cm À2 (Fig. 7b).
In addition, DACs have demonstrated superior performance and advantages over SACs in the synthesis of multi-carbon products.For instance, by constructing dual-atomic Cu sites, Li et al. successfully achieved efficient production of C 2 products (mainly ethylene and acetic acid) on DACs with a high faradaic efficiency of 91% and a partial current density of 90 mA cm À2 . 146Theoretical calculations suggest that the neighboring Cu sites can facilitate the carbon-carbon coupling process of CO.Interestingly, when one of the Cu atoms is replaced by a Ni atom, this highly efficient carbon-carbon coupling pathway is terminated as a result of the excessive CO adsorption by the Ni atom (Fig. 7c).This finding emphasizes the importance of compatibility between bimetallic centers for the CO 2 RR process.Despite some breakthroughs in the activity and selectivity of DACs for the CO 2 RR, there are still difficulties and challenges in the precise synthesis and controllable preparation of dualatomic sites.To this end, Li et al. have developed a dual-core metal molecular pre-synthesis and post-adsorption strategy. 127hey first synthesized a well-defined binuclear Ag complex and then obtained DACs with AgN 3 -AgN 3 sites through the p-p interaction between the complex and graphene, as well as low-temperature treatment (Fig. 7d).These AgN 3 -AgN 3 sites demonstrated greater advantages in synergistic adsorption and activation of CO 2 compared to single Ag sites.
4.2.2Heteronuclear diatomic sites.The synergistic effect of bimetallic interactions occurs not only between two identical metal atoms but also between two different metal atoms.
The coordination of d orbitals between heteronuclear metals allows for flexible modulation of the active center, reducing energy barriers, and thereby generating unexpected catalytic performance.For example, the rate-determining steps for the CO 2 RR on Ni-NC and Fe-NC are CO 2 (g) -*COOH and *CO -CO(g), respectively, due to the weak adsorption of *COOH on Ni-NC and strong adsorption of *CO on Fe-NC.This naturally leads to a reasonable conjecture and consideration: can their advantages in the reaction be maximized by combining them to achieve an optimal performance?Based on this, Ren et al. ingeniously designed an isolated bimetallic Ni-Fe site. 147tudies have shown that during the CO 2 RR process, the bimetallic Ni-Fe center undergoes structural evolution to a COadsorbed moiety, which greatly reduces the formation energy barrier of *COOH and promotes the adsorption of *CO (Fig. 7e  and f).As a result, the synthesized bimetallic Ni-Fe catalyst exhibits superior electrochemical CO 2 to CO activity.The design strategy for this heteronuclear diatomic pair has also been successfully applied to Co-Cu diatomic site catalysts. 148ecause of the weak binding of *COOH on Cu-SAC, a high energy input is required for CO 2 activation to *COOH.On the other hand, Co-SAC exhibits strong *COOH adsorption and enhanced *CO binding strength.When combined, CoCu-DASC(Co) exhibits a relatively moderate d-band center value compared to Co-SAC and Cu-SAC.This implies that the adsorption strength of intermediates on the CoCu bimetallic sites is moderately balanced.As a result, the prepared diatomic CoCu catalyst can maintain over 95% CO selectivity in the range of 100 mA cm À2 to 500 mA cm À2 , with a maximum CO partial current density of 483 mA cm À2 , far exceeding the industrial current density requirement (4200 mA cm À2 ).
Modulating the activation mode of CO 2 molecules through synergistic effects between neighboring atoms is an attractive yet challenging strategy.Pan et al. utilized neighboring Te atoms to induce and regulate the Fe site and constructed Fe-Te diatomic sites for synergistic CO 2 RRs. 149The results demonstrate that the low-valent Fe d+ forms bonds with one carbon and one oxygen atom of the CO 2 molecule, while the adjacent Te d+ acts as an electron donor, regulating the electronic structure of the Fe d+ site and stabilizing the other oxygen atom of the CO 2 molecule.This enhances the 2p u orbital interaction of CO 2 and makes the CO 2 molecule more prone to bending, thereby reducing the activation barrier and improving the CO 2 RR performance (Fig. 7g).In addition to transition metals and main group metals, rare earth metals have shown significant potential in modulating the catalytic activity of metal sites and activating CO 2 due to their unique electronic structure and catalytic properties.Therefore, designing and constructing heteronuclear diatomic sites containing rare earth elements show promise for further tuning the reaction pathways and performance of CO 2 RRs.For instance, Liang et al. synthesized atomically dispersed InCe/NC catalysts using a one-step method. 150Experimental results demonstrated that the introduction of Ce promotes the electrochemical conversion of CO 2 to formate, while the conversion of CO and the hydrogen are suppressed.Theoretical calculations revealed that the incorporation of single Ce atoms not only significantly facilitates electron transfer but also optimizes the In-5p orbitals, enhancing the adsorption energy of the key intermediate *OCHO on the In sites along the formate pathway.

