Electrochemical Reduction of CO2 to CO over Transition Metal/N‐Doped Carbon Catalysts: The Active Sites and Reaction Mechanism

Abstract

Electrochemical CO₂ reduction to value‐added chemicals/fuels provides a promising way to mitigate CO₂ emission and alleviate energy shortage. CO₂‐to‐CO conversion involves only two‐electron/proton transfer and thus is kinetically fast. Among the various developed CO₂‐to‐CO reduction electrocatalysts, transition metal/N‐doped carbon (M‐N‐C) catalysts are attractive due to their low cost and high activity. In this work, recent progress on the development of M‐N‐C catalysts for electrochemical CO₂‐to‐CO conversion is reviewed in detail. The regulation of the active sites in M‐N‐C catalysts and their related adjustable electrocatalytic CO₂ reduction performance is discussed. A visual performance comparison of M‐N‐C catalysts for CO₂ reduction reaction (CO₂RR) reported over the recent years is given, which suggests that Ni and Fe‐N‐C catalysts are the most promising candidates for large‐scale reduction of CO₂ to produce CO. Finally, outlooks and challenges are proposed for future research of CO₂‐to‐CO conversion.

Introduction

Excessive utilization of fossil fuels has caused a severe energy crisis and a sharp rise in CO₂ concentration in the atmosphere, which leads to global concern on the greenhouse effect.^[1] The Paris Agreement strives to achieve a “balance between anthropogenic emissions by sources and removals by sinks” in the second half of this century, which means that CO₂ emitted to the atmosphere by human activity has to be eliminated either through nature-based solutions (such as afforestation) or technological solutions that can store, capture and/or convert CO₂. Especially, it is attractive to develop an effective approach to reduce CO₂ into value-added products, thereby this not only mitigates CO₂ emission but also produces renewable chemicals/fuels to alleviate energy shortage.^[2] Over the past few decades, CO₂ conversion technologies including thermocatalytic,^[3] photocatalytic,^[4] and electrocatalytic have been extensively studied. Electrocatalytic CO₂ reduction to chemicals and fuels driven by renewable energy (e.g., wind, solar and geothermal energy) provides a promising approach for the carbon cycle.^[5,6] It is also considered as an ideal strategy for storing renewable electricity in chemical bonds in addition to hydrogen production by water electrolysis. However, electrocatalytic reduction of CO₂ still stays in its infancy as compared to water electrolysis technology.

Electrochemical CO₂ reduction reaction (CO₂RR) involves multiple electron/proton transfer processes, and CO₂ can be reduced into various gaseous and liquid products including formic acid (HCOOH), carbon monoxide (CO), hydrocarbons (CH₄ and C₂H₄), and alcohols (CH₃OH and C₂H₅OH) through different pathways, which is dependent on the nature of electrocatalysts and the electrolytic conditions (e.g., applied potential, electrolyte, etc).^[7] The first step for CO₂ activation to generate the CO₂•^- radical intermediate is difficult without a catalyst.^[8] But with the help of an electrocatalyst, the CO₂•^- radical can be stabilized through a chemical bond generated between CO₂ and the electrocatalyst, resulting in a less negative redox potential. Generally, selective production of desired products with high efficiency as well as high selectivity is desirable. Among the various CO₂RR products, only CO and HCOOH production have achieved remarkable selectivity close to 100%, showing the great potential for industrial application.^[9–12] Furthermore, as the main component of syngas, CO is a crucial raw material along with H₂ for the Fischer-Tropsch synthesis.^[13] Notably, despite the equilibrium potential for CO₂ reduction to CO is low (-0.11 V vs RHE at pH = 7), a higher overpotential is typically required to overcome the sluggish reaction kinetics.^[14] The high overpotential would also cause an enhanced hydrogen evolution reaction (HER), competing with the CO₂RR, which lowers the CO selectivity.^[15] Therefore, the development of effective electrocatalysts is highly critical to reduce the overpotential and suppress HER to achieve high selectivity and activity in CO₂RR.

Recently, several types of electrocatalysts for efficient CO₂-to-CO conversion have been reported, including noble metals (e.g., Au, Ag, Pd), Zn as well as carbon-based materials,^[14,16–18] among which, single-atom catalysts (SACs) have attracted great attention.^[19] Different from conventional nanocatalysts, SACs downsize to atomic level that can boost the exposed active sites and achieve a maximum atom utilization.^[20,21] In SACs, the supporting substrates, e.g., carbon-based materials, are not only used to anchor single metal atoms but also used to tune the charge distribution and electronic structure of the metal atoms. Besides, for the carbon-rich supports, doping N into the carbon matrix to introduce additional defects offers an efficient way to anchor single metal sites and tune the electronic structure of the carbon surface. Some reports used M-N-C to represent the carbon-based catalysts containing atomically dispersed M sites anchored on N-doped carbon support,^[19] which show great potential for electrochemical CO₂-to-CO reduction. M-N-C catalysts for CO₂ electroreduction is becoming highly attractive in recent years because of their high CO selectivity, good stability, and low cost. Recently, some reviews on the use of SACs for CO₂ electroreduction have been summarized from the aspects of catalyst preparation, characterization, and performance,^[22–25] while some other reviews focus on reviewing different types of catalysts including metal-based and carbon-based electrocatalysts for CO₂ electroreduction to CO.^[14,16] However, so far there is no comprehensive review of different coordination structures of transition metal/N-doped carbon (M-N-C) catalysts that have been reported for CO₂RR. Herein, experimental and theoretical investigations of M-N-C catalysts are summarized here to understand the effect of the active center and the local atomic environments on activity and selectivity, as well as the role of nonmetal moieties and metal nanoparticles in M-N-C in electrochemical CO₂RR. A visual performance comparison of M-N-C catalysts with different central metal atoms for CO₂ reduction reaction (CO₂RR) reported over the recent years is also given.

Reaction Mechanism of Electrochemical CO₂ Reduction to CO

Figure 1 displays the possible reaction pathway for CO₂ electroreduction to CO and hydrogen evolution reaction (HER).^[26] The concerted proton-electron transfer (CPET) reaction (Equation 1-1) is usually considered as the first step toward CO production. Besides, a proton decoupled electron transfer process is proposed, where *COOH is formed through two steps (Equations 1–2 and 1–3). Previous studies suggested that the rate-determining step (RDS) could be the reaction (1-1), (1-2), or (3), depending on the binding energy of CO₂^-* and COOH* as well as the desorption energy of CO*. Therefore, it is critical to tune the adsorption strength of the key intermediates to develop ideal electrocatalysts for CO₂ reduction to CO, which should strongly bind *COOH but weakly bind *CO so that it could promote both the *COOH formation and the *CO desorption steps. However, the adsorption of COOH* and CO* are usually linearly related, especially on transition metals.^[27] Therefore, the binding energy of COOH* (or CO*) was used as a descriptor for CO₂RR to CO, and some studies have shown a volcano relationship between adsorption energy of COOH* (or CO*) and catalytic activity.^[26,28] In addition to CO₂RR intermediates, hydrogen adsorption on the catalyst's surface also needs to be considered since HER is the major competing reaction to CO₂RR. Strong binding of *H is desired as it can lead to a large overpotential in HER and facilitate CO₂ electroreduction. [Image Omitted. See PDF][Image Omitted. See PDF][Image Omitted. See PDF][Image Omitted. See PDF][Image Omitted. See PDF]

Figure 1. Proposed reaction pathway for CO2 reduction to CO and HER (C atoms: gray; O atoms: red; H atoms: orange; and electrode: blue).

