Recent Progress in Two-Dimensional Materials for Electrocatalytic CO2 Reduction

Electrocatalytic CO2 reduction (ECR) is an attractive approach to convert atmospheric CO2 to value-added chemicals and fuels. However, this process is still hindered by sluggish CO2 reaction kinetics and the lack of efficient electrocatalysts. Therefore, new strategies for electrocatalyst design should be developed to solve these problems. Two-dimensional (2D) materials possess great potential in ECR because of their unique electronic and structural properties, excellent electrical conductivity, high atomic utilization and high specific surface area. In this review, we summarize the recent progress on 2D electrocatalysts applied in ECR. We first give a brief description of ECR fundamentals and then discuss in detail the development of different types of 2D electrocatalysts for ECR, including metal, graphene-based materials, transition metal dichalcogenides (TMDs), metal– organic frameworks (MOFs), metal oxide nanosheets and 2D materials incorporated with single atoms as single-atom catalysts (SACs). Metals, such as Ag, Cu, Au, Pt and Pd, graphene-based materials, metal-doped nitric carbide, TMDs and MOFs can mostly only produce CO with a Faradic efficiencies (FE) of 80~90%. Particularly, SACs can exhibit FEs of CO higher than 90%. Metal oxides and graphene-based materials can produce HCOOH, but the FEs are generally lower than that of CO. Only Cu-based materials can produce high carbon products such as C2H4 but they have low product selectivity. It was proposed that the design and synthesis of novel 2D materials for ECR should be based on thorough understanding of the reaction mechanism through combined theoretical prediction with experimental study, especially in situ characterization techniques. The gap between laboratory synthesis and large-scale production of 2D materials also needs to be closed for commercial applications.


Introduction
CO 2 is the main component of greenhouse gases which lead to environmental concerns, and its concentration has increased from approximately 280 ppm in the early 1800s to 410 ppm today [1][2][3]. Without proper strategies to combat this problem, the steadily growing CO 2 emissions will arouse the increase in the global average temperature, loss of glaciers, rise of sea level and other climate issues [4]. To reverse these negative courses, technologies such as carbon capture and storage (CCS) have been pursued [5][6][7]. Despite tremendous efforts on CCS, including separating CO 2 from air or flue gas, storing the captured CO 2 is still a daunting challenge due to the risk of leakage, massive energy consumption, high cost and social acceptance [8][9][10][11][12]. An alternative approach is to utilize CO 2 as raw material and convert it into value-added products, which approach has received extensive attention in the past decades [13][14][15]. Therefore, the conversion of CO 2 has been regarded as a promising route for closing the carbon cycle and producing value-added chemicals and fuels. than, for example, water-splitting reactions. As ECR reactions are normally carried out in aqueous solutions, another critical issue is the low solubility of CO 2 in water, which impedes the diffusion-controlled reactions. There are certain ways to improve CO 2 dissolution, including using nonaqueous electrolytes and increasing CO 2 partial pressure. For instance, some metal catalysts with low activity for ECR under atmospheric pressure, however, can reduce CO 2 to CO and HCOOH efficiently at high pressures [58].
Electrocatalysts can effectively reduce activation energy, accelerate reactions or increase desired product selectivity in ECR. It should also dissociate water in solution to promote proton-electron transfer because the proton-assisted multiple-electron transfer can be beneficial for CO 2 activation. Depending on the different pathways and the number of protons and electrons transferred, a range of products can be formed, including C 1 products, such as carbon monoxide (CO), methanol (CH 3 OH), formic acid (HCOOH), methane (CH 4 ), or C 2 products, such as ethylene (C 2 H 4 ), ethanol (C 2 H 5 OH) and others [59]. C 2 products with higher energy densities are more valuable, but the synthesis of C 2 products is more difficult than that of C 1 products because of larger numbers of required protons during conversion. Meanwhile, high formation energy of the C-C bonds can decrease the efficiency of the reaction. To achieve specific product selectivity, the design and synthesis of highperformance ECR electrocatalysts is essential.

Single-Electron vs. Proton-Coupled Electron Reactions
Research on ECR has been conducted ever since the early 19th century. ECR contains two half reactions occurring in the anode and the cathode, where different numbers of proton-coupled electron transfers occur. The standard electrode potentials of different reactions with reference to standard hydrogen electrons (SHEs) for different products in aqueous solutions are summarized in Table 1. For a typical single-electron ECR, CO 2 is reduced to the CO 2 anion radical (CO − 2 ) at the cathode and water is oxidized to oxygen at the anode. Since the kinetics of the reactions are quite sluggish, the first step for converting CO 2 into reduced carbon species is difficult. The thermodynamic potential for driving single-electron CO 2 reduction to CO − 2 is about −1.90 V vs. SHE at a PH of 7, indicating that this reaction is highly energetic and unfavorable [60][61][62]. In addition, this first step is also the rate-limiting step, and the generation of the CO − 2 intermediate plays a significant role in forming 2e − reduction products. Interestingly, a more favorable route that bypasses the formation of CO − 2 by proton-coupled electron transfer processes has been identified. The transfer of proton-coupled electrons benefits ECR within the potential range of −0.20 to −0.60 V vs. SHE, and the final products are determined by the choices of electrocatalysts and electrolytes as well as the numbers of electrons and protons transferred [63,64]. For instance, two-proton-coupled electron transfer in a CO 2 hydrogenation reaction mainly produces HCOOH and CO. As shown in Figure 1, ECR starts by transferring a proton-electron pair to form carboxyl (*COOH) or formate (*OCHO) intermediate species [53]. These two intermediates can be further reduced by accepting different numbers of proton-electron pairs. Note that there is much competition between intermediate species. For example, the OHCO* species can be converted to HCOOH, and the biformate (H 2 COO*) can also be reduced to HCOOH. Meanwhile, the different products also show similar reduction potentials. Thus, further reactions of these intermediate species are important for the reaction rate and formation of the final products. Moreover, as the equilibrium potential of a hydrogen evolution reaction (HER) (0 V vs. reversible hydrogen electrode (RHE)) is more positive than that of CO 2 reduction to HCOOH (−0.17 V vs. RHE), H 2 will be the by-product, which seriously limits ECR efficiency [65].
atalysts 2022, 12, x FOR PEER REVIEW 5 of 33 As shown in Figure 1, ECR starts by transferring a proton-electron pair to form carboxyl (*COOH) or formate (*OCHO) intermediate species [53]. These two intermediates can be further reduced by accepting different numbers of proton-electron pairs. Note that there is much competition between intermediate species. For example, the OHCO* species can be converted to HCOOH, and the biformate (H2COO*) can also be reduced to HCOOH. Meanwhile, the different products also show similar reduction potentials. Thus, further reactions of these intermediate species are important for the reaction rate and formation of the final products. Moreover, as the equilibrium potential of a hydrogen evolution reaction (HER) (0 V vs. reversible hydrogen electrode (RHE)) is more positive than that of CO2 reduction to HCOOH (−0.17 V vs. RHE), H2 will be the by-product, which seriously limits ECR efficiency [65]. Potential electrochemical CO2 reduction pathways to form the C1 products CO, HCOOH and CH4 and the C2 products C2H4 and C2H5OH. Reprinted from [53], Copyright (2019), with permission from Wiley.