Tandem catalytic regulation
Currently, electrochemical CO 2 reduction for the production of simple 2e À products such as CO or formate has achieved high selectivity and activity, particularly when utilizing single-atom catalysts.However, significant challenges persist in the production of hydrocarbons or alcohols from the CO 2 RR.It is generally understood that the conversion of CO 2 to multi-electron deep reduction products typically involves a two-step sequential process: 151 (1) CO 2 reduction to CO; (2) further reduction of CO to obtain deep reduction products.The electrochemical CO 2 reduction process comprises multiple coupled or consecutive proton-electron transfer steps, and the design and synthesis of tandem catalytic systems using single-atom catalysts as CO-forming or water activation sites show significant potential for facilitating the production of high-value products in these two steps.
4.3.1 SACs serving as CO formation centers.Theoretical studies have revealed that the overall selectivity of C 2+ products is largely dependent on the reaction kinetics of *CO-*CO coupling, which is directly proportional to the square of the *CO coverage.While the *CO coverage can be enhanced by improving local CO concentration. 152,153As discussed above, SACs exhibit exceptionally high selectivity and activity in the process of CO 2 reduction to CO. Utilizing SACs for the first step in tandem catalysis, converting CO 2 to CO, and subsequently coupling it with CO to other product catalytic sites (such as copper), has great promise for the electrochemical reduction of CO 2 to multi-electron deep reduction products.
The potential matching between different materials needs to be carefully considered prior to designing such a tandem catalytic system, where the high selectivity generation of CO and subsequent C-C coupling processes need to be simultaneously achieved at a specific potential.Generally, for Cu-based catalysts, C-C coupling often occurs at relatively negative potentials (4À1 V vs. RHE).Therefore, it is essential for SACs to have a broad potential range while maintaining a high CO faradaic efficiency.According to reports, Ni SACs have been shown to maintain a CO selectivity of over 97% even at À2.4 V vs. RHE, making them an ideal candidate material for these tandem catalytic systems. 140For instance, Yin et al. prepared hybrid tandem catalysts by combining Ni single atoms loaded on high surface area carbon supports with Cu nanowires (Fig. 8a). 154In situ SEIRAS studies revealed that the incorporation of Ni single atoms significantly enhanced the enrichment of CO on the Cu surface, thereby facilitating the C-C coupling process.By adjusting the ratio of hybrid components, the optimized catalyst achieved a 66% ethylene faradaic efficiency in a flow-cell device, which was improved by five times compared to pure copper.Similarly, Meng et al. prepared a nonprecious metal-based tandem electrocatalyst, PTF(Ni)/Cu, by anchoring Ni single atoms and loading Cu nanoparticles onto a porphyrin-based triazine framework. 155During the CO 2 RR process, the single-atom Ni efficiently reduces CO 2 to the intermediate CO, which is immediately converted to ethylene through highly efficient C-C coupling reactions catalyzed by nearby Cu particles.Therefore, compared to the non-tandem catalyst PTF/Cu (triazine framework with no metal Ni atoms in the porphyrin center) that mainly produces methane, the faradaic efficiency of ethylene has increased by fivefold (À1.1 V vs. RHE) (Fig. 8b).
Potential matching determines the compatibility of different components, while spatial distribution determines the efficiency of the tandem reaction. 156The generated CO on SACs often needs to spill over to adjacent Cu sites, and if the distance between them is too far, the tandem process will be hindered.

