Asymmetric Coordination Environment Engineering of Atomic Catalysts for CO2 Reduction

Single-atom catalysts (SACs) have emerged as well-known catalysts in renewable energy storage and conversion systems. Several supports have been developed for stabilizing single-atom catalytic sites, e.g., organic-, metal-, and carbonaceous matrices. Noticeably, the metal species and their local atomic coordination environments have a strong influence on the electrocatalytic capabilities of metal atom active centers. In particular, asymmetric atom electrocatalysts exhibit unique properties and an unexpected carbon dioxide reduction reaction (CO2RR) performance different from those of traditional metal-N4 sites. This review summarizes the recent development of asymmetric atom sites for the CO2RR with emphasis on the coordination structure regulation strategies and their effects on CO2RR performance. Ultimately, several scientific possibilities are proffered with the aim of further expanding and deepening the advancement of asymmetric atom electrocatalysts for the CO2RR.


Introduction
Since the first industrial revolution, the massive consumption of fossil fuels has promoted the innovation of modern science and technology and greatly raised the quality of human lifestyles [1][2][3][4][5]. However, the challenges of energy scarcity and atmospheric pollution are in irreversible progress, and it is imperative to formulate a clean and sustainable energy technology route [6][7][8][9][10]. In particular, one of the possible alternatives is to employ electrocatalytic conversions reactions to manufacture clean energy from renewable or sustainable green sources, e.g., the hydrogen evolution reaction (HER) [11,12], oxygen evolution reaction (OER) [13,14], oxygen reduction reaction (ORR) [15,16], CO 2 reduction reaction (CO 2 RR) [17,18], nitrogen reduction reaction (NRR) [19,20], nitrate reduction reaction (NO 3 RR), etc. [21][22][23]. In these proposed electrochemical reactions, the introduction of catalysts significantly reduces the extra electric energy consumption [24][25][26][27][28]. For traditional bulk and/or nanoparticles (NPs) electrocatalysts, only a few layers of the atoms on the surfaces of the catalysts are involved in the reactions, resulting in a large percentage of the metal atoms being ineffective and wasted. Consequently, next-generation electrocatalysts need to meet the strict demands of higher atomic utilization efficiency compared with the original ones [29][30][31][32].
In 2011, Zhang et al. prepared an atomically dispersed Pt 1 /FeO x , which was the first time the concept of a "single-atom catalyst" (SAC) had ever been applied worldwide [33]. In recent years, due to their almost 100% atomic utilization rates and exceptional activity

Advantages of Asymmetrically Coordinated SACs
Initially, in 2015, Varela et al. reported that the carbon-based catalysts (Fe/Mn-N-C) with doped N and Fe/Mn species demonstrated a higher CO selectivity of up to 80% at −0.5 V vs. RHE, which is superior to polycrystalline Au electrodes [58] and other comparable metal-N-C materials (e.g., the Co protoporphyrin, Ni 2+ on N-doped graphene, Rhporphyrin-like functionalized graphene, etc.) [59][60][61]. Then, the researchers identified that FeN4 moieties were the primary active sites for the selective CO2 reduction into CO with up to 80% FECO in an aqueous solution (Figure 2a,c) [62], whereas Fe nanoparticles (NPs) mostly catalyzed hydrogen evolution [63]. Through using DFT calculations, the CO2RR mechanism of Fe-N4-embedded N-doped graphene was further investigated, finding that the CO2 molecules preferred to adsorb on the Fe catalytic site, while the neighboring N atoms promoted the formation of COOH* and the release of CO*, thus improving the conversion efficiency (Figure 2b) [62].