Factors Determining Product Selectivity
Factors such as the type of electrocatalyst (morphology, composition, chemical state and crystallographic structure), type of electrolyte (composition, concentration and pH), temperature, pressure and applied potential all influence ECR. For instance, solid and aqueous electrolytes have been used in ECR, exhibiting remarkable differences in efficiency [66,67]. To date, NaHCO3 and KHCO3 solutions are frequently chosen as electrolytes because they can maintain the pH at the electrode surface. Meanwhile, the pH of the solution can affect the generation of by-products such as H2.
ECR, with a more positive potential vs. SHE, is thermodynamically more favourable. In this regard, the conversion of CO2 to alcohol or hydrocarbon products should be thermodynamically more favourable than to other products, including CO, HCOOH and HCHO (Table 1). Nevertheless, ECR not only needs to overcome the thermodynamic barrier but also the kinetic challenge determined by the concentration of available protons in reaction. Namely, the electrocatalysts prefer to transfer electrons from their catalytic sites to adjacent sites which can provide protons. Thus, the formation of hydrogenated C1 intermediates such as HCOOH can be kinetically more favourable than the formation of C-C bonds, which hinders the selectivity to C2 and higher hydrocarbon products. To date, only the Faradaic efficiency (FE) of C1 products CO and HCOOH can reach 100%, while the maximum FE of C2 product C2H4 is 80% and the maximum FE of C3 product C3H7OH is 30% [68,69]. Potential electrochemical CO 2 reduction pathways to form the C 1 products CO, HCOOH and CH 4 and the C 2 products C 2 H 4 and C 2 H 5 OH. Reprinted from [53], Copyright (2019), with permission from Wiley.

Factors Determining Product Selectivity
Factors such as the type of electrocatalyst (morphology, composition, chemical state and crystallographic structure), type of electrolyte (composition, concentration and pH), temperature, pressure and applied potential all influence ECR. For instance, solid and aqueous electrolytes have been used in ECR, exhibiting remarkable differences in efficiency [66,67]. To date, NaHCO 3 and KHCO 3 solutions are frequently chosen as electrolytes because they can maintain the pH at the electrode surface. Meanwhile, the pH of the solution can affect the generation of by-products such as H 2 .
ECR, with a more positive potential vs. SHE, is thermodynamically more favourable. In this regard, the conversion of CO 2 to alcohol or hydrocarbon products should be thermodynamically more favourable than to other products, including CO, HCOOH and HCHO (Table 1). Nevertheless, ECR not only needs to overcome the thermodynamic barrier but also the kinetic challenge determined by the concentration of available protons in reaction. Namely, the electrocatalysts prefer to transfer electrons from their catalytic sites to adjacent sites which can provide protons. Thus, the formation of hydrogenated C 1 intermediates such as HCOOH can be kinetically more favourable than the formation of C-C bonds, which hinders the selectivity to C 2 and higher hydrocarbon products. To date, only the Faradaic efficiency (FE) of C 1 products CO and HCOOH can reach 100%, while the maximum FE of C 2 product C 2 H 4 is 80% and the maximum FE of C 3 product C 3 H 7 OH is 30% [68,69]. Many theoretical studies proposed that for multielectron transfer reduction of CO 2 , the adsorption energies of intermediates follow linear scaling relationships, and breaking such linear scaling relationships can improve catalytic performance. Some strategies have been proposed, including reducing coordination numbers, doping with p-block elements, engineering oxyphilic sites and coating the catalyst surface with active ligands. Up to now, the majority of the ECR electrocatalysts have been metal-and carbon-based materials. It has been demonstrated that some heterogeneous electrocatalysts follow the two-electron-transfer mechanisms in CO 2 reduction to HCOOH or CO. For example, Hori et al. investigated a series of metals for ECR reaction and indicated that metallic electrocatalysts, such as Cd, Sn, Hg and Pb, prefer HCOOH formation [70]. However, noble metals, including Pt, Ag and Au, can convert CO 2 into CO. Copper-based electrocatalysts are the only electrocatalysts that have been reported for converting CO 2 into alcohols and hydrocarbons with good activity and selectivity. However, as mentioned above, the interaction between intermediates and electrocatalysts plays a crucial role in forming final products, demonstrating the importance of electrocatalysts.

Key Parameters for Evaluating ECR Efficiency
In experimental ECR study, there are several key parameters for evaluating electrocatalyst performance, including overpotential, current density, Tafel slope, FE, turnover frequency (TOF) and stability of electrocatalysts. The overpotential can be regarded as the energy needed to drive the reaction. In other words, the higher the overpotential, the higher the energy required for the reaction and the more difficult the reaction to be realized. Current density reflects the rate of electrochemical reaction, and the larger the current density, the faster the electrocatalytic reaction. The Tafel slope can be obtained by fitting the linear region of the Tafel curve. With the increase in overpotential, if the current density increases dramatically, the slope will be small. FE is regarded as the fraction between charges transferred to specific products and the total charges in the electrocatalytic process, accounting for the selectivity and efficiency of the reaction. Larger FE suggests less energy loss, and smaller FE indicates more energy loss. TOF is the catalytic activity of each site under a certain overpotential, indicating the intrinsic activity of catalysts. The stability of electrocatalysts can be evaluated by long-term use under a specific voltage.
In summary, the electrocatalytic reaction is a complicated multistep reaction occurring at the multiphase interface. Small overpotential and high current density can boost ECR, but it is difficult to satisfy them simultaneously. Developing excellent electrocatalysts to overcome these problems is the key.

Two-Dimensional Materials as Efficient ECR Electrocatalysts
Compared to bulk materials, 2D materials have a disordered state in their normal direction and a long-range order in the plane, showing a sheet structure. In addition, 2D materials can be exfoliated to single-or multiple-atom thickness, exhibiting interatomic covalent bonds in the plane. The physicochemical properties of 2D materials are different from that of bulk materials. The high atomic exposure rate, large specific surface area and flexible structure make them promising ECR catalysts. Some reported 2D-materialsbased electrocatalysts for ECR are summarized in Table 2 and are illustrated in detail in the following sections. Many metals (Ag, Cu, Au, Pt, Pd, etc.), graphene-based materials, Ni/Fe/Zn-doped nitric carbide, metal-doped TMDs and MOFs can produce CO. Particularly, single TM atoms embedded in 2D materials as SACs can exhibit great CO selectivity. Metal oxides and graphene-based materials can produce HCOOH. Interestingly, only Cu-based materials can produce CH 4 , C 2 H 4 and other C 2+ products.