EES Catal.
Wang et al. systematically studied the influence of spatial distribution between two components on the tandem reaction process. 157They chose CoPC molecules, known as excellent COproducing small molecule catalysts, and Cu nanocubes (Cu cub ) as a model, and prepared two distinct mixed catalysts.One catalyst was CoPC-Cu cub /C, where CoPC and Cu cub shared an interface; the other was CoPC-C/Cu cub , where CoPC was initially loaded onto carbon black and subsequently mixed with Cu cub .Electrochemical tests revealed that CoPC-Cu cub /C exhibited nearly double the selectivity for C 2 products compared to Cu cub /C, whereas the faradaic efficiency of CoPC-C/Cu cub decreased by half.This outcome elucidates the surface transport mechanism of CO intermediates between the two components of the tandem catalyst, highlighting the importance of their relative spatial distribution for achieving efficient tandem reactions.In addition to the tandem systems between single atoms and metal Cu, tandem catalytic pathways can also be formed between different single-atom sites.For instance, Lin et al. reported a tandem catalytic system using CoPC and Zn-N-C for the deep hydrogenation reduction of CO 2 to produce methane. 158The research results demonstrated that CO 2 was initially reduced to CO on CoPC, then diffused to Zn-N-C, where it undergoes further reduction to CH 4 at the ZnN 4 site (Fig. 8c).Compared to using CoPC or Zn-N-C alone, the tandem catalyst showed more than 100-fold increase in the CH 4 /CO production rate ratio.
4.3.2SACs serving as water activation centers.The selectivity of specific products is also influenced by the rate-determining step dominated by PCET for the formation of key intermediates.For instance, *CO protonation to *CHO for methanol, methane, or ethylene production; and *HC-COH protonation to *HC-CHOH for the production of CH 3 CH 2 OH. 20,159,160These processes are closely related to the coverage of adsorbed protons (*H ad ), which typically originate from the dissociation of water molecules.Therefore, the introduction of atomically dispersed water activation sites can facilitate the dissociation of water molecules to optimize *H ad coverage, thereby promoting the generation of key intermediates.In comparison to metal clusters or particles, atomically dispersed sites may have the advantage of avoiding an excessive increase in *H ad coverage, which could otherwise lead to the promotion of the HER side reaction. 161n the reaction pathway of CO 2 reduction to CH 4 , the further hydrogenation of the intermediate *CO is often considered the rate-determining step, while the slow water dissociation kinetics limits the rate of proton supply, thereby impeding the electrochemical methanation of CO 2 .To address this issue, Chen et al. successfully introduced single-atom Ir into copperbased catalysts, resulting in the synthesis of Ir single-atom-doped Cu 3 N/CuO hybrid materials. 162This hybrid catalyst exhibits exceptionally high methane selectivity and activity (with a 75% methane Faraday efficiency at a current density of 320 mA cm À2 ), surpassing catalysts lacking Ir single-atom doping.Experimental and theoretical results indicate that Ir single atoms facilitate the dissociation of water into protons and supply them to Cu 3 N/ CuO sites for the protonation of *CO to *CHO, significantly accelerating the kinetics of CO 2 methanation (Fig. 8d).It has been indicated that methanol and methane products share a common intermediate, *CHO.After overcoming the energy barrier of the rate-determining step along the C 1 pathway (i.e., *COto-*CHO), the critical branching step for generating methanol or methane involves either continued hydrogenation (resulting in *CHOH, *CH 2 OH) or dehydroxylation (leading to *CH, *CH 2 , etc.). 163,164Therefore, further optimizing the *H adsorption on the catalyst surface shows promise for enhancing methanol selectivity.Building on this insight, Zhang et al. synthesized a Rh-doped Rh 1 Cu 4 single-atom alloy catalyst. 161DFT results demonstrate that isolated Rh sites enhance *H coverage on the Rh 1 Cu 4 catalyst and facilitate the hydrogenation of *CH2OH, leading to methanol production.The Rh 1 Cu 4 catalyst exhibits the highest methanol partial current density of 111.7 AE 12.8 mA cm À2 with a Faraday efficiency of 46.2 AE 5.3%.Furthermore, comparative experiments reveal that the introduction of Rh clusters results in excessive *H coverage on the catalyst surface, leading to a decrease in methanol selectivity and an increase in hydrogen production (Fig. 8e).This further shows the advantage of single-atom sites as water activation or *H adsorption sites.
In summary, SACs can promote the supply of CO and protons during the CO 2 RR process by facilitating the generation of CO and the dissociation of water.Subsequently, these intermediate products can be captured by another active site in the tandem catalyst, forming *CO and *H ad species, leading to the acceleration of the rate-determining steps involving C-C coupling and PCET.