Advantages of Asymmetrically Coordinated SACs
Initially, in 2015, Varela et al. reported that the carbon-based catalysts (Fe/Mn-N-C) with doped N and Fe/Mn species demonstrated a higher CO selectivity of up to 80% at −0.5 V vs. RHE, which is superior to polycrystalline Au electrodes [58] and other comparable metal-N-C materials (e.g., the Co protoporphyrin, Ni 2+ on N-doped graphene, Rh-porphyrin-like functionalized graphene, etc.) [59][60][61]. Then, the researchers identified that FeN 4 moieties were the primary active sites for the selective CO 2 reduction into CO with up to 80% FE CO in an aqueous solution (Figure 2a,c) [62], whereas Fe nanoparticles (NPs) mostly catalyzed hydrogen evolution [63]. Through using DFT calculations, the CO 2 RR mechanism of Fe-N 4 -embedded N-doped graphene was further investigated, finding that the CO 2 molecules preferred to adsorb on the Fe catalytic site, while the neighboring N atoms promoted the formation of COOH* and the release of CO*, thus improving the conversion efficiency ( Figure 2b) [62].
Furthermore, in a work on a N-doped porous carbon matrix including isolated FeNx moieties with different N coordination numbers, it was revealed that when compared with perfect FeN4 and the Fe-N3 atomic site, the low-coordinated FeN3V active center effectively reduced the free energy barrier for promoting the formation of intermediates *COOH of rate determination step, so that Fe1NC/S1-1000 with Fe-N3V atomic sites demonstrated a high 96% FECO [66]. In addition to controlling the N coordination number, heteroatoms exist in different coordination shell locations or spatial orientations to alter the catalytic activity of the atomic metal center. For instance, in contrast with the traditional Sn-N4 configuration, which mostly produces HCOOH and H2 products, N-rich carbon catalysts made up of atomically dispersed SnN3O1 active sites have CO as the main product ( Figure 2d). According to DFT calculations, SnN3O1′s atomic configuration lowers the activation energy barrier for *COO and *COOH generation (△G = 1.37 eV) while significantly raising the energy barrier for the production of *OCO (△G = 2.48 eV), which (a) Normalized Fe K-edge X-ray adsorption near-edge structure (XANES) spectra of 0.5Fe/NG-750 catalyst. (b) Free energy diagram for electrochemical CO 2 reduction to CO. (c) CO Faradaic efficiency (FE CO ). Reprinted with permission from ref. [62]. Copyright: (2018) WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. (d) The Fourier transform extended X-ray absorption fine structure (FT-EXAFS) fitting curve for Sn-NOC in R space. The calculated Gibbs free energy diagrams for (e) CO 2 -to-CO conversion and for (f) CO 2 -to-HCOOH conversion. Reprinted with permission from ref. [64]. Copyright: (2021) The Authors. Advanced Science published by Wiley-VCH GmbH. (g) Atomic models of Cu-Fe-N 6 , Fe-N 4 , and Cu-N 4 catalytic centers. (h) Free energy profiles for CO 2 electroreduction reactions. (i) FE CO of Fe-N-C, Cu-N-C, and Cu-Fe-N 6 -C. Reprinted with permission from ref. [65]. Copyright: (2020) Wiley-VCH GmbH.
Furthermore, in a work on a N-doped porous carbon matrix including isolated FeN x moieties with different N coordination numbers, it was revealed that when compared with perfect FeN 4 and the Fe-N 3 atomic site, the low-coordinated FeN 3 V active center effectively reduced the free energy barrier for promoting the formation of intermediates *COOH of rate determination step, so that Fe 1 NC/S 1 -1000 with Fe-N 3 V atomic sites demonstrated a high 96% FE CO [66]. In addition to controlling the N coordination number, heteroatoms exist in different coordination shell locations or spatial orientations to alter the catalytic activity of the atomic metal center. For instance, in contrast with the traditional Sn-N 4 configuration, which mostly produces HCOOH and H 2 products, N-rich carbon catalysts made up of atomically dispersed SnN 3 O 1 active sites have CO as the main product ( Figure 2d). According to DFT calculations, SnN 3 O 1 s atomic configuration lowers the activation energy barrier for *COO and *COOH generation ( G = 1.37 eV) while significantly raising the energy barrier for the production of *OCO ( G = 2.48 eV), which encourages the conversion of CO 2 to CO and inhibits the production of HCOOH, opening up a new route for the selectivity of particular products (Figure 2e,f) [64]. Single atoms can also form metal-metal bonds with other metals in order to build binary atom sites with the unique advantage of synergistic effects. The resulting Fe/Cu-N-C catalysts exhibited excellent >95% FE CO with a wide potential range from −0.4 to −1.1 V vs. RHE (Figure 2g,i), which was primarily attributed to the bifunctional active sites, i.e., the Cu site has a greater affinity for molecular CO 2 (CO 2 adsorption energy: Cu-Fe-N 6 -C, −0.03 eV; Fe-N-C, 0.43 eV; Cu-N-C, 0.61 eV), increasing the CO 2 concentration, while the Fe site inhibits hydrogen evolution (Figure 2h) [65,67]. In conclusion, through the modification of the MN 4 coordination environment, including the metal center, the coordination atom numbers, the coordination atom species, and the position or spatial orientation of the coordination shell had a significant influence on the microscopic properties and catalytic activity of the SACs. Based on a more detailed comparative evaluation, owing to the high electronegative N components in the local structure, the symmetric MN 4 moiety may lead to an inappropriate adsorption strength with the intermediate and can slow down the entire reaction process, causing problems such as low product selectivity and a high reaction overpotential. Meanwhile, the active site's electronic and geometric asymmetry can control the adsorption-desorption behavior of the crucial reaction intermediates, which is beneficial in lowering the reaction energy barrier, preventing the occurrence of side reactions, and accelerating the reaction rate, all of which contribute to improved catalytic activity.

The Synthesis Strategies for Breaking MN 4
There are two main routes to breaking the M-N bond of MN 4 . One is the carbonization of precursors such as metal-organic frameworks (MOFs) and polymers with nitrogencontaining specie ligands or other mixtures containing metal salt and nitrogen sources (such as urea, melamine, dicyandiamide, N 2 , NH 3 , etc.), in which the coordination number can be controlled by finely modulating temperature [68]. Another is the pre-construction of Ndoped graphene with abundant defects, vacancies, or edges, and subsequently introducing metal sources that are immobilized in the pre-existing positions of graphene via pyrolytic dispersion to produce low-coordinated metal atoms [69,70]. For instance, the weaker M-O bond is broken in the M-N x O y hybrid coordination structure, followed by the removal of the oxygen atom, producing an atomically scattered M-N x catalyst [71]. In addition, M-N bonds can also be broken by using sophisticated plasma treatments [72,73].
The introduction of coordinated heteroatoms beyond N can be achieved via the following methods: (i) pre-storing in the precursor through coordination, impregnation, encapsulation, grinding, or other physical mixing styles; (ii) using biomass precursors or polymers rich in a variety of non-metallic elements (purslane, chitosan, cyanobacteria, etc.); (iii) for materials with low boiling points, such as S, P, and I, one can diffuse onto the carbon substrate via the sublimation effect during carbonization, coordinating with metals, or replacing carbon atoms; and (iv) using reactants that contain certain heteroatoms, such as S powder, thiourea, and sulfur salts for S, triphenylphosphine, NaH 2 PO 2 , and polypyrrole for P, NH 4 I and I 2 for I, HCl, NaCl, and KCl for Cl, borane and boronic acid for B, and polytetrafluoroethylene for F [74][75][76][77][78][79][80][81][82][83].
The construction of homonuclear/heteronuclear bimetallic atoms is commonly achieved via increasing the concentration or species of the metal salts, usually accompanied by monometallic sites. For the fabrication of homonuclear sites, the increased metal amount means raising the possibility of NPs [84]. Ligand-protected diatomic nodes can be designed using MOF substrates acting as metal precursors [85]. Recently, an anion replacement deposition-precipitation method was reported, in which water-soluble M 2 precursors become insoluble after anion replacement and were thus deposited onto a pre-prepared carrier [86]. For heteronuclear sites, one method is to encapsulate two metal salts into the pore channels of the MOF template [87], and another is to prepare a mixed precursor containing a nitrogen source, carbon source, and double metallic source [88]. Preparing high-quality bimetallic structures using atomic layer deposition (ALD) technologies, where Nanomaterials 2023, 13, 309 6 of 28 the second metal can only be attached to the primary metal by modulating the deposition conditions, has also been demonstrated to be effective [89].