Over the past decades, many novel electrocatalysts have been reported for selective CO₂-to-CO conversion. Noble metal-based electrocatalysts such as Au,^[29] Ag,^[30] and Pd^[31] were regarded as one of the major classes for efficient CO₂RR to CO with high selectivity (>80%) and activity because they could stabilize *COOH intermediate. However, for further industrial applications, low-cost and earth-abundant electrocatalysts with excellent performance need to be exploited to replace the precious metals. Some non-noble metal-based catalysts like Zn,^[32] bimetallic Cu–Sn^[33] and Cu–In^[34] were found capable of selectively reducing CO₂ to CO. Recently, a new class of heteroatom-doped carbon-based materials,^[35–37] have been shown very active in CO₂RR to CO in aqueous solution. For the metal-free N-doped carbon-based catalysts, the pyridinic-N and/or graphitic-N defects were identified as the active sites for CO₂ reduction.^[38–41] Notably, the binding energies of *COOH and *CO intermediates on the N-doped sites no longer follow the linear scaling relationship that generally appears in pure transition metals, which is possibly due to the heterogeneity caused by the N dopants into the carbon matrix.^[38] Furthermore, inspired by heterogeneous organometallic complexes with high activity toward CO₂ electroreduction,^[42–44] introducing single transition metal sites into N-doped carbon matrix to obtain heterogeneous M-N-C electrocatalysts, was observed capable of significantly improving the current density and CO selectivity.^[45–47] For example, Strasser et al.^[48] reported that introducing metal centers (Fe or Mn) into N-doped carbon matrix could significantly enhance the CO₂RR activity in contrast with metal-free N-doped carbon catalysts and the single M-N_x sites were regarded as the dominant active centers responsible for high CO₂-to-CO activity.

Regulation of Active Center and Local Atomic Environment in M-N-C Catalysts

M-N-C catalysts are typically synthesized through high-temperature pyrolysis (usually above 700 °C), during which precursor containing a transition metal, C (and N) was decomposed under inert gas (i.e., N₂, Ar) or NH₃ gas atmosphere. During the high-temperature treatment, the metal center was coordinated with N atoms to form active M-N_x sites, whereas the structures were usually uncontrolled. Heterogeneous molecular M-N-C catalysts with well-defined M−N₄ moieties could be obtained via attaching metallomacrocycles onto carbon supports.^[49–51] The central metal atom and the coordination environment have a significant effect on the intrinsic electrocatalytic activity of M-N-C catalysts in CO₂RR (Figure 2). The coordination environment includes the first shell, second shell, and even higher shell of the metal center.

Figure 2. Schematic illustration showing regulation of central metal atom and local atomic environment in M-N-C catalysts (C atoms: gray; N atoms: blue; and metal atoms: purple).

Here we divided the regulation strategies into five classes, including regulation of central metal atom, engineering the first shell of the metal coordination environment, engineering the second and higher shell of the metal coordination environment, other heteroatom doping, and establishing dual-atom site catalysts. The regulation of the central metal atom is easy to be operated by changing metal salt precursor, which is closely related to the main CO₂RR products. In contrast to the central atoms, the surrounding coordination atoms own a broader regulation range. Currently, most studies focused on engineering the first shell of the metal coordination environment, and the approach of accurately synthesizing atomically dispersed M-N₄ sites has been developed. However, the accurate synthesis of a more compact coordination structure is still an obstacle due to the inhomogeneity caused by the high temperature used in the preparation of M-N-C catalysts. In addition to C or N coordinated atoms in M-N-C, other heteroatom doping provides more opportunities to adjust the electronic structure of the central atoms. Establishing dual-atom sites offers another way to reduce the barriers for the key reaction intermediates formation in CO₂RR and may be promising for syngas or C₂₊ products production.

Regulation of Central Metal Atom Activity Comparison of Different Metal Centers

The CO₂RR selectivity largely relies on the nature of the metal center in M-N-C electrocatalysts. The central metal atoms in M-N-C materials reported for CO production include Mn, Fe, Co, Ni, Cu, and Zn. Strasser et al.^[47] developed a variety of M-N-C eletrocatalysts (M = Mn, Fe, Co, Ni, Cu) for CO₂ electroreduction to explore the influence of the central metal atom. DFT calculations (Figure 3a) employing typical M-N₄ sites (Figure 3b) indicated that the M-N-C electrocatalysts could be classified into two groups: Fe, Mn, and Co-N-C catalysts with strong binding to *COOH requiring low overpotentials for CO₂-to-CO conversion, and Cu and Ni-N-C catalysts with weak binding to *COOH requiring considerably higher overpotentials. Moreover, the Ni-N-C with weak binding to H* inhibited HER, while the Fe, Mn, Co-N-C with stronger binding of H* displayed a higher HER activity in the large potential range, leading to a lower CO selectivity. Notably, Cu-N-C is an exception because of the thermodynamic instability of Cu-N_x sites at large overpotentials. Consequently, the CO selectivity in the whole potential range follows: Ni-N-C > Fe-N-C > Mn-N-C > Cu-N-C ≈ Co-N-C (Figure 3b). Considering the trade-off between energy input and product yield, the Ni-N-C catalyst achieving the highest CO selectivity at high overpotentials and the Fe-N-C catalyst achieving relatively high CO selectivity at low overpotentials are the most promising candidates. Other researchers obtained similar results.^[52–54] It is noted that most of the M-N-C electrocatalysts are prepared by a high-temperature pyrolysis approach, it is hard to rule out the effect of the surrounding coordinating atoms when comparing different metal active centers. Jiang et al.^[55] synthesized a series of M-N-C (M = Ni, Fe, Co, Cu) catalysts from multivariate metal-organic frameworks (MTV-MOFs) with similar metal coordination environments (pyridinic-type M-N₄), which provided an ideal model for the investigation of the intrinsic activities of various single metal atoms for CO₂RR. Among the synthesized catalysts, Ni-N-C achieved a maximum CO selectivity of 96.8%, followed by Fe-N-C (86.5%), Cu-N-C (14.0%), and Co-N-C (17.8%). DFT calculations were performed and the limiting potential difference between CO₂RR and HER (U_L(CO₂)-U_L(H₂)) was calculated (Figure 3c), used as a descriptor for product selectivity, which was well consistent with the experimental results. Moreover, a universal ligand-mediated approach for transition metal (including Cr, Mn, Fe, Co, Ni, Cu, Zn, Ru, Pt) single-atom catalysts synthesis containing MN₄ sites over carbon support was developed.^[56] Interestingly, a volcano relation between CO FE and metal atomic number was observed (Figure 3d), which might be related to the d-band center of metal in the M-N-C catalysts.

Figure 3. M-N-C catalysts with different central metal atoms for CO2-to-CO conversion. a) Three potential regions with distinctly different rate-determining mechanistic features for M-N-C (M = Mn, Fe, Co, Ni, Cu) catalysts. b) CO FE versus applied potentials for M-N-C catalysts (M = Mn, Fe, Co, Ni, Cu). c) The values of UL(CO2)-UL(H2) for all M-N-C catalysts. d) CO FE for M-N-C at -1.2 V versus RHE. a,b) Reproduced with permission.[47] Copyright 2017, Springer Nature. c) Reproduced with permission.[55] Copyright 2020, WILEY-VCH. d) Reproduced with permission.[56] Copyright 2019, Springer Nature.

The CO₂RR performance of the recently reported M-N-C catalysts with different central metal atoms is summarized in Tables S1–S6 (Supporting Information) and Figure 4. As displayed in Figure 4, the CO selectivity and onset potential of different M-N-C catalysts are compared. Ni-N-C, Fe-N-C and Zn-N-C catalysts are the most promising candidates for large-scale CO₂ conversion to produce CO. The Ni-N-C catalysts achieved maximum CO FEs over 90% in the potential range from -0.7 to -0.9 V versus RHE, while Fe-N-C reached an average CO FE of ≈80% under lower overpotentials. Notably, the Zn-N-C catalysts, which are rarely reported in contrast with Ni-N-C and Fe-N-C, display the lowest onset potential and tend to reach high CO selectivity at low overpotentials. For instance, a Zn-N-G catalyst, in which the single Zn atomic site was coordinated with four N atoms on graphene, exhibited a maximum CO selectivity up to 91% at -0.5 V.^[57] DFT calculation suggested that the Zn-N_x sites facilitated COOH* intermediate formation and CO* desorption, leading to enhanced activity for CO₂ electroreduction. As atomically dispersed Zn sites showed excellent activity for CO₂-to-CO conversion, it is crucial to exclude the existence of atomically dispersed Zn sites when Zn-based MOF materials were used as precursors to prepare M-N-C catalysts for CO₂RR. Other M-N-C catalysts (M = Co, Cu, Mn) displayed quite different CO FEs for distinct samples, some of which showed comparable selectivity to Ni and Fe-N-C, while others exhibited lower values. This distinction may be attributed to the different coordination environments of central atoms. Despite Cu-N-C catalysts showed low selectivity for CO₂ to CO production, they deserve more research attention as they have the potential to reduce CO₂ to multi-carbon products.^[58–60]

Figure 4. a) Reported maximum CO FEs and b) onset potentials for various M-N-C catalysts (M = Mn, Fe, Co, Ni, Cu, Zn). The point at the upper right of the dotted line in Figure 4a indicates that high CO selectivity can be achieved at a low applied potential. Data displayed in this figure are obtained from Tables S1–S6 (Supporting Information).