Catalyst support regulation
Choosing the appropriate support material is essential to providing strong binding through ion or covalent interactions, thus securely immobilizing single atoms.Additionally, the specific characteristics of the support can strongly influence the electronic states of SACs, further impacting the catalytic performance of the CO 2 RR.The majority of SACs used for the CO 2 RR are carbon-based due to their excellent conductivity.Metals, metal oxides, and other supports with unique structural properties and coordination environments can offer more possibilities for tuning the performance of SACs.
We are well aware that copper is the only metal that exhibits appreciable activity towards the multi-electron transfer hydrocarbon products.In particular, Cu SACs exhibited distinct performance compared to non-Cu-based SACs in the generation of products such as methane, ethylene, acetone, etc. 82 However, Cu SACs usually undergo reconstruction involving atomic re-arrangement and compositional change during the CO 2 RR process, and this uncontrollable electrochemical reconstruction not only leads to the inherently poor stability of Cu-based SACs, but also makes the investigation of real active sites and the establishment of structure-performance relationships shrouded in fog.Traditional carbon-based supports have shown some limitations in stabilizing atomically dispersed metal Cu. 165 Therefore, there is an urgent need to develop novel supports that can stabilize copper-based SACs to advance the field of CO 2 RRs.Recently, inspired by traditional Cu/SiO 2 heterogeneous catalysts, Tan et al. have developed a method to stabilize atomic Cu by constructing a Cu-O-Si interface (Fig. 9a). 166The as-prepared CuSiO x catalyst achieved an efficient electrochemical CO 2 methanation with a high faradaic efficiency of CH 4 (72.5% at À1.27 V vs. RHE) and excellent stability.To address the issue of poor conductivity of the oxidebased support, the authors also assembled a flow-cell device and achieved a remarkable CO 2 -to-CH 4 conversion rate of 0.22 mmol cm À2 s À1 .More importantly, in situ X-ray absorption spectrum experiments indicated that the atomic Cu-O-Si interfacial sites exhibited ultra-high electrochemical stability and reconstruction-resistance over a broad potential window or long-time potentiostatic electrolysis, which attribute to the strong interactions between Cu and silica (Fig. 9b).Using CeO 2 as a support to enhance the stability of Cu is also a feasible approach.Zhou et al. reported a CeO x -stabilized Cu 2+ strategy (Cu-CeO x ) in which CeO 2 acts as a self-sacrificing component to protect the atomic Cu 2+ active species. 167They found that after CO 2 RRs, the concentration of Ce 3+ in Cu-CeO x increased significantly, as evidenced by in situ Raman spectroscopy experiments, indicating that electrons preferentially reduce Ce 4+ to Ce 3+ in Cu-CeO x .Meanwhile, the newly formed Ce 3+ /Ce 4+ couple acts as a conductive network to enhance the conductivity of the overall solid solution.It can accelerate electron transfer and suppress the accumulation of electrons at Cu 2+ sites, preventing them from self-reduction.
Oxygen vacancies or Lewis acid sites on metal oxides can also be utilized to modulate the activity of SACs.For instance, Wang et al. found that a single Cu atom substituted on the CeO 2 (110) surface could stably enrich three oxygen vacancies around each Cu site, leading to an efficient CO 2 adsorption and activation center that promotes methane production. 106ES Catal.