The Characterization of Asymmetric Atom Sites
Currently, there are two main characterization tools for describing single atomic sites: high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and XAFS.
The resolution of HADDF-STEM images can reach up to 1 Å, allowing a fine distinction between individual atoms [90]. For instance, Zhang et al. reported a conception in which graphene defects capture Ni single atoms as active sites, as shown in Figure 3a, and identified the presence of di-vacancy defects that trap an atomic Ni species (aNi@Divacancy) [91]. Furthermore, Liu et al. synthesized an atomic dispersed cobalt-based catalyst with a doped I species that has low-coordination CoN 3 and IN 3 sites (Co-I-N/G). Through atomic-scale HAADF-STEM images ( Figure 3b) and a Z-contrast analysis, two sets of uniformly distributed bright spots with different intensities (marked by green and yellow circles, respectively) were attributed to Co and I atoms, respectively [92]. For dual-atom sites, Zhou et al. presented a bimetallic (Rh-Fe) interbond as an ideal chemical facilitator (FR-NCS) that can facilitate the dispersion of Fe atoms. In Figure 3c, most of the Fe and Rh are uniformly and atomically dispersed on the nitrogen-doped carbon hollow spheres shell (highlighted by blue circles), and Fe-Rh "hetero-pair" atoms are found in Figure 3d, demonstrating the formation of an Rh-Fe interband [93]. In short, it can be noted that the STEM technique does not acquire accurate structural information about the atom species and coordination number.
EXAFS can be used to calibrate the atomic species and obtain the neighboring atom structure, and, thus, is an indispensable tool for investigating asymmetric coordination [94]. Sui et al. synthesized an effective CO 2 RR catalyst featuring Ag atoms coordinated with three nitrogen atoms (Ag 1 -N 3 , Figure 3e). In the EXAFS spectra, a significant peak in the Ag-SACs' EXAFS spectra at 1.5 Å is attributed to Ag-N scattering, but there is no Ag-Ag scattering at 2.6 Å, indicating an atomically dispersed Ag species (Figure 3f). In the EXAFS fitting curves (Figure 3g,h), the Ag-N coordination numbers of Ag 1 -N 3 /PCNC and Ag 1 -N 2 /PCNC are fitted from 3.1 to 1.9, respectively, demonstrating the formation of Ag 1 -N 3 and Ag 1 -N 2 coordination structures. The electrochemical test results showed the trend of the CO Faradaic efficiency (FE CO ) sequence, Ag 1 -N 3 /PCNC > Ag 1 -N 2 /PCNC > Ag NPs/C > CFP, confirming that Ag 1 -N 3 was the best active site among them ( Figure 3i) [95].

Low-Coordination Structure
Eliminating part of the coordinated N atoms from the original symmetric MN4 results in the corresponding defective sites, i.e., the low-coordinated MNx configuration. The low-coordination sites of MNx (M = Fe, Co, Ni, Cu, Zn, Mn, Ag, etc.; x < 4) are mainly constructed in two strategies: one is a one-step method to obtain the desired coordination numbers, and the other is breaking the original metal-coordination bond.

Low-Coordination Structure
Eliminating part of the coordinated N atoms from the original symmetric MN 4 results in the corresponding defective sites, i.e., the low-coordinated MNx configuration. The low-coordination sites of MN x (M = Fe, Co, Ni, Cu, Zn, Mn, Ag, etc.; x < 4) are mainly constructed in two strategies: one is a one-step method to obtain the desired coordination numbers, and the other is breaking the original metal-coordination bond.
In abundant atomically dispersed based metal-nitrogen co-decorated carbon (M-NC) materials, some transition metals, i.e., Fe, Co, Ni, and Cu, can facilitate charge transfer to account for their unfilled 3d orbitals. Wang et al. reported the gas diffusion of ferrocene and thermal activation (800-1000 • C) to successfully fabricate isolated Fe atoms loaded on Ndoped porous carbon polyhedrons ( Figure 4a) with different N coordination numbers (FeN 4 , FeN 3 ). According to DFT calculations and experimental measurements, the Fe-N 3 V active centers have most balanced free energy barriers of intermediates, CO 2 to *COOH (0.47 eV), *COOH to *CO (0.15 eV), and *CO to CO (0.14 eV), strengthening its CO 2 RR performance. At the same applied potential of −0.5 V vs. RHE, the FE CO of Fe 1 NC/S 1 -1000, Fe 1 NC/S 1 -900, and Fe 1 NC/S 1 -800 are 96%, 90%, and 82%, respectively (Figure 4b,c) [66]. Similarly, three atomically dispersed Co catalysts (800 • C, Co-N 4 ; 900 • C, Co-N 3 ; and 1000 • C, Co-N 2 ) can be selectively synthesized by adopting various pyrolysis temperatures of bimetallic Co/Zn ZIFs. The Co-N 2 could acquire a better selectivity and activity compared with other Co-N x , which demonstrated that the inert reactive sites could also be activated by adjusting the coordination environment (Figure 4d,e) [96]. Rong et al. adopted the other strategy that differs from those mentioned above to construct a vacancy-defect Ni site. The single-atom Ni-N 3 O mixed-coordination precursors were first synthesized at 500 • C and then heated to 800 • C, where the oxygen atoms could be subsequently eliminated on account of the weaker Ni-O interactions, leading to atomically dispersed Ni-N 3 -V catalysts ( Figure 4f). It was discovered that the electrocatalytic activity of the CO 2 RR can be significantly enhanced after the introduction of vacancy defects in Ni-N 3 -V SACs, compared with the difficulty of the NiN 3 site to release CO ( G = 1.264 eV), and this is mainly attributed to Ni-N 3 -V's optimized and moderate CO 2 to COOH* reaction free energy ( G = 0.680 eV) (Figure 4g) [71]. Calculations further disclose that the active site is the particular defect-Ni-N 3 structure [97]. Moreover, the Ni-pyridinic N 2 V 2 is quickly generated as a result of the high-energy plasma's constant collision and presents a more advantageous comprehensive performance than other Ni-pyridinic N x C 3−x species, resulting in a higher CO 2 RR efficiency and worse HER activity (Figure 4h,i) [72]. Lin et al. found that the FE CO of both the NiN 2 site and FeN 4 site is proportional to the reaction temperature, but that the NiN 2 site is more effective, which stems from the variation in the adsorption strength of the key intermediates between the two [98]. Cu-based nanocatalysts are frequently utilized for the construction of a wide range of C 1+ and C 2+ products [99]. The work of Guan et al. showed the significance of the distance between isolated atomic copper sites for the electroreduction of CO 2 to various hydrocarbons ( Figure 4j). DFT calculations indicate that high 4.9% mol Cu concentrations with adjacent Cu-N 2 sites are more advantageous for the formation of C 2 H 4 due to the reduced free energy of the C-C coupling, while all isolated Cu-N 4 , neighboring Cu-N 4 , and isolated Cu-N 2 sites tend to generate C 1 products when the concentration of Cu is less than 2.4% mol [100].
To date, Zn-NC has received less attention compared with other transition metals (Fe, Co, Ni, and Cu), and it is generally believed that when the 3d 10 orbital is fully occupied, it can significantly impair electron transport and lower the electrocatalytic activity of Zn. Li et al. reported a N-anchored low-valence Zn single-atom catalyst that contains saturated four-coordinate (ZnN 4 ) and unsaturated (Zn-N 3 ) active sites. The unsaturated Zn-N 3 could dramatically reduce the energy barrier by stabilizing the COOH* intermediate, thus allowing simultaneous large current densities of up to 1 A cm −2 and high CO selectivity of up to 95% using flow cell devices (Figure 4k) [68]. Compared with the poor selectivity and activity (65% FE CO , 3.3 mA cm −2 of j CO ) of Mn-N 4 [58], the Mn-N 3 center in Mn-C 3 N 4 /CNT can attain an excellent CO 2 RR performance (98.8% FE CO , 14.0 mA cm −2 of j CO ). In situ X-ray absorption spectra and DFT calculations reveal that the Mn-N 3 site can lower the free energy barrier for the production of the crucial intermediate COOH*, which is responsible for high catalytic performance (Figure 4l) [101]. Additionally, altering the coordinate environment of the Ag atoms can boost their activity. It was shown that the Ag 1 -N 3 site performed better than Ag 1 -N 2 and Ag NPs, owing to its lower CO binding energy [95]. To date, Zn-NC has received less attention compared with other transition metals (Fe, Co, Ni, and Cu), and it is generally believed that when the 3d 10 orbital is fully occupied, it can significantly impair electron transport and lower the electrocatalytic activity of Zn. Li et al. reported a N-anchored low-valence Zn single-atom catalyst that contains saturated four-coordinate (ZnN4) and unsaturated (Zn-N3) active sites. The unsaturated Zn-N3 could dramatically reduce the energy barrier by stabilizing the COOH* The investigation of the above works reveals that the reduced oxidation state of the monoatomic metal center with low coordination numbers makes it easier to adsorb reactant molecules, and optimize the adsorption strength of intermediates, especially COOH* in the CO product process. In addition, we also speculate that the extra space caused by low coordination also facilitates a more flexible matching between reactant molecules and active sites.