Two-Dimensional Metallic Materials
Bulk metal catalysts in ECR have been investigated for decades [109][110][111]. Metal electrocatalysts can be divided into two groups: noble metal (Au, Ag, Pt, etc.) and other earth-abundant transition metals (Co, Ni, Cu, etc.). Theoretical work has well disclosed that noble metals are better electrocatalysts compared with earth-abundant metals. Nevertheless, some intermediates, such as CO, could poison the active sites, which affects their stability. Large-scale application of precious metals is also not possible due to their high price and scarcity.
It has been revealed that precious metals with 2D nanosheet structures can enhance their electrocatalytic performance effectively and improve the utilization of noble  [112]. The high performance could be attributed to nanosheet structure with more active sites, better electronic structure and larger surface area, as well as high atom utilization. Zhu et al. synthesized ultrathin Pd nanosheets for ECR, which exhibited an FE of 94% for CO production at a potential of −0.50 V vs. RHE [73]. Meanwhile, experimental and theoretical work revealed that ultrathin Pd nanosheets with high edge exposure ratios are beneficial for CO 2 adsorption and CO desorption as well as suppressing HER. Nanostructured Au also shows great activity in ECR. Mistry et al. studied the electrocatalytic activity of Au nanoparticles (NPs) in the range of 1-8 nm for CO 2 reduction to CO [113]. The results suggested that the current density increases with the decrease in particle size.
Moreover, multimetal composite nanosheets also exhibit excellent performance in electrocatalytic reactions because of the synergistic effect of different metals compared to single-metal electrocatalysts [114][115][116]. Wang et al. showed that a 2D Pd-Au bimetallic catalyst has lower activation energy than monometallic Au in ECR via density functional theory (DFT) calculation [74]. Very recently, Au-Cu bimetallic thin films with various ratios were synthesized for ECR [117]. The activity of the Au-Cu alloy was found to be correlated with the variation of electronic structures determined by alloy composition. As shown in Figure 2a, the surface valence bands of the Au-Cu catalysts present different patterns with different alloy compositions. The d-band center gradually moves away from the Fermi level with the increase in Au concentration (Figure 2b), which could influence the occupancy of antibonding states. Meanwhile, the interaction strength between substrate and intermediates decreases. The d-band center does not shift after ECR, implying great stability. Therefore, the catalytic performance for CO formation is enhanced as Au concentration increases, with the highest current density and FE at −0.80 V vs. RHE (Figure 2c,d). Nonprecious metals, including Fe, Co, Ni, etc., have also attracted much attention in electrocatalysis due to low prices, abundant resources and promising electrochemical properties as well as tunable electronic structures [118,119]. Studies have shown that Fe, Ni and Ti produce H2 as the major product in ECR due to their low HER overpotential Nonprecious metals, including Fe, Co, Ni, etc., have also attracted much attention in electrocatalysis due to low prices, abundant resources and promising electrochemical properties as well as tunable electronic structures [118,119]. Studies have shown that Fe, Ni and Ti produce H 2 as the major product in ECR due to their low HER overpotential and strong CO adsorption capability. Sn, Pb and In have poor binding energy with CO − 2 intermediates and the final products are usually formate or formic acid. Moreover, Cu is the only metal catalyst that can produce C 3 hydrocarbons by a C-C coupling mechanism [120]. Generally, modulating the electronic and surface structures of these metals could effectively enhance their electrocatalytic activity. Kuang et al. synthesized ultrathin Ni nanosheet arrays by an in situ topology reduction technique. In their study, the electrocatalyst showed excellent HER activity compared with Pt/C because ultrathin nanosheet structure could expose more active sites which boost electrocatalytic reactions [121]. In addition, 2Dstructured bismuth (Bi) exhibited more active sites and promising selectivity to formate compared to bulk Bi. Zhang et al. prepared ultrathin Bi by a liquid phase stripping technique. The ultrathin Bi displayed excellent electrocatalytic performance for formate generation with an FE of 86% and a current density of 16.5 mA/cm 2 at −1.1 V vs. RHE, which is obviously higher than bulk Bi and acetylene black decorated carbon paper (AB/CP) (Figure 3a,b). As presented in Figure 3c,d, CO 2 conversion to HCOOH underwent lower Gibbs free energy change on edges (003) and (012) [76]. Special surface structure can also improve the catalytic efficiency and selectivity. It has been reported that ultrathin porous Cu nanosheets synthesized by a simple replacement method can achieve an FE of 74.1% and current density of 23.0 mA/cm 2 at −1.0 V vs. RHE for CO 2 conversion to CO, remarkably better than that of traditional bulk Cu [77]. DFT calculations clarified that the porous Cu nanosheet structure can accelerate the formation of CO intermediate, thus promoting ECR efficiency. Similarly, Wang et al. synthesized Cu nanocubes which demonstrated better electrocatalytic activity and selectivity in comparison with Cu nanospheres of the same size. The nanocubes could reach an FE of 60% and a partial current density of 144 mA/cm 2 towards C 2 H 4 production [78]. Zhang et al. prepared the pipet-like bismuth (Bi) nanorods semifilled in nitrogen-doped carbon nanotubes (BiNRs@NCNTs) for ECR [122]. The catalyst acted as a nanoscale conveyor which could facilitate mass transport and reactant adsorption on active sites. As a result, the FE for formate generation reached 90.9% at a moderate applied potential of −0.9 V (vs. RHE).
Therefore, metal catalysts are effective for electrocatalysis due to their intrinsic activity and good conductivity. When the metal catalysts are synthesized with 2D nanosheets or other special nanostructures, the atom utilization, activity and selectivity could be improved, thus increasing electrocatalytic efficiency. Precious metals, such as Pd or Au, or multimetals with different ratios show excellent selectivity for CO generation. Nonprecious metals, such as Fe, Co and Ni, show poor ECR performance, while Sn, Pb and In can have good activity for the production of formate or formic acid. Particularly, Cu-based catalysts with different morphologies can exbibit different activities and selectivities for different products. Accordingly, the choice of metal, small nanosheet size and different morphology will have a positive effect on ECR activity and selectivity.