Chen et al. have demonstrated through theoretical predictions that the Lewis acid sites in metal oxides such as Al 2 O 3 and Cr 2 O 3 not only facilitate the activation of CO 2 but also lower the energy barrier for the *CH 4 O intermediate, providing a favourable pathway for methane production. 168Based on these findings, the authors synthesized Cu SACs supported on ultrathin Al 2 O 3 nanosheets with strong Lewis acid sites.The catalyst exhibited an excellent electrochemical CO 2 -CH 4 conversion performance, with a methane faradaic efficiency of 62% at À1.2 V vs. RHE.Yin et al. demonstrated that strong binding sites for CO 2 adsorption can be formed at the Cu-CeO 2 interface, promoting CO 2 activation and conversion into *CO intermediates on the nearby Cu surface. 169As the Cu loading increases, the CO 2 reduction product can be tailored from C 1 to C 2+ .
Using Cu-based materials as supports (such as metallic copper or copper oxides) and introducing atomically dispersed heteroatoms into Cu is a highly effective approach for modulating the activity of Cu-based catalysts.Previous researches have indicated that Cu + sites played a crucial role in improving the selectivity of C 2+ products. 170It can not only enhance the CO adsorption and facilitate C-C coupling reactions but also stabilize C 2 intermediates.However, Cu + was found to be unstable under negative potentials during the CO 2 RR process.Feng et al. synthesized Gd-doped CuO x catalyst (Gd 1 /CuO x ) using a simple one-step solvent thermal method. 171The research findings demonstrated that atomically dispersed Gd can stabilize Cu + and induce tensile strain in the catalyst.The prepared Gd 1 /CuO x exhibited excellent performance in CO 2 -to-C 2+ product conversion, achieving a faradaic efficiency of 81.4% for C 2+ products at À0.8 V. Experimental and computational results suggested that Gd doping improved the stability of the crucial intermediate O*CCO, reduced the energy barrier for C-C coupling, and resulted in high CO 2 RR-to-C 2+ product performance (Fig. 9c and d).Zheng et al. synthesized a Pballoyed Cu catalyst, denoted as Pb 1 Cu, by incorporating Pb single atoms onto the surface of metallic copper (Fig. 9e). 38The introduction of Pb had a significant impact on the protonation process during the CO 2 RR.Rather than occurring on the oxygen site, the initial protonation was observed to preferentially take place on the carbon site of CO 2 when Pb was present.This led to the formation of the *HCOO pathway instead of the *COOH pathway, resulting in Pb 1 Cu demonstrating exceptional formate activity with an ultra-high faradaic efficiency of approximately 96% at current densities 1 A cm À2 (Fig. 9f).Moreover, the researchers utilized the Pb 1 Cu catalyst to construct a solid electrolyte reactor, enabling the continuous production of pure formic acid solution for a remarkable duration of 180 hours, while maintaining a high current density above 100 mA cm À2 .These findings highlight the potential of Pb 1 Cu as a promising catalyst for the CO 2 RR and its application in the efficient and sustained production of formic acid.
In conclusion, adjusting the type, composition, and structure of the supports can optimize the electronic and geometric properties of SACs, enhancing the catalytic activity and stability.This holds significant importance for the efficient, stable, and sustainable development of SACs for the CO 2 RR.