Lateral Heteroatom Coordination Structure
The doping of non-metallic atoms in carbon supports is an efficient and flexible technique for improving the overall performance of SACs, such as through the introduction of heteroatoms to replace the coordinated N atoms in the symmetrical planar MN 4 sites to form a lateral heteroatom coordination structure. The laterally coordinated non-metallic heteroatoms, such as C, N, S, O, B, P, F, etc., could modulate the coordination environment of the atomic metal active center through direct coordination (first coordination shell) and indirect coordination (second or higher coordination shell), thereby achieving the goal of optimizing the reaction path.
In  [103]. Yang et al. also observed that the N coordination number dropped from 4 to 1, while the C coordination number grew throughout the construction process of the Ni@N x C y catalysts when the temperature was raised from 800 to 1100 • C [104]. Of note is the charge capacity of the site for facilitating the electrochemical steps as well as hydrogen bonding to the intermediate for stabilizing the intermediates, which were often ignored in previous DFT calculations and are indispensable [105]. To lower the production investment of single-atom catalysts, Zu et al. synthesized kilogram-scale Sn δ+ on N-doped graphene (SnN 2 C 2 ) using a rapid freeze-vacuum drying-calcination procedure, and its electroreduction activity basically remained inactive even after 200 h (Figure 5d) [106]. Sulfur and oxygen atoms have been viewed as ideal alternatives for coordinated N in the metal-Nx center because of their lower electronegativity than N, having the ability to dramatically change the active center's local electronic densities [107]. Wang et al. developed an atomically dispersed N, S co-coordinated Bi site catalyst (Bi-N 3 S/C) shown in Figure 5e, which achieved up to 98.3% FE CO at −0.8 V vs. RHE, which is significantly better than Bi-SAs-N/C, and this demonstrated that the introduction of S can greatly lower the energy barrier of the intermediates of the rate-determining step [108]. Sometimes, other strategies are employed in cooperation with heteroatomic coordination to increase a catalyst's activity. In a S/N dual-heteroatom anchored unsaturated Ni site (NiN 2 S-V), the decreased energy barriers for the electroreduction of CO 2 to CO are a result of both doped S atoms and evolved S vacancies [109]. As shown in Figure 5f, this synergistic interaction was also verified by Chen et al. via developing a tandem catalyst consisting of a single Cu site with N and S co-coordinated and atomically dispersed Cu clusters (Cu-S 1 N 3 /Cu x ) [110]. Guo et al. demonstrated the difference between prepared SnN 3 O 1 (CO product) and a classic SnN 4 (HCOOH product) configuration for CO 2 -to-CO conversion [64]. Meanwhile, Sn-C 2 O 2 F catalysts with distinctive non-nitrogen coordination structures also support the opinion just mentioned according to the selectivity of CO 2 RR products [83]. In addition, the smaller radii of the electron-deficient B make it vulnerable to coordinate with the electron-rich N via electron-sharing effects, such as with the B-doping in B/N co-coordinated atomic Fe-based catalyst (Fe-N 3 B) that also facilitated *COOH formation and inhibited hydrogen generation [82]. Indirectly coordinated heteroatoms (secondary or higher coordination shells) can also induce local electric fields and change the catalytic behavior of the catalytic center. Chen Figure 5h, both inducing an enhanced CO 2 RR performance relative to pristine FeN 4 as well as NiN 4 , respectively [111,112]. P, in the N group, as a classic electron donor, through integrating into carbon substrates, can alter the electronic structure of the metal site via its 3p lone pair electrons. In Sun et al.'s work on Fe-SAC/NPC (FeN 4 O 1 -P), by fitting FT-EXAFS spectra in R space, each Fe atom was separated by four Ns and one O, while a single P in high-coordination shells (n ≥ 3) was merged mainly as P-C bonds into the N-doped carbon (Figure 5i), and the P of the third coordination shell enhanced the stability of the key *COOH intermediate on Fe, resulting in an outstanding CO 2 RR performance at a low applied potential [113].
also support the opinion just mentioned according to the selectivity of CO2RR products [83]. In addition, the smaller radii of the electron-deficient B make it vulnerable to coordinate with the electron-rich N via electron-sharing effects, such as with the B-doping in B/N co-coordinated atomic Fe-based catalyst (Fe-N3B) that also facilitated *COOH formation and inhibited hydrogen generation [82]. Indirectly coordinated heteroatoms (secondary or higher coordination shells) can also induce local electric fields and change the catalytic behavior of the catalytic center. Chen et al. reported S doping in the second coordination shell of FeN4 (FeN4S), shown in Figure 5g, and Gu et al. finished the B bridging atomic coordination of NiN4 (NiN4B2) shown in Figure 5h, both inducing an enhanced CO2RR performance relative to pristine FeN4 as well as NiN4, respectively [111,112]. P, in the N group, as a classic electron donor, through integrating into carbon substrates, can alter the electronic structure of the metal site via its 3p lone pair electrons. In Sun et al.'s work on Fe-SAC/NPC (FeN4O1-P), by fitting FT-EXAFS spectra in R space, each Fe atom was separated by four Ns and one O, while a single P in high-coordination shells (n ≥ 3) was merged mainly as P-C bonds into the N-doped carbon (Figure 5i), and the P of the third coordination shell enhanced the stability of the key *COOH intermediate on Fe, resulting in an outstanding CO2RR performance at a low applied potential [113]. In short, heteroatom coordination has a greater influence on CO 2 RR conversion compared with lower coordination strategies and can even completely change the obtained product, e.g., SnN 4 , a classical formic-acid-producing catalyst, transforms into CO after the introduction of a coordinated O or F. Further enhancement of the understanding of the mechanism of heteroatom action can expand the choice of products.