Activity Comparison of Ni SACs

Among various M-N-C electrocatalysts, Ni-N-C catalysts have shown the best performance and drawn great attention for electrochemical CO₂ reduction to CO. Therefore, we further compared the CO₂RR performance of some recently reported Ni-N-C catalysts with different Ni-N_x sites (Figure 5). It can be observed that in all cases CO FE follows a similar trend (Figure 5a), achieving a maximum value higher than 80% in the potential of -0.7–-0.9 V for Ni-N-C catalysts with NiN₄ moieties. Figure 5c compares the CO₂RR activity of the Ni SACs with Ni coordination number to N lower than 4. Most of these catalysts exhibit high CO FEs over 90%, similar to those Ni-N-C catalysts with dominant NiN₄ moieties. But they seem to reach the maximum CO selectivity under different potential ranges, which may be due to the different coordination environments of the central metal atom. For the purpose of practical application, CO partial current density (j_CO) should be the main consideration as it represents the product yield. In addition to improving the intrinsic reactivity of active centers, increasing the density of active sites in the Ni-N-C catalysts can also effectively enhance the electrocatalytic efficiency. Nevertheless, it is notable that the Ni content is not always proportional to the j_CO in different Ni-N-C electrocatalysts, which may be due to the fact that some single Ni atoms are encapsulated within the carbon matrix that are inaccessible to CO₂ molecules. The future direction towards improving the j_CO should maximize exposure of active single metal sites at the catalyst's surface.

Figure 5. Catalytic activity comparison for various previously reported Ni-N-C catalysts with NiN4 moieties and NiNx (x [less than] 4) moieties. a) CO FE and b) jCO versus applied potential for Ni-N-C catalysts with NiN4 moieties. The data were obtained from the following studies: [A] Li et al.;[10] [B] Lu et al.;[61] [C] Yang et al.;[62] [D] Pan et al.;[63] [E] Jeong et al.;[64] [F] Zhao et al.;[65] [G] Zheng et al.;[66] [H] Yuan et al.;[67] [I] Sa et al.;[68] [J] Liu et al.;[50] [K] Chen et al.;[69] [L] Han et al..[70] c) CO FE and b) jCO versus applied potential for Ni-N-C catalysts with NiNx (x [less than] 4) moieties. The data were obtained from the following studies: [A] Zhang et al.;[71] [B] Jiang et al.;[72] [C] Yan et al.;[73] [D] Mou et al.;[74] [E] Fan et al.;[75] [F] Yang et al.;[76] [G] Cheng et al.;[77] [H] Gong et al.;[78] [I] Sa et al.;[68] [J] Rong et al.;[79] [K] Zhang et al.[80]

Engineering the First Shell of the Metal Coordination Environment

The first shell represents the surrounding atoms that directly bond to the central metal atom, which can be adjusted by tuning the coordination number or changing the chemical identity of neighboring atoms. In this section, we mainly introduce M-N-C catalysts with typical M-N₄ structure, M-N₃ structure, M-N_5–6 structure and M-N_xC_4-x (x = 1–3) structures for electrochemical CO₂ reduction.

M-N₄ Moieties

Similar to homogeneous M-cyclam^[42] materials, M-N₄ moieties in heterogeneous M-N-C electrocatalysts were widely considered as the active sites for CO₂ electroreduction to CO.^[47,54,56] For example, Wu et al.^[10] established a pyridine-type Ni-N₄ structure via a topo-chemical transformation method, which could avoid agglomeration of nickel species and ensure maximum maintenance of the Ni-N₄ structure (Figure 6a). The obtained Ni-N₄-C catalyst with abundant active sites exhibited a maximum CO selectivity up to 99% with a high j_CO up to 28.3 mA cm⁻² at -0.81 V. DFT calculation results showed that the high CO₂RR activity could be attributed to the reduced binding energy of COOH* by pyridine-type Ni-N₄ (1.7 eV) sites. However, it should be noted that the calculated *COOH formation energy of 1.7 eV is still a bit too high, higher than ΔG_H,^[10,47,55] which is contradictory to the experimental observation of high CO selectivity over H₂. Liu and co-workers proved atomically dispersed low-valent Ni(I) as the active site for CO₂ electroreduction by performing operando XAS measurements.^[62] During electrolysis, the unpaired electron in the Ni 3d_x2−y2 orbital would be delocalized and the charge would spontaneously mitigate from Ni(I) to the carbon 2p orbital in CO₂ to generate a CO₂^δ− species (Figure 6b). The obtained SAC maintained a superior CO FE up to 98% with a j_CO of 22 mA cm⁻² for 100 h at -0.72 V.

Figure 6. M-N-C with different MN4 sites for CO2-to-CO conversion. a) Schematic illustration showing the topo-chemical transformation strategy for Ni-N4-C synthesis (Ni atoms, green; N atoms, blue; C atoms, gray; O atoms, red). b) Structural evolution of the Ni(I) active site in electrochemical CO2 reduction. c) Calculated Gibbs free energy diagrams for CO2 electroreduction to CO on Fe-N4 moieties. d) First shell fitting of Fourier transformation of EXAFS spectra for NiPor-CTF and the model NiN4 structure in NiPor-CTF. e) The first derivative of the Fe-edge XANES spectra at different potentials. a) Reproduced with permission.[10] Copyright 2017, American Chemical Society. b) Reproduced with permission.[62] Copyright 2018, Springer Nature. c) Adapted with permission.[81] Copyright 2018, WILEY-VCH. d) Reproduced with permission.[61] Copyright 2019, WILEY-VCH. e) Adapted with permission.[46] Copyright 2019, AAAS.

Pyridine-type Fe-N₄ sites have been shown to bind CO strongly, leading to limited CO selectivity in CO₂RR. Pyridine-type Fe-N₄ catalysts with different graphitic nitrogen doping configurations were further studied (Figure 6c).^[81] The N atoms substitution on graphene was found able to reduce the energy barrier for COOH* formation over FeN₄ sites (from 0.63 to 0.29 eV), and thus facilitated CO* desorption (from 0.5 to 0.3 eV).

In addition to the pyridine-type M-N₄ sites, pyrrole-type M-N₄ sites can also reduce CO₂ to CO. For instance, a Ni porphyrin-based covalent triazine framework (NiPor-CTF) catalyst with abundant atomically dispersed pyrrole-type NiN₄ sites (Figure 6d) was prepared through a conventional ionothermal strategy for effective electrochemical CO₂ reduction.^[61] The NiPor-CTF achieved a high CO productivity with a maximum CO FE up to 97% and a j_CO of 51.3 mA cm⁻² at −0.9 V versus RHE for 20 h. Theoretical calculations indicated that the Gibbs free energy of *COOH on pyrrole-type NiN₄ sites (1.49 eV) embedded in CTF skeleton slightly reduced as compared to that on pyridine-type NiN₄ (1.55 eV), explaining the improved CO₂RR performance as compared to Ni/N-PC with pyridine-type NiN₄. Moreover, the type of N ligands could affect the oxidation state of the central metal. Fe³⁺ atoms coordinated with pyrrolic N (Fe²⁺-N-C) exhibited a high CO FE of over 80% in CO₂RR in a low potential range of -0.2– -0.5 V versus RHE, reaching a j_CO of 20 mA cm⁻² at -0.47 V, better than Fe²⁺ atoms coordinated with pyridinic N (Fe²⁺-N-C).^[46] This result was attributed to the pyrrolic N ligands, which enabled stabilizing Fe³⁺ during electrocatalysis, promoting CO₂ adsorption and CO desorption. If the applied potential was decreased to less than -0.5 V versus RHE, Fe³⁺ sites would be reduced to Fe²⁺ with coordination number decreased from 4 to 3 (Figure 6e), leading to a significant decrease in CO₂RR activity.