Conclusions and perspectives
The unique geometric/electronic structures of SACs qualify them as a new family of catalytic materials and a new frontier in the research field of CO 2 RRs.In this review, we introduced the research status of SACs and summarized the synthesis strategies for SACs.Furthermore, we briefly introduced the fundamental parameters in the CO 2 RR and attempted to summarize the possible reaction mechanisms for generating C 1 and C 2 products in the CO 2 RR.Most importantly, we focused on discussing the structure-function relationship between the structure of SACs and the performance of CO 2 RR, mainly manifested in the following four aspects of regulation: coordination environment regulation, diatomic sites regulation, tandem catalytic regulation, and catalyst support regulation.Although significant progress has been made in the synthesis of SACs as well as in the application of SACs in the CO 2 RR and the establishment of their structure-function relationships, there are still certain limitations or challenges to the practical application of SACs in CO 2 RRs.Summarized below are some potential research directions and challenges for SACs in CO 2 RR applications in the future (Fig. 10).

Precise synthesis and characterization of SACs
The precise synthesis of SACs has always been a challenge for their application in various fields, including the CO 2 RR.Metal single atoms are often fixed by the coordination of heteroatoms on the support owing to their high surface free energy.At present, the preparation of SACs often resorts to hightemperature calcination steps, which may lead to a certain degree of randomness in the generation of SACs.Many studies have shown that minor alterations in calcination temperature can lead to major changes in the coordination environment of SACs, and needless to say, their catalytic activity also changes accordingly.Therefore, when a high-temperature calcination step is used to prepare SACs, strict control of calcination conditions is necessary to ensure the repeatability of the preparation of different batches of catalysts.Obviously, a more precise strategy for preparing SACs has always been pursued, especially for the preparation of high-loading SACs, which currently faces greater difficulties.In addition, as the research on SACs in different applications deepens, there is increasing interest on multi-site catalysts (such as DACs).However, research on DACs is currently still in its early stages, and many basic concepts and standards have yet to be established.For instance, the characterization of SACs mainly depends on AC-HAADF-STEM and XAS technologies.AC-HAADF-STEM records the projected images of metal single atoms under the incident electron beam, and cannot distinguish the spatial arrangement of metal single atoms in close proximity; on the other hand, XAS records the average signal of the test sample, and cannot separate the information of a single active site.The limitations of these characterization techniques are particularly prominent in the analysis of DACs.Especially in the AC-HAADF-STEM images of DACs, the spatial arrangement of adjacent metal atoms cannot be determined by simply reading the highcontrast bright spots in the projected image while ignoring their spatial relationship along the incident electron beam.In this regard, one of our recent works attempts to identify ''true/false'' DACs by using in situ rotation operation in AC-HAADF-STEM characterization. 172Therefore, more precise synthesis strategies and more advanced characterization methods need to be further explored, as the precise synthesis and characterization of SACs are prerequisites for understanding their structure-function relationship.

Rational design of SACs for producing non-CO products
As summarized in this review, considerably satisfactory CO selectivity has been achieved from the CO 2 RR over a number of SACs, such as Fe, Ni, Mn, and Ag, but it still poses a great challenge for high-value-added products beyond CO.There are two main reasons for this: (1) on the one hand, a single metal active site has moderate binding strength to the intermediate of CO 2 RR, especially *COOH and *CO.Thus, *CO has a moderate desorption energy barrier, which is beneficial for the production of CO. (2) On the other hand, the atomically dispersed metal active sites are isolated from each other, leading to the physical separation of the intermediates, thus blocking further C-C coupling.Owing to its complex reaction mechanism and diverse reaction pathways, it is rather difficult to precisely control the distribution of the reaction products in the CO 2 RR.In order to expand the product range of the CO 2 RR and to obtain high-value-added products, two promising strategies could be explored.One is to rationally design DACs with multiple reaction active sites, which can provide mutually related synergistic active sites.The interrelated multiple active sites could provide a heteroatomic interface for asymmetric C-C coupling, which is expected to yield specific multi-carbon (C 2+ ) products.effective strategy is to construct tandem catalytic systems.The general principle is to combine SACs with high CO selectivity with Cu-based catalysts to achieve continuous conversion of CO 2 to CO and further to multi-carbon products, thus achieving high C 2+ Faraday efficiencies.This method can promote the transfer of reaction intermediates and electrons, thereby improving reaction efficiency and selectivity.In addition, the tandem catalytic system lessens the difficulties in the catalyst design, and makes the catalytic reaction more controllable and predictable.