Graphene and Graphene-Based Materials
Generally, graphene is a single layer of graphite with a zero-band gap. It is a promising electrocatalyst for CO 2 reduction due to its high electron mobility, conductivity, unique electronic structure and large surface area. In addition, the high thermal conductivity can improve heat diffusion during the exothermic process, benefiting electrocatalysis. However, pristine graphene exhibits some drawbacks in ECR. For instance, the delocalized π bonding network of graphene negatively affects the adsorption of intermediates such as *COOH or *OCHO; therefore, graphene cannot effectively activate the CO 2 molecule and presents high-energy barriers for intermediate formation, leading to low ECR activity [123]. Some studies have clarified that the undoped zigzag edge of graphene shows a metastable adsorbed CO 2 state, and the energy barrier is quite high at about 1.3 eV [124]. As the modification of electronic structures can efficiently tackle these problems, many researchers have focused on tailing its electronic structures. For example, incorporating heteroatoms, such as B or N, can effectively modify graphene structures and decrease the CO 2 adsorption barrier. Studies have disclosed that N-doped graphene can exhibit a low overpotential of around 0.19 V for converting CO 2 into CO [75]. Duan et al. reported that doping boron can stabilize the negatively polarized O atoms of CO 2 and improve CO 2 chemisorption on carbon surfaces [125]. Similarly, Sreekanth et al. reported B-doped graphene as a metal-free electrocatalyst for CO 2 reduction in 0.1 M KHCO 3 solutions. They concluded that the presence of boron dopants can introduce spin density distribution and these atoms can be active sites for intermediate adsorption [79]. Some doped systems have been investigated in more detail by DFT calculations. Wu and coworkers systematically studied the effect of N-doped graphene for ECR. Three different N atom sites, including graphitic N, pyrrolic N and pyridinic N, were investigated as active sites for ECR. The authors suggested that the three different N sites can significantly decrease the free energy barrier for intermediate adsorption. The triple-pyridinic N can most effectively reduce the barrier for COOH adsorption, indicating that pyridinic N is a highly active site for converting CO 2 to CO [75]. However, there are different views on the mechanism of N-doped graphene for ECR. Chai et al. reported that the graphitic N-doped edges of graphene sites have low CO 2 activation barriers and are the most active sites for ECR among graphene-based materials [124]. Liu et al. demonstrated that a pyrrolic N site performs the best for CO 2 reduction to HCOOH with a low overpotential [126]. It has also been reported that N-doped graphene quantum dots with predominant pyridinic N at the edges can electrochemically convert CO 2 to C 2 compounds with low overpotentials [59].  [122]. The catalyst acted as a nanoscale conveyor which could facilitate mass transport and reactant adsorption on active sites. As a result, the FE for formate generation reached 90.9% at a moderate applied potential of −0.9 V (vs. RHE). Therefore, metal catalysts are effective for electrocatalysis due to their intrinsic activity and good conductivity. When the metal catalysts are synthesized with 2D nanosheets or other special nanostructures, the atom utilization, activity and selectivity could be improved, thus increasing electrocatalytic efficiency. Precious metals, such as Pd or Au, or multimetals with different ratios show excellent selectivity for CO generation. Nonpre- Compared with nonmetal dopants, introducing single or multiple metal dopants to construct special graphene nanostructures could also exploit materials with excellent catalytic performance. Au nanoparticles embedded in graphene nanoribbon exhibit low overpotential, high FE for CO generation and excellent stability compared to amorphous carbon-supported Au nanoparticles, attributed to the change in electronic properties and the increase in active sites [127]. Liu et al. synthesized Pd and Cu mono-and bimetallic nanoparticle-embedded graphene and suggested that 1 wt.% Pd-2 wt.% Cu/graphene had the highest ECR efficiency [80]. Su et al. prepared nickel-nitrogen comodified graphene (Ni-N-Gr) with more active centers by short-duration heat treatment and found that the Ni-N site was the active center for CO 2 reduction [81]. As shown Figure 4a,c, the Ni-N-Gr and the Ni foil exhibit obvious enhancement of current density in the presence of CO 2 . However, there is a remarkable difference in FE of CO. The Ni-N-Gr showed an FE of 90% at −0.7 to −0.9 V vs. RHE for CO production, where Ni foil produced mostly H 2 (Figure 4b,d), demonstrating that the synergistic effect of Ni and N is critical for improving CO selectivity.  Graphene-based composite materials also possess excellent conductivities and larger surface areas compared to pristine graphene. Huang et al. synthesized N-doped graphene monolayer-coated Sn foil, showing excellent flexibility with a high FE of 92% for formate at −1.0 V vs. RHE [82]. Lei et al. prepared Sn quantum sheets confined in graphene, offering more active sites for CO2 adsorption [83]. Li et al. successfully prepared SnS2 nanosheets supported on reduced graphene oxide for CO2 conversion to formate, displaying a low overpotential of 0.23 V and a maximum Faradaic efficiency of 84.5%. The spectroscopic and electrochemical characterizations suggested that the improvement of electrocatalytic performance can be attributed to the residual SnS2 [84].
For graphene-like materials, graphitic carbon nitride (g-C3N4) typically exhibits good chemical and thermal stability under ambient conditions. However, poor conductivity and less active sites limit its potential application in ECR. Strategies, including doping metal and constructing composite, have been developed to increase the activity of g-C3N4.
Metal atoms can effectively modulate the electronic structure of g-C3N4, thus improving the catalytic activity. The g-C3N4 nanosheet and multiwalled carbon nanotubes composite Graphene-based composite materials also possess excellent conductivities and larger surface areas compared to pristine graphene. Huang et al. synthesized N-doped graphene monolayer-coated Sn foil, showing excellent flexibility with a high FE of 92% for formate at −1.0 V vs. RHE [82]. Lei et al. prepared Sn quantum sheets confined in graphene, offering more active sites for CO 2 adsorption [83]. Li et al. successfully prepared SnS 2 nanosheets supported on reduced graphene oxide for CO 2 conversion to formate, displaying a low overpotential of 0.23 V and a maximum Faradaic efficiency of 84.5%. The spectroscopic and electrochemical characterizations suggested that the improvement of electrocatalytic performance can be attributed to the residual SnS 2 [84].
For graphene-like materials, graphitic carbon nitride (g-C 3 N 4 ) typically exhibits good chemical and thermal stability under ambient conditions. However, poor conductivity and less active sites limit its potential application in ECR. Strategies, including doping metal and constructing composite, have been developed to increase the activity of g-C 3 N 4 . Metal atoms can effectively modulate the electronic structure of g-C 3 N 4, thus improving the catalytic activity. The g-C 3 N 4 nanosheet and multiwalled carbon nanotubes composite have been shown to exhibit excellent stability and good electrocatalytic performance for CO 2 reduction to CO [85]. The electrocatalytic activity of the composite arises from the C-N bonds, and the high conductivity allows numerous electrons to transfer rapidly to the C-N sites. Guo et al. synthesized Co 3 O 4 -CDots-g-C 3 N 4 tricomponent electrocatalysts for syngas production [86]. They suggested that different catalytic components have different functionalities. For instance, g-C 3 N 4 and Co 3 O 4 provide active sites for ECR and HER, respectively, whereas CDots are the sites for proton generation.
It can be concluded that the electronic structure of graphene-based materials can be tuned to effectively improve ECR selectivity. Doping nonmetal or metal atoms is an efficient way for the design and synthesis of high-performance ECR catalysts. To date, nitrogen and metal atom codoping has been widely reported and shown excellent activity for CO production. Constructing highly active interfaces in graphene-based nanocomposite is also a good strategy in ECR catalyst design.

Two-Dimensional Transition Metal Dichalcogenides
Transition metal dichalcogenides (TMDs), such as MoS 2 , MoSe 2 and WS 2 , have the general formula MX 2 , where M is a transition metal atom and X represents a chalcogen atom. The layers of X-M-X are stacked together by the van der Waals interaction. This is another class of 2D materials with exceptional physical and chemical properties to be applied as electrocatalysts [88]. The large surface area of 2D TMDs can provide more active sites for reactions such as CO 2 reduction and water splitting [128,129]. DFT calculations suggested that the excellent catalytic properties for metal-terminated edges of 2D TMDs can be attributed to its metallic character and high d-electron density [87]. It has been confirmed that intermediates, such as COOH and CHO, prefer to adsorb on MoS 2 and MoSe 2 edges than the bridge S or Se atoms, and CO species adsorb selectively on the metal atoms during ECR [130,131]. Four different TMD materials, MoS 2 , MoSe 2 , WS 2 and WSe 2 , have been investigated as ECR catalysts; of which, WSe 2 nanoflakes exhibit the highest activity at an overpotential of 65 mV with an exceptional current density of 330 mA/cm 2 and an FE of 85% for CO generation, surpassing other electrocatalysts [89]. The excellent performance can be attributed to WSe 2 nanoflakes' much lower charge transfer resistance and low work function. DFT calculations showed that the formation of CO is kinetically favorable.
Further efforts have been made to enhance intrinsic activity, suppress the competing HER reaction and improve catalyst stability. It has been reported that doping and alloying different metal atoms with TMDs could enhance electrocatalytic efficiency by tuning the binding strengths of reaction intermediates and the reaction energy barrier. For instance, in one study, 5% niobium (Nb)-doped vertically aligned MoS 2 showed the largest CO 2 reduction current density of ca. 237 mA/cm 2 at the potential of −0.80 V, which was about two to three times higher than that of pristine and Ta-doped MoS 2 (Figure 5a). Nb-doped MoS 2 displayed an order of magnitude higher TOF for CO generation than pure MoS 2 at an overpotential of 20-150 mV and two orders of magnitude higher TOF for CO formation than Ag nanoparticles at the potential range of 0-650 mV (Figure 5b). The Nb atom could facilitate rapid CO release from the TMD edge, which remarkably increased ECR efficiency [87]. Theoretical calculation elucidated that there is a linear scaling relationship between the adsorption energies of the crucial intermediates, including COOH*, *CHO and *CO. It means that the stabilities of these intermediates are closely related, which causes a persistent overpotential for ECR. Theoretically, to improve ECR activity, breaking the linear scaling relationship between the binding energies of intermediates would be a good solution. Norskøv et al. have revealed that if the intermediates bind to different active sites, the linear scaling relation can be broken at the edge site of MoS 2 and MoSe 2 , which can increase ECR activity. They further showed that the doped S edge of MoS 2 can meet this requirement with *CO bonding on the doped metal atoms and *COH, *CHO and *COOH binding on the S sites [130]. Therefore, the binding energies of these three intermediates are higher than that of CO*, suggesting the breaking of the linear scaling relationship. Among four different TMD materials, MoS2, MoSe2, WS2 and WSe2, WSe2 exhibits good CO selectivity because of its lower work function and charge transfer resistance. Metal doping is also an effective way to improve ECR performance of TMD materials. More importantly, breaking the linear scaling relationship between the binding energies of intermediates will be a good strategy for designing excellent TMD catalysts.