M-N₃ Moieties

Although it has been reported that HER dominates on the NiN₃ site (pyridinic),^[73] there are a few studies predicting that the Ni@N₃ (pyrrolic) sites are likely the major active sites for CO₂ electroreduction.^[75–76] Single Ni atom catalyst with a possible NiN₃ structure on N-doped carbon nanotubes (named as NC-CNTs (Ni)) was prepared through an in situ thermal transformation method.^[75] The obtained NC-CNTs (Ni) possessed NiN₃ moieties, proved by extended X-ray absorption fine structure (EXAFS), and showed a CO FE over 90% as well as a j_CO of ≈10 mA cm⁻² at −1.0 V versus RHE. Importantly, this study provided a DFT comparison of the two different Ni@N₃ sites including Ni@N₃(pyridinic) and Ni@N₃(pyrrolic) site for CO₂-to-CO reduction (Figure 7a). The calculations revealed that the binding energy of *COOH on the Ni@N₃(pyrrolic) site (1.09 eV) was lower than that on the Ni@N₄ site (1.54 eV), implying that Ni@N₃(pyrrolic) site was more active for CO₂ reduction to CO than the Ni@N₄ site. Besides, the desorption of CO on the Ni@N₃(pyrrolic) site was exothermic, suggesting less *CO poisoning. Moreover, ΔG(*COOH) was more negative than ΔG(*H) on both pyridinic and pyrrolic Ni@N₃ moieties, illustrating that CO₂RR was thermodynamically more preferred than HER. Optimization of Ni atom structures on pure and N-doped graphene was performed via DFT by Wu et al.,^[76] which revealed that the most stable adsorption sites were pyrrolic N (binding energy 6.98 eV) and pyridinic N (binding energy 7.75 eV) (Figure 7b). Therefore, three structures including Ni-4N and Ni-3N (Ni@N₃(pyrrolic)) representing Ni SAs and Ni (111) standing for Ni NPs were comparatively studied, which indicated that Ni-3N (Ni@N₃(pyrrolic)) was more active than Ni-4N in CO₂RR to CO due to its lower *COOH binding energy while Ni NPs were inactive due to its high required energy for CO desorption. Apart from Ni-N-C, Mn-N-C with Mn-N₃ sites embedded in g-C₃N₄ on carbon nanotubes (Mn-C₃N₄/CNT) was also constructed, which exhibited a CO selectivity up to 98.8% with a j_CO of 14.0 mA cm⁻² at −0.55 V versus RHE, outperforming all of the Mn-N-C electrocatalysts previously reported.^[82] In situ XAS analysis as well as DFT calculations revealed that the outstanding performance was derived from the Mn-N₃ sites that facilitated the COOH* intermediate formation by decreasing the free energy barrier compared to Mn-N₄ sites (Figure 7c,d).

Figure 7. M-N-C catalysts with MN3 sites for CO2-to-CO conversion. a) Optimized atomic structures including NiN3, NiN3V, and NiN2V2 as well as DFT calculated free energy profiles for CO2 electroreduction and comparison of ΔG(*H) and ΔG(*COOH). C, N, and Ni atoms are represented by gray, blue, and purple spheres, respectively. b) The adsorption energies of Ni adatom and Ni dimer vertical of pyrrolic N, pyridinic N, quaternary N, and graphene. All values are in eV per Ni atom. c) EXAFS spectra at the Mn K-edge of Mn–C3N4/CNT under various conditions. d) Calculated Gibbs free energy diagrams for CO2RR on Mn-N3-C3N4 and Mn-N4-G. a) Reproduced with permission.[75] Copyright 2019, WILEY-VCH. b) Adapted with permission.[76] Copyright 2018, WILEY-VCH. c,d) Adapted with permission.[82] Copyright 2020, Springer Nature.

M-N_5-6 moieties

M-N-C electrocatalysts with the central metal atom M coordinated with more than four N atoms were also recently studied. For instance, Co-N-C and Fe-N-C with atomically dispersed CoN₅ and FeN₅ sites, respectively were both reported to be extremely efficient in CO₂-to-CO conversion.^[83,84] Pan et al.^[84] synthesized a class of M-N₅/HNPCSs (M = Co, Fe, Ni, Cu) catalysts with dominant atomically dispersed M−N₅ sites and compared their performance for CO₂ to CO production. Co-N₅/HNPCSs exhibited the highest CO FE of 99.4% at −0.79 V versus RHE (Figure 8a), outperforming most other previously reported Co-N-C electrocatalysts. The Co-N₅ moieties were determined to be the active sites for CO₂RR, which promoted COOH* formation as well as CO desorption (Figure 8b-c). The Fe-N₅ catalyst with superior CO FE and j_CO to the Fe-N₄ catalyst was also reported by Zhang et al. (Figure 8d,e).^[83] Theoretical calculations (Figure 8f) show that the rate-determining step on FeN₄ is the *CO desorption that requires an energy of 1.35 eV, while that over FeN₅ is the *COOH formation step with an energy barrier of 0.77 eV. Moreover, as shown in Figure 8g, the axial pyrrolic-N ligand on the FeN₅ sites could deplete the electron density of Fe 3d orbitals and reduce the Fe-CO p back-donation as compared to FeN₄, hence leading to fast desorption of CO. Furthermore, Fe-N₆ catalyst was compared to Fe-N₅ by Chen and co-workers,^[85] in which, Fe-N₅ catalyst reached a lower overpotential (50 mV) and exhibited a higher CO FE (up to 99%), j_CO and CO TOF than the Fe-N₆ catalyst in a broad potential range of −0.35–−1.05 V versus RHE. The superior performance was attributed to the Fe-N₅ site, which facilitated the *COOH formation (Figure 8h,i).

Figure 8. M-N-C with M-N5 structure for CO2-to-CO conversion. a) CO FEs and H2 FEs of M-N5/HNPCSs (M = Co, Fe, Ni, Cu). b) Optimized structures for the intermediates. c) Calculated free energy diagrams for CO2 electroreduction to CO. d) CO FEs versus different applied potentials. e) jCO versus different applied potentials. f) Calculated Gibbs free energy diagrams for CO2 electroreduction to CO on Fe-N4 and Fe-N5. g) Local DOS and partial charge density of the FeN4 and FeN5 with adsorbed CO. h) Calculated free energy diagrams for CO2 electroreduction to CO on Fe-N5 and Fe-N6. i) Optimized structures of Fe-N5 and Fe-N6 and a slice of calculated charge densities with adsorbed *COOH. a–c) Adapted with permission.[84] Copyright 2018, American Chemical Society. d–g) Adapted with permission.[83] Copyright 2019, Wiley-VCH. h,i) Adapted with permission.[85] Copyright 2020, The Royal Society of Chemistry.

M-N_xC_4-x (x = 1–3) Moieties

In addition to the typical M-N₄ structure, the M-N and M-C mixed structures were also studied for electrochemical CO₂ reduction to CO. Wu et al.^[86] prepared atomically dispersed Co on N-doped carbon with different M-N/C mixed structures (Co-N₄, Co-N₃C, or Co-N₂C₂) by changing pyrolysis temperature. It was found that the Co-N₂ (Co-N₂C₂) catalyst reached a CO selectivity up to 94% with a total current density of 18.1 mA cm⁻² at -0.63 V versus RHE, far superior to that for Co-N₃ and Co-N₄ (Figure 9a). DFT results (Figure 9b) indicated that the low coordination number of Co-N₂ could facilitate CO₂ activation and thus lead to enhanced CO₂RR activity. However, the strong binding of CO* on the Co-N₂ site still inhibited CO formation. On the contrary, some works reported that Co-N₄ sites owned superior performance to Co-N_4-xC_x sites.^[45]

Figure 9. M-N-C catalysts with different MNxC4-x sites for CO2-to-CO conversion. a) CO FE versus applied potential for Co-N2, Co-N3, and Co NPs catalysts. b) Calculated Gibbs free energy diagrams for CO2 electroreduction to CO on Co-N2 and Co-N4. c) EXAFS fitting curves for Ni SAs/N-C showing NiN3C structure. d,e) EXAFS and fitting curves for NiSA-N2-C showing NiN2C2 structure f) CO FE versus applied potential for NiSA-NX-C catalyst. g) Optimized atomic structures including NiN4, NiN3C, NiN2C2, and NiNC3 as well as DFT calculated free energy profiles for CO2 electroreduction. C, N, and Ni atoms are represented by gray, blue, and purple spheres, respectively. a,b) Adapted with permission.[86] Copyright 2018, Wiley-VCH. c) Reproduced with permission.[71] Copyright 2017, American Chemical Society. d–f) Adapted with permission.[78] Copyright 2020, Wiley-VCH. g) Reproduced with permission.[87] Copyright 2019, Elsevier Inc.