Deciphering the catalytic mechanisms of CO 2 RRs
At present, significant progress has been made in both the synthesis of SACs and their application in CO 2 RRs.However, owing to the diverse reaction pathways and complex reaction intermediates, the dynamic process and reaction mechanism of the CO 2 RR are still difficult to understand.The catalytic activity of the CO 2 RR is highly dependent on its active sites; therefore, the structure of the active site and its dynamic evolution during the catalytic reaction process are crucial.However, conventional characterization methods often detect the state of catalysts or catalytic reactions at a certain point in time but cannot monitor their dynamic evolution process.Therefore, it would be very helpful to use in situ technologies to identify the dynamic evolution of active sites and capture catalytic reaction intermediates, but this also requires more stringent test conditions.Currently, in situ tests are often conducted separately, such as in situ XAS, in situ TEM, and in situ Raman.It is expected that more advanced coupled in situ test techniques could be developed in the future to facilitate in-depth exploration of reaction mechanisms.In addition, machine learning has made significant progress in the preparation of nanomaterials and its application in the CO 2 RR, and it is expected to be introduced into the construction and performance prediction of SACs in the future.Machine learning can achieve high-throughput material performance prediction and assist in analyzing catalytic mechanisms, thus greatly reducing the trial-and-error costs.

Industrial-level applications
The industrial application of SACs in the CO 2 RR is mainly focused on two aspects: large-scale synthesis of SACs and industrial-level performance of SACs in the CO 2 RR.At present, a number of methods have been developed in the laboratory for the preparation of SACs, but a significant proportion of these methods are rather complex and cumbersome.The complex or harsh preparation conditions would result in poor repeatability of SACs preparation, thereby limiting its large-scale production.Moreover, apart from considerations regarding its repeatability, large-scale production of SACs also needs to balance economic benefits, which drives the preparation of SACs towards simpler, more efficient, and more reliable directions.On the other hand, in addition to achieving industrial-level preparation, SACs also need to meet industrial-level test conditions and performance indicators.As we already know, the low solubility of CO 2 in aqueous electrolytes greatly constrains its mass transport at high current densities, which is also a bottleneck for the industrialization of the CO 2 RR.To address these issues, it has been attempted to combine SACs with gas diffusion layers and the assembly of membrane electrode components in flow cells, which helps to meet the requirements of industriallevel reactions and has made some notable progress thus far.However, reactors often fail during operation owing to various problems such as catalyst aggregation and membrane flooding, and these issues still need to be addressed one by one in future device design.In addition, the stability of SACs is also one of the evaluation metrics for their potentials of industrialization.Currently, the research on SACs in the CO 2 RR is mainly focused on catalytic activity and selectivity, while stability tests often fail to meet the industrial requirements.Therefore, it is necessary to strengthen the research on catalyst stability in the future, to explore the deactivation mechanism of catalysts, and to design efficient SACs that feature balanced activity, selectivity, and stability.

Fig. 2
Fig. 2 Schematic diagram of the reported elements in SACs for CO 2 RR.

Fig. 3
Fig.3The possible reaction pathways of C 1 and C 2 products in the CO 2 RR.

©
2023 The Author(s).Published by the Royal Society of Chemistry EES Catal.

©
2023 The Author(s).Published by Royal Society of Chemistry EES Catal.

Fig. 10
Fig. 10 Schematic illustration of future directions and challenges of SACs for the CO 2 RR.
Peterson et al. demonstrated that the most favorable thermodynamic pathway for the CO 2 RR on the Cu(211) surface is the formation of *CO and subsequent hydrogenation to generate *HCO, *H 2 CO, and *H 3 CO. 75Among them, *H 2 CO can selectively desorb to form HCHO or form *H 3 CO, and the *H 3 CO can further form CH 3 OH through the PCET process.Nie et al.