MOF Materials
MOFs are a new type of porous material with metal ions or clusters coordinated with organic ligands. Due to their high porosity, large specific surface area and flexible structure, MOFs have been applied in electrocatalytic research recently [43,132]. However, some disadvantages, such as poor conductivity and less active centers, hamper their electrocatalytic activity. Studies have shown that low-dimensional MOFs can achieve high electrocatalytic performance because MOF nanosheets can expose more active atoms and exhibit excellent electron transfer. Moreover, the metal atoms, ligand and the connection of metal centers with ligands also play vital roles in the electrocatalytic activity of MOFs. Kornienko et al. synthesized a thin film cobalt-porphyrin MOF on a carbon substrate. The FE of this sample was 76% at −0.70 V vs. RHE for CO2 conversion to CO. In addition, it possessed a low Tafel slope of 165 mV/decade and the electrocatalytic stability was excellent [90]. A Zn MOF synthesized by ionic exchange of Zn nodes with adsorbed Ni salts achieved an FE of 71.9% for CO at 10.48 mA/cm 2 under the potential of −0.89V vs. RHE [91]. Zhang et al. prepared a series of transition-metal-based MOF electrocatalysts by using phthalocyanine (Pc), which can convert CO2 to CO efficiently [92]. Under controlled potential electrolysis tests of different samples in CO2-saturated solution, the Co-based MOF exhibited the lowest onset potential and largest current density (Figure 6a). Remarkably, Co-based MOF also showed the largest CO current density in a wide potential range ( Figure 6b). Moreover, Co-based MOF exhibited 99% FE of CO formation under the potential of −0.80 V, as shown in Figure 6c and d, demonstrating that Co-based MOF is the best catalyst for CO2 conversion to CO among others. DFT calculations suggested that the rate-limiting step for CO2 conversion to CO had the lowest Gibbs energy change ( Figure  6e). As shown in the inset of Figure 6e, the adsorption energy of *COOH on N and C sites was 0.95 and 0.90 eV, which is much larger than that of −0.1 eV on the Co site, indicating that metal center is the most stable adsorption site. Moreover, two linear relationships disclosed a volcano curve (Figure 6f) in which the location of Co-based MOF is closest to the volcano peak. Therefore, it shows the best intrinsic activity for ECR into CO. Among four different TMD materials, MoS 2 , MoSe 2 , WS 2 and WSe 2 , WSe 2 exhibits good CO selectivity because of its lower work function and charge transfer resistance. Metal doping is also an effective way to improve ECR performance of TMD materials. More importantly, breaking the linear scaling relationship between the binding energies of intermediates will be a good strategy for designing excellent TMD catalysts.

MOF Materials
MOFs are a new type of porous material with metal ions or clusters coordinated with organic ligands. Due to their high porosity, large specific surface area and flexible structure, MOFs have been applied in electrocatalytic research recently [43,132]. However, some disadvantages, such as poor conductivity and less active centers, hamper their electrocatalytic activity. Studies have shown that low-dimensional MOFs can achieve high electrocatalytic performance because MOF nanosheets can expose more active atoms and exhibit excellent electron transfer. Moreover, the metal atoms, ligand and the connection of metal centers with ligands also play vital roles in the electrocatalytic activity of MOFs. Kornienko et al. synthesized a thin film cobalt-porphyrin MOF on a carbon substrate. The FE of this sample was 76% at −0.70 V vs. RHE for CO 2 conversion to CO. In addition, it possessed a low Tafel slope of 165 mV/decade and the electrocatalytic stability was excellent [90]. A Zn MOF synthesized by ionic exchange of Zn nodes with adsorbed Ni salts achieved an FE of 71.9% for CO at 10.48 mA/cm 2 under the potential of −0.89V vs. RHE [91]. Zhang et al. prepared a series of transition-metal-based MOF electrocatalysts by using phthalocyanine (Pc), which can convert CO 2 to CO efficiently [92]. Under controlled potential electrolysis tests of different samples in CO 2 -saturated solution, the Co-based MOF exhibited the lowest onset potential and largest current density (Figure 6a). Remarkably, Cobased MOF also showed the largest CO current density in a wide potential range (Figure 6b). Moreover, Co-based MOF exhibited 99% FE of CO formation under the potential of −0.80 V, as shown in Figure 6c,d, demonstrating that Co-based MOF is the best catalyst for CO 2 conversion to CO among others. DFT calculations suggested that the rate-limiting step for CO 2 conversion to CO had the lowest Gibbs energy change ( Figure 6e). As shown in the inset of Figure 6e, the adsorption energy of *COOH on N and C sites was 0.95 and 0.90 eV, which is much larger than that of −0.1 eV on the Co site, indicating that metal center is the most stable adsorption site. Moreover, two linear relationships disclosed a volcano curve (Figure 6f) in which the location of Co-based MOF is closest to the volcano peak. Therefore, it shows the best intrinsic activity for ECR into CO. Recently, Perfecto-Irigaray et al. developed Cu-based MOF by doping metals (Zn, Pd and Ru) into HKUST-1 with different ratios [133]. They reported that the Ru-doped sample (10 wt.% Ru) showed an FE of 47.2% for methanol and ethanol generation, while other metal-doped samples performed worse compared to the pristine HKUST-1. Nevertheless, all doped samples exhibited increased ethanol selectivity. This is ascribed to the formation of C-C bonds on metal sites that boost the formation of C2 or C3 intermediates, which are finally converted into alcohol or other high-carbon products. Very recently, 2D copper  [133]. They reported that the Ru-doped sample (10 wt.% Ru) showed an FE of 47.2% for methanol and ethanol generation, while other metal-doped samples performed worse compared to the pristine HKUST-1. Nevertheless, all doped samples exhibited increased ethanol selectivity. This is ascribed to the formation of C-C bonds on metal sites that boost the formation of C 2 or C 3 intermediates, which are finally converted into alcohol or other high-carbon products. Very recently, 2D copper tetrahydroxyquinone conductive MOF (Cu-THQ) was reported for ECR at a low overpotential. Cu-THQ nanoflakes with an average particle size of 140 nm showed the largest current density of about 173 mA/cm 2 under −0.45 V vs. RHE, an average Faradaic efficiency FE CO of about 91% and a remarkable TOF of 20.82 s −1 . Spectroscopic characterization and DFT calculations uncovered the importance of reduced Cu + during ECR, which was reversibly oxidized to Cu 2+ after the reaction [134].
Overall, the metal atoms, ligand and the connection of metal center with ligand of MOF materials can determine electrocatalytic activity. Doping of a different metal to the original MOF materials, on the other hand, could either increase or decrease ECR activity.