Additionally, a Ni single-atom catalyst (SAC) with a dominant NiN₃C structure was prepared by an ionic exchange process.^[71] The obtained Ni single-atom catalyst (Ni SAs/N-C), in which the Ni atoms were anchored with three N atoms and one C atom in the carbon matrix according to the XAFs fitting (Figure 9c), was able to reduce CO₂ to CO with a selectivity of ≈71.9% at -1.0 V. The NiN₃C structure was considered as the active site that exhibited strong bonding to CO₂^•−. In another work, NiN₂C₂ structure was reported more active in CO₂RR than NiN₃C and NiN₄.^[78] A series of Ni_SA-N_x-C catalysts (x = 2–4) with different N coordination numbers (Figure 9d,e) were prepared by changing the pyrolysis temperature and studied for CO₂RR. The Ni_SA-N₂-C catalyst with the lowest N coordination number suggested the most superior performance for CO production with CO FE up to 98% (Figure 9f). DFT calculations revealed that the low N coordination number was favorable to COOH* intermediate formation. Similar DFT calculations for different NiN_xC_4-x structures including NiN₄, NiN₃C, NiN₂C₂, NiNC₃ for CO₂-to-CO electroreduction were performed in other works.^[78,87–88] As shown in Figure 9g, the results suggested that NiNC₃ (0.59 eV) was the most favorable site for COOH* formation, followed by NiN₃C (1.10 eV), NiN₂C₂ (1.11 eV), and NiN₄ (1.64 eV). Besides, for these different NiN_x structures, CO* desorption step is exothermic, revealing easy desorption of adsorbed CO* intermediate to release molecular CO.

Furthermore, the CO₂RR performance of some M-N-C catalysts with these four structures (typical M-N₄ structure, M-N₃ structure, M-N_5∼6 structure and M-N_xC_4-x (x = 1–3) structures) was compared as shown in Table 1. For the CO₂RR performance comparison of M-N-C catalysts with different metal centers, the nature of the metal centers is still the decisive factor. While for catalysts with the same metal center, the coordination structure has different effects. Notably, most of the coordination structures reported in the literature were only an average structure determined by EXAFS, not a homogeneous coordination structure. Therefore, it is necessary to further develop M-N-C catalysts with a homogeneous coordination structure to accurately compare the activity of different coordinative active sites

Table 1 Comparison of CO₂RR performance for M-N-C catalysts with four structures (typical M-N₄ structure, M-N₃ structure, M-N_5∼6 structure and M-N_xC_4-x (x = 1–3) structures)

M-N-C catalyst	Active site(s)	Metal content(wt%)	Onset potential (vs RHE)	CO FE[%]	jCO[mA cm−2]	TOF[h−1]
Ni-N4-C[10]	Pyridine-type Ni-N4	1.41	−0.4 V	99 @−0.81 V (vs RHE)	28.3 @−0.81 V (vs RHE)	/
A-Ni-NG[62]	Pyridine-type Ni-N4	4.6	−0.35 V	97 @−0.72 V (vs RHE)	30.6 @−0.72 V (vs RHE)	8000 @−0.72 V (vs RHE)
A-Ni-NSG[62]	Pyridine-type Ni-N4	2.5	−0.25 V	97 @−0.72 V (vs RHE)	22 @−0.72 V (vs RHE)	14 800 @−0.72 V (vs RHE)
Fe/NG-750[81]	Pyridine-type FeN4	1.25	−0.3 V	80 @−0.57 V (vs RHE)	2 @−0.6 V (vs RHE)	/
NiPor-CTF[61]	Pyrrole-type Ni-N4	2.4	−0.44 V	97 @−0.9 V (vs RHE)	51.3 @−0.9 V (vs RHE)	1690 @−0.9 V (vs RHE)
NC-CNTs (Ni)[75]	NiN3	1.01	−0.4 V	90 @−0.8 V (vs RHE)	9.2 @−1 V (vs RHE)	11 650 @−0.89 V (vs RHE)
SE-Ni SAs@PNC[76]	NiN3 NiN4	/	/	95.7 @−0.8 V (vs RHE)	18.3 @−1 V (vs RHE)	47 805 @−1 V (vs RHE)
Mn-C3N4/CNT[82]	Mn-N3	0.17	−0.3 V	98.8 @−0.55 V (vs RHE)	14 @−0.55 V (vs RHE)	/
Ni-N5/HNPCSs[84]	NiN5	3.32	/	50 @−0.79 V (vs RHE)	/	/
Fe-N5/HNPCSs[84]	FeN5	3.03	/	82 @−0.66 V (vs RHE)	/	/
Co-N5/HNPCSs[84]	CoN5	3.54	/	99.4 @−0.79 V (vs RHE)	4.5 @−0.73 V (vs RHE)	480.2 @−0.73 V (vs RHE)
Cu-N5/HNPCSs[84]	CuN5	3.75	/	2 @−0.73 V (vs. RHE)	/	/
Fe-N5[83]	FeN5	0.8 at%	−0.26 V	97 @−0.46 V (vs RHE)	1.81 @−0.46 V (vs RHE)	/
Fe-N4[83]	FeN4	/	/	66 @−0.56 V (vs RHE)	0.3 @−0.56 V (vs RHE)	/
Fe-N5[85]	FeN5	0.57	−0.16 V	99 @−0.35 V (vs RHE)	20.8 @−1.05 V (vs RHE)	5006 @−1.05 V (vs RHE)
Fe-N6[85]	FeN6	0.58	−0.2 V	96 @−0.35 V (vs RHE)	5.5 @−1.05 V (vs RHE)	1324 @−1.05 V (vs RHE)
Ni SAs/N-C[71]	NiN3C	1.53	−0.57 V	71.9 @−0.9 V (vs RHE)	7.37 @−1 V (vs RHE)	5273 @−1 V (vs RHE)
Co-N2[86]	CoN2C2	0.25	−0.22 V	94 @−0.63 V (vs RHE)	17 @−0.63 V (vs RHE)	18 200 @−0.63 V (vs RHE)
Co-N3[86]	CoN3C	/	−0.45 V	63 @−0.53 V (vs RHE)	1.56 @−0.53 V (vs RHE)	1250 @−0.63 V (vs RHE)
Co-N4[86]	Pyridine-type CoN4	/	/	4.2 @−0.83 V (vs RHE)	/	84 @−0.83 V (vs RHE)
Co1-N4[45]	Pyridine-type Co-N4	0.6	−0.3 V	82 @−0.8 V (vsRHE)	15.8 @−0.8 V (vs RHE)	1455 @−1.0 V (vs RHE)
Co1-N4-x[45]	Co-Nx	0.63	−0.32 V	47 @−0.8 V (vs RHE)	8.8 @−0.8 V (vs RHE)	763 @−1.0 V (vs RHE)

Engineering the Second and Higher Shell of the Metal Coordination Environment Vacancy Engineering

Some studies found that coordinatively unsaturated M-N sites formed by engineering surface vacancies in carbon matrix could improve the binding strength of reaction intermediates and thus reduce the reaction free energy of CO₂RR.^{[73–74,89–90]} For instance, a series of improved CO₂RR electrocatalysts derived from Zn/Ni bimetallic ZIF-8 with different proportions of Zn and Ni were synthesized (Figure 10a), in which coordinatively unsaturated Ni-N sites were embodied in porous carbon matrix.^[73] In this work, four Ni-N architectures were compared by DFT, including NiN₄, NiN₃, NiN₃V, and NiN₂V₂, where V represents coordination vacancy (Figure 10c). According to the calculation results, the higher activity of CO₂RR (Figure 10b) could be attributed to the significantly lower free energy of *COOH on the coordinatively unsaturated Ni-N sites as compared to that on NiN₄. Similar calculation results were also reported in other recent studies.^[74,80] Besides, Sa et al.^[68] and Lu et al.^[79] prepared vacancy-defect Ni SAC to compare with typical NiN₄ SAC, and found that the vacancy-defect Ni-N-C with dominant NiN₃V sites established high CO turnover frequency (13860 h⁻¹ at −0.94 V vs RHE and 135 000 h⁻¹ at −0.9 V vs RHE, respectively) for CO₂ reduction to CO, far superior to the catalysts with NiN₄ sites (Figure 10d). Additionally, a novel Cu-N₂/GN catalyst with atomically dispersed undercoordinated Cu-N₂V₂ sites on graphene matrix was reported, which exhibited high electrocatalytic activity and selectivity (a maximum CO FE of 81%) for CO₂RR, showing an onset potential of −0.33 V versus RHE. ^[91] According to DFT calculation, the shorter bonding length of the Cu-N₂ site as compared to that of the Cu–N₄ site boosted the *COOH and then the *CO formation by accelerating electron transfer from Cu-N₂ site to *CO₂ (Figure 10e).