Axial Heteroatom Coordination Structure
Heteroatoms in the conductive carbon supports can function as bridges to axially couple to MN 4 sites, while isolated heteroatoms can also build axial coordination patterns in the form of unsaturated coordination (dangling atoms). The various heteroatoms can directly coordinate with the single atomic M center in the symmetric MN 4 to construct an asymmetric axial coordination structure of MN 4 X 1 (X = N, O, S, Cl, F), which is essentially identical to the function of the lateral structural configuration, i.e., to obtain a more suitable electron structure and performance than pristine MN 4 .
Zhang et al. designed a graphene-confined atomic dispersed FeN 5 catalytic site in Figure 6a-c, in which the electron density of the Fe 3d orbitals is further consumed through axial pyrrolic N, thus weakening the Fe-CO π back donation, resulting in a fast desorption toward CO and high selectivity (only at −0.46V vs. RHE, up to~5 mA cm −2 , FE CO ca. 97.0%) [114]. After the substitution of axial N by O, the corresponding catalytic activity of the obtained Fe 1 N 4 O 1 was further strengthened (−0.56 V to −0.87 V vs. RHE, up to~15 mA cm −2 , FE CO nearly 100%), as shown in Figure 6d-f [115]. Simultaneously, similar satisfactory results were also achieved for NiN 4 O (−0.5 V to −1.1V vs. RHE, up to~30 mA cm −2 , FE CO above 90%), which could be responsible for the axial traction effect with an additional axial oxygen atom in Figure 6g-i [116]. Additionally, the one-pot template sacrificial pyrolysis method used by Huang et al. promises the development of a series of metal-N 4 -O catalysts, such as well-defined Ni-N 4 -O, for a variety of catalytic applications [117]. Aiming to attain the targets of high selectivity, high current density, and long-term stability, Wu et al. designed and prepared a CdN 4 S 1 /CN catalyst that is effective for the electrochemical reduction of CO 2 to CO (−2.1 V to −2.7 V vs. Ag/Ag + , FE CO more than 95%) in H-type cells with [Bmim]PF 6 -MeCN cathodic electrolytes and also reaches an outstanding current density of 182.2 mA cm −2 , which is majorly owing to the reason that Cd can inhibit HER as well as S with a high spin density and charge delocalization can lower the key free energy barrier. Thus, the CdN 4 S 1 site has a lower free energy barrier for the CO 2 RR (0.27 eV) and higher free energy barriers for the formation of H* for the HER (0.49 eV) (Figure 6j-l) [118]. Furthermore, Huang et al. designed a new Ni-N 5 -C single-atom catalyst with an enzyme-like catalytic active site and then evaluated its performance using a flow cell device, achieving a maximum current density of 1.23 A cm −2 at −2.4 V vs. RHE with a 99.6% FE CO and 100 h continuous stable operation [119]. In addition, in some reported works, Cl and F are also used as axial coordination atoms of MN 4 to improve the electroreduction properties of CO 2 , such as FeN 4 Cl/NC and MnN 4 Cl, and, although both of them can achieve a high selectivity of >90%, they are limited to a narrow lower potential window [56,78]. In another specialized configuration catalyst of SnC 2 O 2 F, the F atom bonded to Sn has a significant capability to inhibit hydrogen evolution [83].
By summarizing the above-mentioned works, it can be found that the activity of the CO 2 RR is gradually enhanced by modifying the axial coordination atoms or optimizing the choice of metal centers, and the axial coordination atoms seem to be more advantageous in inhibiting hydrogen precipitation. Importantly, the choice of electrolyte, such as ionic liquids, also has a significant effect on the improvement of CO 2 RR yield.