Metal Oxide Nanosheets
Metal/metal oxide composites and oxide-derived metal have also been developed as ECR electrocatalysts with high performance [33,135,136]. Nevertheless, metal oxides still suffer problems such as instability and poor conductivity. It has been demonstrated that low-coordinated surface metal cations can be beneficial for CO 2 adsorption during ECR. First principles calculation revealed that both the density of states (DOS) and the charge density around the conduction band edges can be greatly enhanced [137]. Therefore, faster carrier transport along the 2D ultrathin layer during ECR can be expected. One excellent example is the ultrathin Co 3 O 4 layers, with thicknesses of 1.72 nm and 3.51 nm, synthesized by fast-heating treatment [93]. The ultrathin Co 3 O 4 layer exhibited higher DOS at the conduction band edges than the bulk structure (Figure 7a,b). Moreover, the charge density of the ultrathin Co 3 O 4 layer around the conduction band edge was larger and more delocalized than the bulk Co 3 O 4 (Figure 7c,d). These changes can be beneficial for faster carrier transport, thus accelerating ECR. Co 3 O 4 layers with 1.72 nm thickness displayed a maximum current density of 0.68 mA/cm 2 and FE of 64.3% for formate production at −0.88 V vs. the saturated calomel electrode (SCE) (Figure 7e,f). Figure 7g indicates that the electrochemically active surface area (ECSA) increased with the decrease in thickness from bulk to 1.72 nm, which clearly evidences the close relationship between structure and activity. The intrinsic activity of Co 3 O 4 with 1.72 nm thickness was over 1.5 times higher than the 3.51 nm thick Co 3 O 4 layers and 20 times higher than the bulk structure ( Figure 7h). Moreover, the authors elucidated an atomic-level relationship between oxygen vacancies and ECR performance by constructing a model of oxygen vacancies confined in atomic Co 3 O 4 single-unit cell layers [94]. DFT calculations uncovered that the main defect was oxygen (II), and X-ray absorption fine structure spectroscopy (XAFS) and X-ray photoelectron spectroscopy (XPS) indicated the distinct oxygen vacancy concentration. CO 2 adsorption isotherms demonstrated that the presence of oxygen (II) vacancy can promote CO 2 adsorption, while DFT calculations verified that oxygen (II) vacancy benefits spontaneous HCOO* desorption and prevents catalyst deactivation. More importantly, oxygen (II) vacancy facilitated the rate-limiting proton transfer step by stabilizing the HCOO* intermediate, which was confirmed by the drop of the activation energy barrier from 0.51 to 0.40 eV. This may also improve ECR kinetics because of the decreases in onset potential from 0.81 to 0.78 V vs. SCE and Tafel slope from 48 to 37mV dec −1 . Thus, oxygen (II)-vacancy-rich Co 3 O 4 nanosheets (0.84 nm) exhibited a current density of about 2.7 mA/cm 2 with 85% formate selectivity for 40 h, which is promising for an ECR electrocatalyst. The researches further prepared a Co-based nanosheet with a thickness of only four atoms (about 0.84 nm) via solvothermal growth. The partial oxidation of the four-atom-thick Co layer can be formed on its surface by controlling the reaction time [28]. Interestingly, the four-atom-thick Co sheets exhibited great potential in ECR because the thinner 2D structures presented more surface-active sites and higher electrical conductivity. In LSV experiments, the partially oxidized four-atom-thick Co layers displayed a distinct CO 2 reduction peak at −0.85 V vs. SCE with a current density of 10.59 mA/cm 2 . It was a performance enhancement by about 10, 40 and 260 times in comparison with Co four-atom-thick layer without oxidation, oxidized bulk Co and unoxidized bulk Co, respectively. More importantly, the partially oxidized four-atom-thick Co layers showed great stability with a current density of 10 mA/cm 2 over 40 h and a high FE for formate generation of about 90% under a low overpotential of 0.24 V. Such a noticeable improvement can be partly attributed to the increase in ECSA. Another reason is that the partially oxidized four-atom-thick Co layer (atomic Co 3 O 4 ) further increased their intrinsic activity and selectivity compared with pristine four-atom-thick Co layer.
Zhang's group prepared mesoporous SnO 2 nanosheets by a facile combination of hydrothermal reaction and calcination on carbon cloth [95]. As shown in Figure 8a,b, dense SnO 2 nanosheets grew uniformly on carbon cloth, and these nanosheets also contained numerous pores. The study showed a high partial current density of about 50 mA/cm 2 at −1.6 V vs. Ag/AgCl, with a high Faradaic efficiency of about 87% for formate generation, which outperformed most gas diffusion electrodes (Figure 8c,d). After electrolysis for 24 h, the current density and FE of formate were still stable (Figure 8e). Moreover, Figure 8f indicates that the current density also remained unchanged after folding or twisting the electrode 10 times. The excellent performance can be attributed to the following factors: Firstly, the SnO 2 nanosheet could provide intrinsically active sites for ECR; secondly, the highly porous structure could increase the specific surface area, which facilitated charge and mass transfer during ECR; the third reason is that the robustness of the hierarchical structures ensured high stability of the electrocatalyst during long-term use.
Therefore, metal oxide nanosheets mainly produce formate during ECR. The lowcoordinated surface metal cations can be active centers for ECR. Ultrathin metal oxides exhibit higher DOS at the conduction band compared to the bulk structure. The charge density is more delocalized, which benefits the carrier transport. In addition, building oxygen vacancy could also increase ECR performance by improving the reaction kinetics.

Two-Dimensional Materials Incorporated Single-Atom Catalysts
Single-atom catalysts (SACs) have enormous advantages in electrocatalysis in terms of 100% atom utilization and intriguing electronic structures [138,139]. The traditionally supported nanoparticles or clusters exhibit structural irregularities on the nanoparticle surface. The surface reactivity has a close relationship with surface free energy, and the ratio between fully coordinated surface atoms and the number of vacancies distinguishes the reactivity of atoms with identical chemical compositions but different positions [140][141][142]. Therefore, the conventional clusters with different surface reactivities can result in poor product selectivity. SACs are different from clusters because the surface free energy is homogeneous and the number of vacancies in the nearest neighbors is the same. Thus, SACs can show good stability and special intrinsic activity [143]. Single-atom catalysts exhibit uniform active sites as ideal catalysts; however, one major challenge for SACs is synthesis because SACs are unstable and can agglomerate quickly due to high surface energy. Furthermore, the supported structures in the neighboring environment increase the heterogeneity of SACs' active sites, and thus uniform activity cannot be easily achieved. However, 2D materials can be utilized as supports to improve the uniformity and performance of SACs in electrocatalytic reactions [144,145]. Studies have evidenced that 2D materials, such as MoS 2 , graphene and MXene, could stabilize the SACs and maintain their single-atom state [96,[146][147][148][149][150][151]. For instance, Back et al. investigated single transition metal atoms anchored on graphene with single or double vacancies as ECR catalysts by DFT calculations [152]. They suggested that many SACs exhibit high selectivity for ECR over the competitive HER because of favorable adsorption of *COOH or *OCHO over *H on the catalysts (Figure 9). SACs for CH 3 OH production, Ni and Pt, showed limiting potentials of −0.41 V and −0.27 V, respectively, while Os and Ru systems had the same limiting potential (−0.52 V) for CH 4 production. The activity improvement of SACs can be partially attributed to the unique electronic structure of the SACs and orbital interaction. Moreover, the authors elucidated an atomic-level relationship between oxygen va cancies and ECR performance by constructing a model of oxygen vacancies confined in atomic Co3O4 single-unit cell layers [94]. DFT calculations uncovered that the main defec  Therefore, metal oxide nanosheets mainly produce formate during ECR. The lowcoordinated surface metal cations can be active centers for ECR. Ultrathin metal oxides exhibit higher DOS at the conduction band compared to the bulk structure. The charge density is more delocalized, which benefits the carrier transport. In addition, building oxygen vacancy could also increase ECR performance by improving the reaction kinetics.