Figure 10. M-N-C catalysts with coordinatively unsaturated M-N structures for CO2-to-CO conversion. a) Scheme of C-ZnxNiy ZIF-8 synthesis. b) CO FE versus applied potential for various C-ZnxNiy ZIF-8 catalysts. c) Optimized atomic structures including NiN3, NiN3V, and NiN2V2 as well as DFT calculated free energy profiles for CO2RR. C, N, and Ni atoms are represented by gray, blue, and purple spheres, respectively. d) Illustration showing higher CO2-to-CO TOFs of H-NiPc/CNT with NiN3V sites than NiPc/CNT with NiN4 sites. e) Calculated Gibbs free energy diagrams for CO2 electroreduction to CO on Cu-N2 and Cu-N4. a–c) Reproduced with permission.[73] Copyright 2018, The Royal Society of Chemistry. d) Reproduced with permission.[68] Copyright 2020, American Chemical Society. e) Reproduced with permission.[91] Copyright 2019, WILEY-VCH.

Edge-Anchored Structure

It should be noted that the above studies only paid attention to the bulk M-N₄ sites for the CO₂RR while neglecting the active sites at the edges. In this regard, Wang's group prepared M-N-C (M = Fe, Co) catalysts containing both bulk and edge M−N₄ sites by employing Fe- or Co-doped MOF as precursors.^[92] The electrocatalytic activity of bulk and edge M-N₄ sites were compared in DFT calculations (Figure 11a), which found that the COOH dissociation to CO* and OH* was endothermic on M-N₄-C₁₀ (>1.11 eV) while it was exothermic on M-N₂₊₂-C₈ (←1.18 eV). Therefore, the edge M-N₂₊₂ site was found more active in catalyzing CO₂RR than the bulk M-N₄ site. The same conclusion was also reached in the Ni-N-C catalyst system.^[63]

Figure 11. M-N-C catalysts with edge-anchored Ni-Nx structures or with amino modification for CO2-to-CO conversion. a) The initial and final states for the *COOH dissociation reaction on M-N4-C10 and M-N2+2-C8 sites (M = Fe, Co). b) AC-STEM images showing that the single-atom Ni sites are predominately anchored at the edges of nanopores inside Ni-N-MEGO. DFT calculation results: c) Different Ni-N active site structures on the edges of graphene sheets. d) The reaction energy is in a linear relationship with the adsorption free energy of *CO (G*CO) on the active sites. Lower reaction energy indicates higher reactivity. A lower G*CO indicates weaker adsorption. e) Reaction pathway on the NiN2(NH2) and NiN3 site, with the free energy shown on top. (C: gray; N: blue; Ni: yellow; O: red; and H: white). f) Schematic of the local coordination environment for Ni-N4/C-NH2. g) Calculated Gibbs free energy diagrams for CO2 electroreduction to CO on Ni-N4/C-NH2 and Ni-N4/C. h) Projected DOS of Ni 3d in Ni-N4/C-NH2 and Ni-N4/C. a) Adapted with permission.[92] Copyright 2018, American Chemical Society. b–e) Reproduced with permission.[77] Copyright 2019, Elsevier B.V. f–h) Reproduced with permission.[69] Copyright 2021, The Royal Society of Chemistry.

In addition, unsaturated edge-anchored metal single atoms for electrochemical CO₂ reduction were also studied in comparison to the unsaturated in-plane M-N structures. Jiang et al.^[77] used microwave exfoliated GO as the porous support to prepare single Ni atom catalysts (Ni-N-MEGO), generating mangy edge-anchored single Ni atoms distributed along the edges of nanopores (<6 nm) (Figure 11b). The Ni-N-MEGO catalyst achieved a high CO selectivity of 92.1% at the potential of -0.7 V versus RHE in CO₂RR. Several different possible Ni-N_x (x = 2–4) structures were compared (Figure 11c) and the reaction energies were computed to assess CO₂RR activity (Figure 11d). As shown in Figure 11e, it could be deduced that edge-coordinated unsaturated NiN₃ and NiN₂(NH₂) architectures possessed the highest activity because of their optimized energies for both CO₂ activation (0.7 and 1.02 eV) and CO desorption (0.63 and 0.63 eV), while NiN₂ was inactive because of the high CO desorption energy.

Amino Modification

As amino-modified carbon materials showed enhanced CO₂ adsorption capacity, Liu's group^[69] modified the M-N-C electrocatalysts with amino functional groups with the well-maintained atomically dispersed structure to improve the CO₂RR activity (Figure 11f). The obtained Ni-N₄/C-NH₂ catalyst reached CO selectivities over 90% in a broad potential range, and achieved a j_CO up to 63.6 mA cm⁻² at -1.0 V versus RHE in a H-type cell, 2.5 times that of Ni-N₄/C catalyst. Moreover, an industrial level current density up to 440 mA cm⁻² with a relatively high CO selectivity over 85% was achieved in a flow cell. The superior performance was due to the reduced free energy of the COOH* intermediate derived from the regulation of the electronic distribution by amino modification (Figure 11g). The more positive d-band center of the Ni 3d DOS in Ni-N₄/C-NH₂ than that in Ni-N₄/C benefited the adsorption strength of the reaction intermediates due to declined occupation (Figure 11h).

Other Heteroatom Doping

The introduction of other heteroatoms (such as B, O, F, P, S, and Cl) into N-doped carbon to tune the electronic structure of single metal atoms provides another effective method to enhance electrocatalytic activity of the reaction.^[93–96] This strategy can manipulate the first shell of the coordination environment adjacent to the central metal atom or the higher shell bonded to the atom of the first shell to tune the electronic structure. For instance, although Mn-N-C catalysts with dominant MnN₄ sites showed low CO selectivity in CO₂RR, halogen (Cl/F/Br) and nitrogen dual-coordinated single Mn atom catalysts were found efficient in reducing CO₂ to CO.^[97] The single Mn atom in (Cl, N)-Mn/G was determined to be coordinated with a Cl atom in the axial direction and four N atoms in plane by EXAFS fitting analysis (Figure 12a). DFT calculation suggested obviously decreased free energy barrier for the rate-limiting step of CO desorption (Figure 12b), at the same time inhibiting HER (Figure 12c), which was because of the modified electronic structure of the single Mn atom active site via electron transfer between Mn and Cl (Figure 12d). A Ni-based single-atom catalyst (Ni-N₄-O/C) with a similar coordination structure, in which atomically dispersed Ni sites were coordinated with an O atom in the axial direction and four N atoms in plane, also exhibited outstanding CO₂RR performance with a maximum CO selectivity up to 100%.^[98] Besides, heteroatom doping into carbon matrix beyond the first shell of the central metal also has an impact on the distribution of electron density around the central metal atom, leading to changes in catalytic activity. For example, a F-tuned single-Ni-atom catalyst with NiN₄ structure (Figure 12e) was found to be capable of reducing CO₂ to CO in a higher efficiency than its counterpart without F-doping (Figure 12f), which was attributed to the lower energy barrier for CO₂ activation (Figure 12g).^[70] N, S co-doped Ni SAC showed a reduced overpotential compared to N-doped Ni SAC catalyst (Figure 12h).^[62]

Figure 12. M-N-C catalysts with other heteroatoms doping for CO2-to-CO conversion. a) EXAFS fitting curves of the (Cl, N)-Mn/G in R space. Inset shows the schematic model of (Cl, N)-Mn/G. b) Calculated free energy of CO2RR. c) Calculated free energy of hydrogen adsorption. d) Electron density difference for COOH* adsorbed on (Cl, N)-Mn/G (left) and N-Mn/G (right). The blue and red region denote the electron accumulation and electron depletion, respectively. e) Different model structures of the F-tuned single-Ni-atom catalyst. f) jCO versus applied potential. g) Calculated Gibbs free energy diagrams for CO2-to-CO conversion. h) CO Faradaic efficiency at various applied potentials. a–d) Adapted with permission.[97] Copyright 2019, Springer Nature. e–g) Adapted with permission.[70] Copyright 2021, Elsevier B.V. h) Adapted with permission.[62] Copyright 2018, Springer Nature.