Dual-Metal Coordination Structure
Bimetallic atomic sites, i.e., neighboring monoatomic sites, provide more opportunities for the manipulation of the electrical and geometric structures of SACs on the basis of preserving atomic dispersion. Differing from SACs, bimetallic heteroatomic site catalysts, mainly including M 1 N 4 -M 2 N 4 , M 1 M 1 N x , M 1 M 2 N x , etc., utilize two adjacent metal atoms to achieve a functional association and synergistic effect and, furthermore, through the elec-tronic interactions between them, to efficiently tailor the binding strength of intermediates as well as optimize CO 2 RR activity. electroreduction properties of CO2, such as FeN4Cl/NC and MnN4Cl, and, although both of them can achieve a high selectivity of >90%, they are limited to a narrow lower potential window [56,78]. In another specialized configuration catalyst of SnC2O2F, the F atom bonded to Sn has a significant capability to inhibit hydrogen evolution [83].  (Figure 7a). The neighboring Ni redistributes the electrons of Sn and results in the reduced free energy barrier of *OCHO and subsequent thermodynamically spontaneous determining step, and by taking double advantage of the synergistic effect of the NiSn diatomic site and the increased utilization, exhibits excellent selectivity for the generation of formate (86.1% at −0.82 V RHE ) compared with Sn-SAC (70.4% at −0.82 V RHE ) as well as Ni-SAC (<1 %), as shown in Figure 7b [120]. Hao et al. designed a non-bridged Ni dual-atom catalyst (Ni 2 N 6 ) with a local structure of N 3 -Ni-Ni-N 3 for CO 2 to CO (Figure 7c), and in situ environmental STEM captured the Ostwald ripening process of the Ni species generating the bulk phase and the subsequent atomization dynamics process of the transformation into Ni diatomic sites (Figure 7d). The actual active sites for the fast CO 2 RR kinetics and the slower HER kinetics are Ni diatomic sites (Ni 2 N 6 OH) induced via hydroxyl adsorption, according to in situ XAFS studies and theory calculations (Figure 7e). [84] In addition, there is a similar active site on the O-bridge adsorption structure (O-Ni 2 N 6 ) of the 2N-bridged Ni 2 N 6 site for the dynamic catalysis of CO 2 [85]. Moreover, several other homoatomic binuclear site catalysts with the special local environment and a non-bridged Ni 2 N 4 C 2 and Pd 2 dual-atom site with a PdN 2 O 2 coordination structure have shown superior, nearly 100% CO selectivity [86,121].   (Figure 7f). The optimal CO 2 RR activity and selectivity were identified to be at the 2N-bridged (Fe-Ni)N 6 sites, which can account for these two metal sites' synergistic interaction that could simultaneously promote the adsorption of *COOH and the desorption of *CO while inhibiting the HER [122]. More in-depth insights can be provided through an electronic structure analysis: in the NiFe heteroatom sites, Fe acts as the catalytic center with a higher oxidation state resulting from its orbital coupling with Ni, which facilitates CO 2 activation and does not change the binding strength to *COOH but the weakened binding strength with *CO intermediates, and accelerates CO desorption, thus increasing the catalytic CO 2 RR activity [125]. Jiao et al. concluded that in non-bonding Fe 1 -Ni 1 -N-C, the Fe atomic site can also be highly activated by the adjacent single-atom Ni through non-bonding interactions between them to act as the active center [126]. Meanwhile, Ren et al. believe that the key to the CO 2 RR is the collaborative coordination of CO at non-bridged diatomic NiFeN 6 sites fixed on nitrogenated carbon [127].

Xie et al. synthesized a neighboring bi-metal atom catalyst (NiN 4 -SnN 4 ) with each Ni and S coordinated with four Ns
Furthermore, two types of (Cu-Ni)N 6 for non-bridged Ni/Cu-N-C ( Figure 7g) and 2N-bridged CuNi-DSA/CNF (inset of Figure 7h) have been reported: for non-bridged Ni/Cu-N-C, the introduction of the secondary Cu metal accelerates the rate-determining step of *COOH formation by transferring the Ni 3d orbital energy to the Fermi energy level; for 2N-bridged CuNi-DSA/CNF, electronegativity shifts between Cu and Ni, causing strong electronic interactions and the offset effect, which can efficiently optimize the adsorption strength of *COOH intermediates. The activation energy of *COOH creation on CuN 4 -NiN 4 is 0.08 eV, as shown in Figure 7h, which is much lower than that on CuN 4 and NiN 4 (0.58 eV and 0.48 eV), proving that the CuNi dual-atom sites do actually facilitate the kinetic process of *COOH production [123,124,128]. In addition, co-constructed M 1 M 2 N x composed of other metals, e.g., the ZIF-8 containing Fe/Cu sources derived a Cu-Fe diatomic site catalyst that coordinated with the N and was doped in the carbon matrix (non-bridged (Cu-Fe)N 6 ) [65,67]; direct pyrolysis of a combination of carbon black, metal salt, and a urea nitrogen source yielded a Co-Cu diatomic site catalyst (2N-bridged (Co-Cu)N 6 ) [129]; a mixture of chitosan and Ni/Zn metal salts derived atomically dispersed Ni-Zn bimetal sites coordinated on N-doped carbon (non-bridged (Ni-Zn)N 6 ) [130]; and a precursor containing Zn/Co nitrate, urea, and activated carbon black derived diatomic Zn-Co monomers supported on nitrogen-doped carbon (2N-bridged (Zn-Co)N 6 ) [131], have also shown better CO 2 electroreduction performances.
The analysis of the above different strategies for improving the CO 2 RR reveals that each strategy has its own unique advantages, and the combined use of these strategies may achieve unexpected results. It is important to note that the final evaluation index of almost all the strategies focuses on the pair energy barrier and adsorption energy changes due to electronic interactions but ignores factors such as hydrogen bonding, solvation effects, and spatial structure that are often simplified or ignored in theoretical analyses. In addition, mesoscopic factors, such as the modulation of the solid-liquid-gas interface, should also be taken into account.