Two-Dimensional Materials Incorporated Single-Atom Catalysts
Single-atom catalysts (SACs) have enormous advantages in electrocatalysis in terms of 100% atom utilization and intriguing electronic structures [138,139]. The traditionally supported nanoparticles or clusters exhibit structural irregularities on the nanoparticle surface. The surface reactivity has a close relationship with surface free energy, and the ratio between fully coordinated surface atoms and the number of vacancies distinguishes the reactivity of atoms with identical chemical compositions but different positions [140][141][142]. Therefore, the conventional clusters with different surface reactivities can result in poor product selectivity. SACs are different from clusters because the surface free energy is homogeneous and the number of vacancies in the nearest neighbors is the same. Thus, SACs can show good stability and special intrinsic activity [143]. Single-atom catalysts exhibit uniform active sites as ideal catalysts; however, one major challenge for SACs is synthesis because SACs are unstable and can agglomerate quickly due to high surface energy. Furthermore, the supported structures in the neighboring environment increase the heterogeneity of SACs' active sites, and thus uniform activity cannot be easily achieved. However, 2D materials can be utilized as supports to improve the uniformity and performance of SACs in electrocatalytic reactions [144,145]. Studies have evidenced that 2D materials, such as MoS2, graphene and MXene, could stabilize the SACs and maintain their single-atom state [96,[146][147][148][149][150][151]. For instance, Back et al. investigated single transition metal atoms anchored on graphene with single or double vacancies as ECR catalysts by DFT calculations [152]. They suggested that many SACs exhibit high selectivity for ECR over the competitive HER because of favorable adsorption of *COOH or *OCHO over *H on the catalysts (Figure 9). SACs for CH3OH production, Ni and Pt, showed limiting potentials of −0.41 V and −0.27 V, respectively, while Os and Ru systems had the same limiting potential (−0.52 V) for CH4 production. The activity improvement of SACs can be partially attributed to the unique electronic structure of the SACs and orbital interaction.  Constructing a 3d transition metal-Nitrogen atom (M-N x ) system based on 2D materials for ECR is prevailing. Huan et al. investigated activity and selectivity of N-doped Fe-N 4 and Fe nanoparticles' moieties. They reported that samples with Fe-N 4 active sites had an FE of 80% for CO generation, but Fe nanoparticles could mostly only produce H 2 [97]. Zhang et al. synthesized Fe single atoms anchored on N-doped graphene and revealed that Fe-N 4 with an Fe 2+ oxidation state was the active site for ECR [98]. They also recognized that N dopants on the graphene substrate can be beneficial for ECR. However, there are some debates on the active sites and the corresponding ECR mechanisms on Fe SACs. For instance, Zhang et al. synthesized single Fe atoms anchored on N-doped graphene by prolonged thermal pyrolysis [99]. The SAC exhibited a high FE of about 97.0% for CO production under a low overpotential of 0.35 V. The excellent performance resulted from the presence of highly efficient dispersed Fe-N 5 active sites. Theoretical calculations disclosed that a fifth N atom coordinated to Fe 3+ to form Fe-N 5 moieties, where the additional axial pyrrolic N ligand further depleted the electron density of Fe 3d orbitals and reduced the π back-donation of Fe-CO bonding, which resulted in rapid CO desorption and high selectivity for CO formation.
Zhang et al. reported on single Fe atoms confined in carbon foams forming Fe-N 4 active sites [100]. However, they also detected an Fe 3+ state in their samples because Fe 2+ can be oxidized to Fe 3+ when the samples are exposed to air, but the difference was not distinguishable in ECR. Gu et al. investigated single Fe atoms loaded on N-doped carbon with an overpotential as low as 0.08 V for CO production [101]. They confirmed the persistent presence of Fe 3+ in the whole ECR process by in situ X-ray absorption near edge structure (XANES) measurements. Further studies indicated that Fe 3+ -N-C showed better performance for CO formation than that of Fe 2+ -N-C because of weaker CO absorption on the Fe 3+ sites. Yang et al. prepared a N-anchored Zn single-atom catalyst on carbon [102]. From HAADF-STEM characterization, it can be clearly seen that an individual single Zn atom was supported on carbon (Figure 10a), and extended X-ray absorption fine-structure (EXAFS) spectra demonstrated the existence of a Zn-N 4 moiety (Figure 10b). The catalyst could reduce CO 2 to CO with high selectivity and a high FE of about 95% under the potential of −0.43 V (Figure 10c). It also exhibited remarkable durability over 75 h without any FE CO decay (Figure 10d). Further experimental work and DFT calculations suggested that the remarkable activity can be attributed to the existence of Zn-N 4 , which was the main active site for ECR and had a low free energy barrier for the rate-limiting step of *COOH formation. Zhao et al. synthesized a single Ni atom loaded on graphene oxide. The catalyst showed an FE for CO production of 96.5% at a potential of −0.63V, with a TOF of 325.9/h [103].
Yang et al. prepared a single Ni atom catalyst supported on N-doped graphene. The TOF reached 14,800/h, and the FE of CO was 97% at a potential of −0.61 V [104]. The catalyst also displayed great stability with very stable electrocatalytic activity for 100 h. Liu et al. constructed a model SAC with precise structure using two-steps by linking Ni-TAPc to CNTs and by C-C coupling [105]. This model Ni SAC exhibited excellent activity, selectivity and durability in CO 2 conversion to CO. Experimental study suggested that Ni + in Ni-TAPc is highly active for CO 2 activation and is the catalytically active site. In addition, the study suggested that *CO − 2 + H + to *COOH is the rate-determining step. Gong et al. prepared a series of single-atom Ni catalysts (Ni-SA-N x -C) with different N coordination numbers by controlling the pyrolysis temperature [106]. Figure 11a,b clearly indicates the existence of isolated Ni atoms from TEM characterizations. The homogeneous distribution of Ni, N and C is demonstrated by energy dispersive spectroscopy (EDS) mapping (Figure 11c). It was found that the sample with Ni-N 2 moiety showed a higher current density than Ni-N 3 -C and Ni-N 4 in CO 2 -saturated 0.5 M KHCO 3 , indicating the best activity of NiSA-N 2 -C among all three SACs (Figure 11d). In addition, it exhibited a maximum FE CO of 98% at −0.8 V, which is also the highest (Figure 11e). The TOF for CO production on NiSA-N 2 -C reached 3467 h −1 at the potential of −1.0 V, which is much higher than those of NiSA-N 4 -C and NiSA-N 3 -C (Figure 11f), elucidating that the coordi-nation environment plays a crucial role in ECR performance. The current density and FE for CO generation were almost unchanged on the NiSA-N 2 -C catalyst after electrolyzation for 10 h at the potential of −0.8 V vs. RHE (Figure 11g), demonstrating excellent stability. DFT calculations suggested that the reaction path is proton-coupled electron transfer (Figure 11h). The low N coordinated single-atom Ni sites in NiSA-N 2 -C are beneficial for the formation of *COOH intermediate, thereby enhancing ECR activity (Figure 11i). on the Fe 3+ sites. Yang et al. prepared a N-anchored Zn single-atom catalyst on carbon [102]. From HAADF-STEM characterization, it can be clearly seen that an individual single Zn atom was supported on carbon (Figure 10a), and extended X-ray absorption finestructure (EXAFS) spectra demonstrated the existence of a Zn-N4 moiety (Figure 10b). The catalyst could reduce CO2 to CO with high selectivity and a high FE of about 95% under the potential of −0.43 V (Figure 10c). It also exhibited remarkable durability over 75 h without any FECO decay (Figure 10d). Further experimental work and DFT calculations suggested that the remarkable activity can be attributed to the existence of Zn-N4, which was the main active site for ECR and had a low free energy barrier for the rate-limiting step of *COOH formation. Zhao et al. synthesized a single Ni atom loaded on graphene oxide. The catalyst showed an FE for CO production of 96.5% at a potential of −0.63V, with a TOF of 325.9/h [103].  Han et al. designed a free-standing ultrathin 2D SAC by self-assembly of a Coporphyrin complex, which showed a CO production FE of 96% under an overpotential of 0.5 V [107]. They indicated that the improvement of activity was attributed to nitrogen coordination from the bottom layer, which increased the d z energy level of the Co atoms. Jiang et al. investigated a series of SACs supported on graphene nanosheets with different defect structures [153]. Ni SAC exhibited a CO formation FE of 95% under an overpotential of 0.55 V and excellent stability. Experimental study and theoretical calculations suggested that Ni sites with slightly larger vacancies benefit CO 2 conversion to CO by decreasing CO desorption barrier. Recently, Zheng et al. investigated ECR performance of Cu SACs on a N-doped graphene matrix with a highly exposed and coordinatively unsaturated Cu-N 2 center, as shown in Figure 12a [108]. Experimental work showed an onset potential of −0.33 V and FE for CO production of about 81% under a low potential of −0.50 V vs. RHE, obviously higher than the sample with Cu-N 4 moiety (Figure 12b). DFT calculations manifested that Cu-N 2 centers can facilitate CO 2 activation and accelerate electron transfer from Cu-N 2 sites to *CO 2 , which can greatly improve *COOH generation and the overall ECR activity. Jiao et al. investigated a single Cu atom catalyst anchored on g-C 3 N 4 as an electrocatalyst for CO 2 conversion to various hydrocarbons [154]. A slightly negative shift of photon energy demonstrated that N accepts extra charges from the Cu atom (Figure 12c). The authors reckoned that the d-orbital of Cu can be efficiently elevated by coordination with the g-C 3 N 4 framework, enhancing the adsorption of carbonaceous intermediates. Therefore, Cu-g-C 3 N 4 had better ECR activity with lower onset potential and exhibited a higher C 2 product rate than conventional Cu supported on nitrogen-doped graphene (Figure 12d).
( Figure 11c). It was found that the sample with Ni-N2 moiety showed a higher curr density than Ni-N3-C and Ni-N4 in CO2-saturated 0.5 M KHCO3, indicating the best tivity of NiSA-N2-C among all three SACs (Figure 11d). In addition, it exhibited a ma mum FECO of 98% at −0.8 V, which is also the highest (Figure 11e). The TOF for CO p duction on NiSA-N2-C reached 3467 h −1 at the potential of −1.0 V, which is much high than those of NiSA-N4-C and NiSA-N3-C (Figure 11f), elucidating that the coordinat environment plays a crucial role in ECR performance. The current density and FE for C generation were almost unchanged on the NiSA-N2-C catalyst after electrolyzation for h at the potential of −0.8 V vs. RHE (Figure 11g), demonstrating excellent stability. D calculations suggested that the reaction path is proton-coupled electron transfer (Figu 11h). The low N coordinated single-atom Ni sites in NiSA-N2-C are beneficial for the f mation of *COOH intermediate, thereby enhancing ECR activity (Figure 11i).  Single-metal-atom-loaded 2D materials are a class of very promising catalysts for ECR due to their high atom utilization and catalytic activity. Their activity is determined by the type of metal atom and the electronic structures of the metal atom. Furthermore, the coordination number of metal atoms can be tuned by the 2D materials' support, which can further optimize the activity and selectivity of the SACs.
as an electrocatalyst for CO2 conversion to various hydrocarbons [154]. A slightly negative shift of photon energy demonstrated that N accepts extra charges from the Cu atom (Figure 12c). The authors reckoned that the d-orbital of Cu can be efficiently elevated by coordination with the g-C3N4 framework, enhancing the adsorption of carbonaceous intermediates. Therefore, Cu-g-C3N4 had better ECR activity with lower onset potential and exhibited a higher C2 product rate than conventional Cu supported on nitrogen-doped graphene (Figure 12d).