Dual-Atom Site Catalysts

Dual-atom site catalysts are emerging as a promising candidate for electrochemical reactions.^[99–102] According to different relative positions of dual-atom sites, they can be classified into isolated dual-atom site catalysts and binuclear dual-atom site catalysts.

Isolated dual-atom site denotes two types of isolated central metal atoms anchored with surrounding chelating atoms, without forming metal–metal bond, which have been reported for CO₂RR to generate syngas (CO/H₂). For example, Chen and co-workers reported a series of NiFe dual-atom catalysts containing bimetallic centers for tunable syngas production (from 1:3 to 4:1) by changing the configurations of the metal-N sites (Ni-N₄ and Fe-N₄), as well as tuning the applied potential.^[103] Theoretical and experimental results revealed that there was a synergy of Ni-N₄ and Fe-N₄ sites for CO₂RR process, where the Fe atoms functioned both as the reactive and adsorption sites of intermediates for CO₂-to-CO conversion, while the introduction of Ni atoms lowered the bond strength of Fe atoms to CO* (Figure 13a,b). Moreover, a CoNi bimetallic atom catalyst with isolated CoN₄ and NiN₄ moieties was developed, which showed a high total current density > 74 mA cm⁻² with CO/H₂ ratios tunable from 0.23 to 2.26 in CO₂RR.^[104]

Figure 13. M1M2-N-C catalysts with dual-atom sites for CO2-to-CO conversion. a) Schematic atomic structure of CO2RR on the Fe site in FePc@NiNC, NiPc@FeNC, and NiNC/FeNC. b) Calculated free energy of CO2RR. c) Fourier transformation of the EXAFS spectra at R space of Ni/Fe-N-C. d) CO FE versus applied potential of Ni/Fe-N-C, Ni-N-C, Fe-N-C catalysts. e) Calculated free energy of CO2RR. f) The catalytic mechanism on diatomic metal-nitrogen site based on the optimized structures of adsorbed intermediates COOH* and CO*. a,b) Adapted with permission.[103] Copyright 2021, Elsevier Ltd. c–f) Adapted with permission.[101] Copyright 2019, WILEY-VCH.

In addition to isolated dual-atom sites, binuclear dual-atom sites can also be formed during catalyst preparation, for which the two central metal atoms are bonded to each other to form dual-metal atom pairs. Recently, a dual-metal-atom electrocatalyst comprising isolated diatomic Ni-Fe sites anchored on N-doped carbon (Figure 13c) was reported for CO₂ electroreduction.^[101] As shown in Figure 13d, the NiFe dual-atom catalyst (Ni/Fe-N-C) outperformed Fe-N-C and Ni-N-C in the potential range of -0.4– -1.0 V versus RHE, achieving a maximum CO FE up to 98% at -0.7 V. Theoretical results revealed that although bare Ni/Fe-N-C tended to be passivated by strongly bounded CO* intermediate, the energy barrier for the rate-determining step on CO-adsorbed Ni/Fe-N-C was significantly reduced in contrast with to bare Ni/Fe-N-C (Figure 13e). Based on which, the CO₂RR mechanism on the dual-metal-atom site was proposed (Figure 13f) that the neighboring NiFe dual-site was first passivated by strong bonding to CO* and then undergoing another CO₂ molecular reduction on the Fe site.

The Role of Nonmetal Moieties in M-N-C

In addition to single metal centers, nonmetal moieties in the carbon plane of M-N-C (e.g., N-doped sites and intrinsic defects) may also serve as active sites for CO₂RR. Over the past few years, metal-free N doped carbon catalysts have shown good performance for CO₂RR, and the pyridinic-N and/or graphitic-N defects were identified as the active sites.^[38–41] Some studies have also shown that the introduction of metal to prepare M-N-C catalysts can further increase the CO₂RR activity.^[45–48] However, the activity of N functionalities in M-N-C catalysts cannot be ignored. The role of N moieties on CO₂RR was studied by Strasser et al. through the preparation of a series of catalysts with the addition of secondary nitrogen precursors to polyaniline-based FeNC materials.^[105] They found that both Fe-N moieties and pyridinic N were likely active sites when comparing specific current density with X-ray photoelectron spectroscopy (XPS) data. Additionally, the role of pyridinic and pyrrolic N moieties in M-N-C with different central metals was further studied by near ambient pressure XPS (Figure 14a).^[106] It was revealed that these N-containing sites were not only the active sites for the catalytic reaction but also preferential adsorption sites for CO₂, which was dependent on the nature of the M-N₄ and M-N_xC_y moieties and the electronic structure of the carbon surface. According to Figure 14b, the preferential adsorption sites for CO₂ were quite different on M-N-C with distinct central metals. For example, CO₂ was preferentially adsorbed onto pyridinic N, M-N_xC_y and MN₄ sites on Co-N-C while onto M-N_xC_y sites on Ni-N-C. In contrast, there was no CO₂ adsorption observed on M-N_xC_y and MN₄ sites on Fe-N-C, only CO₂ adsorption on pyridinic N sites.

Figure 14. a) N 1s high-resolution X-ray photoelectron spectra of the Fe, Co, and Ni-N-C under ultrahigh vacuum (UHV) and 13.3 Pa CO2 atmosphere. b) Difference between XPS N 1s spectra of the Fe, Co, and Ni-N-C electrocatalysts in CO2 atmosphere and UHV with the associated structures, and schematic representation of the different preferential adsorption sites for CO2 as a function of M (M = Fe, Co, Ni). c) The values of UL(CO2)-UL(H2) for Fe-N4-N, Fe-N4-NH, 585-Fe-N4, and 585-defect, and view of slab models for *COOH adsorption on Fe-N4-N and 585-Fe-N4 site. a,b) Adapted with permission.[106] Copyright 2019, American Chemical Society. c) Reproduced with permission.[109] Copyright 2020, WILEY-VCH.

Furthermore, the role of intrinsic carbon defects in metal-free N doped carbon materials or M-N-C materials for CO₂RR has also been investigated.^[107–109] Wang and co-workers prepared an intrinsic defect-rich porous carbon embedded with atomically dispersed Fe-N₄ sites through high-temperature pyrolysis of carbon-rich carbon nitride with Fe salt. Compared to NG-SAFe (without intrinsic defect sites) and DNG (without Fe-N₄ site), the material consisting of rich intrinsic defects with the Fe-N₄ center (DNG-SAFe) reached outstanding performance with a maximum CO selectivity up to 90% and CO partial current density of 33 mA cm⁻². Consistently, the DFT calculation also indicated that the 585-Fe-N₄ (representing DNG-SAFe) owned the most positive value of U_L(CO₂)-U_L(H₂) compared to other structures (Figure 14c). Consequently, the intrinsic carbon defects instead of Fe-N₄ moieties in Fe-N-C were determined to be the active sites for CO₂RR.

The Role of Metal Nanoparticles

During the preparation of M-N-C catalysts under pyrolysis, metal nanoparticles (NPs) are inevitably formed. It is well known that exposed metallic species such as Ni and Fe are prone to produce H₂ in aqueous solution due to their low Gibbs free energy to adsorb hydrogen intermediate.^[110,111] Moreover, they may also dissolve to form ions under acidic conditions, not only destroying the structure of the M-N-C catalysts, but also competing for electrons with CO₂ reduction. In this situation, there exists a tradeoff between the high content of single metal atom sites and the formation of metal NPs for CO₂RR.