Asymmetric Atom Sites of Organic/Metal-Based Supports for CO 2 RR
Except for carbon-based ones, organic/metal-based supports are also capable of loading the above-mentioned asymmetric coordination atom sites. Organic crystalline materials, including MOFs, covalent organic frameworks (COFs), metal-organic complexes, etc., have explicit structures and tuned metal sites while retaining a major problem of poor electrical conductivity that cannot be ignored.
To overcome the conductivity of a crystallite, one strategy is to enhance conductivity with itself, e.g., conductive MOFs with conjugated organic ligands, and the other is to integrate it with other better conductive materials. Zhang et al. provided coordinationenvironment-dependent single-site-Cu-O 4 conductive MOFs (Cu-DBC), and, by comparing the Cu-N 4 sites of two other COFs (Cu-TTCOF and Cu-PPCOF), it can be found that the Cu-O 4 site with lower free energy barrier has better CO 2 to CH 4 activity (Faradaic efficiency of CH 4 up to~80% at −0.9 V vs. RHE) than the CuN 4 sites (Figure 8a) [57]. Another conductive COF (PcCu-TFPN) with a high electron density and isolated phthalocyanine CuN 4 sites is more appropriate for the critical step of the C-C coupling of *CH 3 intermediates with CO 2 to generate acetates (Faradaic efficiency of 90.3(2)% at −0.8 V vs. RHE), as shown in Figure 8b [132]. To more precisely elucidate the essential contribution of C-C coupling to the CO 2 deep reduction of C 2+ products, Shao et al. designed two MOFs with various nodes, and, compared with BIF-104 with isolated Cu sites as a reference, BIF-102 for the Clbridging of a dimer Cu unit can provide a higher C 2 H 4 Faraday efficiency as high as~11.3%, while the adjacent Cu atom acts as a modulator, changing the reaction barrier, promoting C-C coupling, and providing a reaction pathway for generating C 2+ products that differ from isolated monomers, which shows that there exists a close correlation between the active-centered coordination structure and CO 2 RR selectivity (Figure 8c) [133]. Zhu et al. hold a comparable view that in robust π-π stacking MOFs (CuBtz) with pyrazolate-bridged dicopper(I) sites, the uncoordinated N atoms act as proton relays and co-catalyze with neighboring Cu sites to promote *CO hydrogenation and the C-C coupling for the highly selective electroreduction of CO 2 to C 2+ products (Figure 8d) [134]. Instead, Lin et al. anchored cobalt phthalocyanine with CoN 4 sites (CoPc) in the FeN 4 sites of graphene (Fe-N-C) for synergistic catalytic CO 2 electroreduction, and the received CoPc@Fe-N-C showed a broader potential window and larger CO current density beyond Fe-N-C [135]. Pan et al. used pre-prepared hollow carbon spheres as supports to successfully construct CoN 5 asymmetric sites through the coordination of Co in CoPC and N-doped carbon spheres (Figure 8e), obtaining a high selectivity of FE CO of over 90% for the CO 2 RR across a broad potential range, from −0.57 V to −0.88 V vs. RHE (Figure 8f) [136]. A similar strategy was successfully employed by Wang et al. to immobilize planar Co II -2,3-naphthalocyanine (NapCo) on graphitic sulfoxide (SO) or carboxyl (COO)-doped graphene (NapCo@graphitic SO; NapCo@COO), i.e., where axial coordination between CoN 4 in NapCo and O from SO or COO occurred (Figure 8g). As a result, there was good electronic communication between NapCo and graphene in NapCo@graphitic SO compared with NapCo@COO, resulting in a high Faraday efficiency of 97% for CO production, yet no activity for CO 2 electroreduction was found for the CoN 4 site of pristine NapCo without graphene [137]. In metal-based supports, Jiao et al. reported Cu 1 0 -Cu 1 x+ pairs stabilized on the Te surface imperfections of Pd 10 Te 3 alloy nanowires, and, with Cu 1 x+ and Cu 1 0 separately corresponding to the adsorption of molecular H 2 O and CO 2 , with an almost completely suppressed HER, this named bi-atomic activating bimolecular effect facilitates the evolution of CO 2 to CO (Figure 8h) [138].
Unlike carbon-based supports, there are still relatively few hetero-coordination atoms loaded on organic/metal-based supports, and the key is to fundamentally address organic substrates' conductivity, stability, and lessened selectivity, as well as metallic carriers' undesirable competitive HER, but satisfactory CO 2 RR data have still been found from the limited reports, which inspires the majority of researchers to continue exploring this vast uncharted territory. Unlike carbon-based supports, there are still relatively few hetero-coordination atoms loaded on organic/metal-based supports, and the key is to fundamentally address organic substrates' conductivity, stability, and lessened selectivity, as well as metallic carriers' undesirable competitive HER, but satisfactory CO2RR data have still been found from the limited reports, which inspires the majority of researchers to continue exploring this vast uncharted territory.

Summary and Outlook
For over a decade, significant advancements have been accomplished in the structural design and fundamental electrocatalytic investigation of SACs as next-generation catalysts with a maximum of nearly 100% atomic utilization [139][140][141]. However, for the CO 2 RR, the simplicity of the conventional MN 4 structure severely restricts the substantial improvement of its electrocatalytic activity. It is commonly accepted that the SACs' coordination structures determine the electronic structures and thus affect the intrinsic electrocatalytic activity.
It follows that the local/expanded coordination environment of active single atoms in SACs has a combined effect on the stability, catalytic performance, and utilization of the loaded materials [142,143]. As shown in Table 1, asymmetric atom sites, which have recently gained popularity among a wide range of researchers, can provide flexible structures with customizable atomic configuration, local/expanded coordination, and electronic architecture without damaging this principle of high atomic utilization, thereby breaking such innate limitations of MN 4 . Therefore, efforts have been made in recent years to develop more advanced asymmetric atom electrocatalysts for the CO 2 RR compared with SACs.  [123] Furthermore, the attention on the CO 2 RR has extended to the catalytic environment, as the catalytic activity can be modulated by altering the structural parameters of the catalyst (solid)-electrolyte (liquid)-molecular reactant (gas) three-phase interface [144][145][146]. Some advances have been realized in these interfacial modulation studies, wherein it was discovered that controlling the proton supply or CO 2 supply can effectively inhibit the HER [147,148], electrodes with improved interfacial structures significantly improve selectivity and partial current density at relatively low overpotentials in H-type electrolytic cells [149,150], and membrane reactors and microfluidic reactors (flow cells) can achieve higher current densities [151,152]. Modification of the catalyst-electrolyte interface region also includes the addition of external molecules (or polymers), changes in electrolyte composition and concentration, the design of novel reactor and electrode structures, and a combination of these approaches, and these modification strategies can significantly affect the intrinsic and extrinsic catalytic activity of the catalyst. Furthermore, the design of syngas (H 2 and CO) catalysts can also give us some guidance, especially by regulating the relative content of different active center species, to achieve the selectivity of H 2 and CO, which can be used to determine the differences in various activities of metal sites for the CO 2 RR and HER and to obtain general design principles [153,154]. Moreover, some predictions about CO 2 RR electrocatalysts from theoretical calculations also provide guidance for the development of new catalysts, which have previously received less attention, for example, Rh@Au(100) and Rh@Ag(100), which may produce CH 4 [155], and the Cu 2 N 4 -loaded C 2 N layer, which can form CH 4 and C 2 H 4 [156], are expected to perform well.
Recent studies on asymmetric atom sites in the electrocatalysis of the CO 2 RR were reviewed in this paper, with importance placed on the comparison of asymmetric atom sites and traditional SACs for the CO 2 RR, including an examination of the crucial role that coordination structures play in the intrinsic electrocatalytic activity of supports-loaded SACs, including local coordination (metal atom center and first coordination shell) and extended coordination (second and higher coordination shell), as well as the resulting alternation in the adsorption parameters, the decrease in the energy barrier of rate-determining step, and the optimization of the reaction pathway. It can be straightforwardly found that asymmetric atom sites have significant advantages in improving the electrocatalytic properties when performing the CO 2 RR over symmetrical atomic catalysts, including catalytic activity, Faraday efficiency, selectivity, stability, etc. It is worth noting that, for the CO 2 RR, this is a universal rule, but some prominent properties of the symmetric coordination should not be neglected, and thus it is concluded that asymmetric atom catalysts are expected to break the theoretical boundary of SACs [157,158]. Furthermore, we proposed the current urgent challenges to be solved and future expected directions in the field of asymmetric atom sites.
The accurate management of single atoms' desired coordination environments on the supports remains a major challenge. After the coordination environment is optimized, the corresponding active monoatom is securely trapped in the surface anchor site via the charge transfer effect, interaction absorption, etc., which could successfully avoid the metal atom movement and agglomeration caused by high surface energy during the construction procedure and reaction system, impairing its catalytic efficacy [159][160][161]. Currently, SACs with outstanding performance have been successfully established using some reported synthetic approaches, including wet chemistry, ALD deposition, and high-temperature pyrolysis. However, metal nanoclusters or even NPs are occasionally doped into the resultant SACs. The conventional wet impregnation methods are easy to implement, low-cost, and have a scalable production, but obtaining highly loaded and precisely regulated single-atom coordinated sites is challenging [162][163][164]. The popular high-temperature pyrolysis method generally involves the high-temperature calcination of a pre-designed precursor, but it often has unpredictable and ill-defined multiple coordination structures simultaneously due to the influence of the pyrolysis temperature, heating rate, inert atmosphere, and other factors [165][166][167]. The ALD method, which has the advantage of high-precision control procedures, also has the non-negligible disadvantages of complex synthesis steps, a handful of substrate selection, high investment, and low productivity [168][169][170]. Not negligibly, a suitable choice of supports species will influence the vacancies and defects on the surface to be better utilized for stabilizing isolated metal atoms. Carbon-based materials have become the main choice of carriers for SACs due to the easy introduction of O, S, and N elements, but a major drawback is the instability of carbon-based materials at high temperatures and high oxygen concentrations [171][172][173][174]. Correspondingly, the field of supports should be expanded, such as organic crystallization, metals, and derived oxides and nitrides, as well as the selection of the best preparation techniques. Practically speaking, further breakthroughs in the field of SACs lie in the updating of low-cost, easily scalable, and environmentally friendly preparation methods [175][176][177][178][179][180].
Accurate identification of the coordination environment is essential for the development of SACs. STEM techniques allow direct viewing of individual metal atoms to obtain relatively well-defined coordination structures, but the low Z-contrast of the first/second coordination shell (e.g., C, N, S, etc.) causes considerable trouble in mapping the spatial positions of the near-end light atoms. In addition, XAFS characterization can directly distinguish between metal NPs and atomically dispersed atoms, but the acquired coordination structure is generally an averaged result [178][179][180][181][182][183][184][185]. Currently, a combination of characterization techniques, such as electron energy loss spectroscopy (EDS), X-ray photoelectron spectroscopy (XPS), scanning tunneling microscopy (STM), non-contact atomic force microscopy (AFM), and DFT theoretical modeling, is required to obtain a relatively reliable local structure determination [181][182][183][184][185][186][187][188][189]. With the emergence of various in situ/operando characterization techniques in recent years, it has become feasible to directly monitor catalytic reaction processes, which would be extremely potent if applied in conjunction with advanced theoretical modeling. For example, the conversion of dispersed metal atoms into clusters or NPs during electrolysis has been confirmed [184]. A comprehensive insight into the structure-property connection relies on the expression of isolated single atoms rather than statistically averaging data of all the single atoms in diverse bonding structure distributions. The combination of electron microscopy techniques, spectral analysis, and DFT calculations can, firstly, reflect the nature of a single-atom configuration, then, recognize the dynamic evolution of the active components during the formation of products, and, finally, make the reaction pathway and mechanism gradually clear, which is of major interest for catalyst design [190][191][192][193][194][195].
Large-scale industrial applications require the batch construction of SACs with high activity and selectivity using simple and scalable methods, whereby the ultimate goal is to switch out the current noble metal-based electrocatalysts with more accessible transition metals, which is advantageous for the atomic economy and the advancement of green chemistry [188][189][190][191][192][193][194][195][196][197][198]. A practical screening tool is machine learning, which enables the matching and integration of different elements or groups, model construction, and the specific performance prediction of catalysts, and then obtaining optimal results using highthroughput screening, these methods greatly reduce the investment of time and human, material, and financial resources [199][200][201][202][203]. In addition, the factor of long-term thermodynamic stability also needs to be considered [204,205]. Too strong or too weak interactions in the metal-species-anchored supports systems during catalysis can result in negative Ostwald maturation and the migration and aggregation of active sites, respectively [206,207]. In particular, the high temperatures and reducing environments usually faced in industrial applications make catalysts more vulnerable to poisoning or deactivation [199,207]. Overall, the design of high-quality asymmetric atom electrocatalysts via reasonable coordination modulation remains promising.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.

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