Recently, a few works studied the relationship between the metal loading and the overall CO₂RR performance.^[112–113] For example, a series of polymer-derived Ni-N-C catalysts with different Ni contents were prepared for CO₂RR to CO.^[112] As shown in Figure 15a, it can be observed that CO FE maintains a relatively high value (>80%) with Ni loadings in the range of 0–8 wt%, and drops slightly under high loading situation, which may be due to the appearance of Ni NPs. But the j_CO rises with Ni content up to ≈2 wt% and then keeps unchanged at 14 mA cm^-2 with formation of Ni NPs. These results suggest that the dispersed Ni atoms are responsible for the highly active CO₂RR in Ni SACs instead of Ni NPs, and the formation of Ni NPs under high Ni loadings did not influence CO₂RR performance too much as most of the Ni NPs were coated by a thin layer of carbon shells. The adverse influence of Ni NPs on CO₂RR was also studied by another work, which showed that excess Ni NPs led to enhanced HER.^[113]

Figure 15. The effect of the ratio of metal single atoms and metal NPs on CO2RR performance. a) CO FE and jCO at -0.8 V versus RHE for a series of Ni-PACN catalysts with varying Ni loadings (synthesized by changing amount of Ni nitrate added in the synthesis). b) Syngas proportion and total current density as a function of etching time. c) Calculation model and charge density difference graph of 3NGR@Ni (left: side view; right: top view). Yellow and blue represent the charge accumulation and depletion, respectively. d) Difference in limiting potentials for CO2 reduction and H2 evolution. e) A proposed reaction mechanism for CO production via ECR. a) Reproduced with permission.[112] Copyright 2020, WILEY-VCH. b) Reproduced with permission.[114] Copyright 2020, Elsevier B.V. c,d) Reproduced with permission.[120] Copyright 2021, WILEY-VCH. e) Reproduced with permission.[121] Copyright 2021, WILEY-VCH.

Acid washing is commonly adopted to remove unstable metal NPs during SACs synthesis. Recently, self-supported catalysts containing abundant single Ni atoms before (F-CPs) and after acid leaching (H-CPs) were prepared through a solid diffusion approach, and then their CO₂RR performances were studied.^[87] H-CPs exhibited a much higher CO FE over 90% across a broad potential range than the F-CPs, indicating inactive Ni NPs for CO₂RR. Furthermore, KSCN poisoning experiment was conducted to show that the exposed isolated Ni sites indeed mainly contributed to the excellent performance for CO₂ electroreduction to CO. In short, these results provide an insight that when preparing M-N-C electrocatalysts, the content of single-atom sites should be increased as much as possible along with avoiding the formation of metal nanoparticles to enhance the CO₂RR activity and at the same time suppress HER. Nevertheless, there comes another idea that M-N-C containing both the single metal atoms and nanoparticles can be directly applied for syngas synthesis via CO₂RR in an aqueous solution.^[114,115] Compared with CO₂ electroreduction, CO₂ and water coelectroreduction into syngas is more practical as syngas has been extensively used for industrial production of various hydrocarbons. A series of mesoporous N-doped carbon nanorods containing both single Ni atoms and Ni NPs were synthesized and used for controlled syngas production from CO₂ and water.^[114] The syngas with different ratios of CO to H₂ could be varied from 1:9 to 19:1 by changing the ratio of single Ni atoms to Ni NPs from 1:4.9 to 1:0 via adjusting the acid leaching time (Figure 15b).

Notably, there are also some works, which proposed that N-doped carbon encapsulated Ni NPs showed comparable CO₂RR performance to the single-atom Ni sites as the inner Ni NPs could tune the electronic structure of the outermost carbon layer.^[116–119] However, the existence of single Ni atom sites could not be completely excluded and the excellent electrolytic performance might actually come from the omitted single Ni atom sites. Very recently, we proved that in the Ni-N-C system consisting of Ni nanoparticles and single Ni atoms, single Ni atoms should be the real active sites for CO₂RR instead of N-doped carbon encapsulated Ni NPs through systematic experiments and theoretical calculations.^[120] DFT results indicated that the nitrogen-doped carbon-encapsulated Ni NPs were selective for the HER rather than for CO₂ electroreduction, and a synergistic effect between the encapsulated Ni NPs and the surface carbon layer was unlikely exist in the N-doped carbon-encapsulated Ni NPs catalysts with more than three carbon layers (Figure 15c,d).

Furthermore, a proton capture strategy by transition-metal nanoparticles adjacent to atomically dispersed Ni-N_x sites was proposed to accelerate proton transfer to the latter to improve CO₂RR, in which a hybrid Ni@NiNCM catalyst containing both Ni NPs and atomically dispersed Ni-N_x sites on supported carbon was prepared (Figure 15e).^[121] The states around the Fermi level showed that the Ni@NiN₄CM possessed a superior electron transport capability to NiN₄CM, thus leading to enhanced CO₂RR activity. This proton capture strategy was extendable to other NPs@TM-NC catalysts.

Summary and Outlook

Combining electrochemical CO₂ reduction with a renewable energy source to produce valuable chemicals/fuels offers a promising way to realize carbon recycle as well as alleviate the energy crisis. Recently, M-N-C catalysts have shown high efficiency for CO₂-to-CO conversion with high selectivity of up to 100%. In this review, regulation of the active center as well as the local atomic environments in M-N-C catalysts to adjust CO₂RR performance has been discussed in terms of experimental and theoretical findings. A direct view of the performance comparison of previously reported M-N-C catalysts with various central metal atoms is provided. Ni-N-C tends to be highly selective to produce CO across a wide potential range while Fe-N-C possesses a low overpotential for CO formation. Considering the trade-off between energy input and product yield, Ni-N-C and Fe-N-C are the most promising candidates for future electrochemical CO₂RR application at the industrial scale. Furthermore, the catalysts’ performance can also be controlled by tuning the coordination environment of the central metal atoms. But the intrinsic activity of the active sites in M-N-C catalysts are still controversial: some reports regarded MN₄ moieties as the active sites while others suggested a higher activity of unsaturated MN_x moieties; some studies believed that the Ni coordinated with carbon could function as the active site,^[122] and the intrinsic defect instead of the FeN₄ sites served as the active site for CO₂RR was also reported.^[109] In all, there are still lots of rooms to explore the potential of central metal atoms and coordination atoms to improve the CO₂RR. Here, some outlooks and challenges are proposed:

• (1)In terms of electrocatalysts, although the M-N-C catalysts have shown high selectivity up to >90% for CO formation, their activities corresponding to current densities still need to be further improved for large-scale applications. In this review, some strategies such as the MOF-assisted approach, in situ thermal transformation approach and solid diffusion method have been included, but more scalable strategies should be invented to synthesize catalysts in facile way and large scales. Besides, electrocatalysts with high surface area and dense single-atom sites are desired to maximize the amount of active sites accessible to reactant to further improve the activity (current density).
• (2)Developing in situ characterization techniques such as Raman spectroscopy, surface-enhanced infrared absorption spectroscopy, X-ray absorption and new theoretical calculation approaches are essential to understand the reaction mechanism and speed up the catalyst design. Many studies applied the “Computational Hydrogen Electrode (CHE)” model for DFT calculation. The proposed active sites sometimes, however, are actually not active or selective for CO₂ reduction according to the calculation results in a strict sense.^[123] Recently, Liu et al. introduced the surface charge and some layers of water into the simulation and applied ab initio molecular dynamics (AIMD) with the “slow-growth” method to analyze the kinetic barriers. We believe that in future, more rigorous calculation models like this should be widely used.
• (3)Design of flow cells with high mass transfer efficiency is equally significant for further industrial application, for which different architectures have been reviewed and widely applied for CO₂ electroreduction.^[124–126] Recently, there were some reports on M-N-C catalysts for large-scale CO₂ electroreduction at industrial current densities adopting the flow cell, which all achieved high CO partial current densities above 100 mA cm^-2.^{[69,72,88,127]} These results prove that M-N-C catalysts are highly promising for efficient CO₂ electroreduction to CO in industrial scale. In addition to selective CO formation, direct production of syngas (CO+H₂) that combines CO₂RR and HER, is also considered as a promising direction for future industrial application. However, the electrodes stability (>1000 h) for CO or syngas production in flow cell system is urgent to be overcome. Furthermore, the energy conversion efficiency of CO₂RR system should be considered.^[128] In the future, the optimization of the commercial CO₂-recycling system design is one of the major research goals, with voltage, energy utilization efficiency, and lifetime of the electrolyzer being taken into consideration.

Acknowledgements

This work was supported by the Fundamental Research Funds for the Central Universities (2019JQ03015), National Natural Science Foundation of China (42075169, U1810209), and the Beijing Municipal Education Commission through the Innovative Transdisciplinary Program “Ecological Restoration Engineering”.

Conflict of Interest

The authors declare no conflict of interest.