The Design of Electrolyzers
Numerous ECR studies are based on H-cells because they can be used conveniently to evaluate the activity of catalysts due to their easy operation and adaptability to different electrode materials and configurations [155,156]. A typical H-cell electrolyzer contains two chambers that are separated by an ion exchange membrane. The counter electrode is in the anodic chamber, whereas the working electrode and reference electrode are in the cathodic chamber. During the ECR process, CO 2 is first dissolved in the liquid catholyte, such as KHCO 3 and NaHCO 3 , and then transferred to the surface of the working electrode where ECR occurs. However, the thick diffusion layer and low solubility of CO 2 in water limit CO 2 mass transport to the catalyst surface, impeding its practical applications [157].
Commercial ECR application requires continuous operation to realize high production efficiency. A microfluidic reactor for ECR has been designed where the electrolyte can continuously flow through the electrolyzer [158,159]. The cathode and anode are separated by a very thin channel that is filled with flowing electrolytes. CO 2 is supplied from the back of the cathodic gas diffusion electrode (GDE), and ECR occurs when CO 2 reaches the surface of catalysts. The operation parameters, such as the composition and pH of the electrolyte, can be adjusted, which promotes ECR rates. GDEs are high-surfacearea and porous electrodes. A typical GDE consists of a catalytic layer, a gas diffusion layer and a gas flow field. The structure and composition of GDE layers play a vital role in the transport of reactants and products because the transport processes influence the accumulation or depletion of intermediates on the catalysts, which determine ECR performance [160,161]. Lim et al. prepared Sn electrocatalysts with dense tips (SnDT) GDE, showing formate productivity of 65 mg h −1 . It was stable for 72 h without changes in FE or current density [162]. A 2D nanosheet Cu-loaded GDE achieved the highest partial current density of 131 mA cm −2 for acetate production [163]. Although high current density was realized, the microfluidic reactor still suffered product crossover, where the reduction of oxygen diffused to the cathode and the reoxidation of cathodic products diffused to the anode, decreasing the overall energy efficiency and productivity.
The membrane electrode assembly (MEA) electrolyzers have been exploited to solve problems of the microfluidic reactors. The catalytic layer of cathodic GDEs is in direct contact with the polymer electrolyte membranes. MEAs can decrease ohmic losses and prevent the poison of cathode catalysts by impurities through removing catholyte between the membrane and GDE. As a result, MEAs can exhibit excellent stability and high energy efficiency [164]. Gas and liquid products flow out of the cathode flow plate with little diffusion into catholyte through the membrane, resulting in high concentration liquid product. Cobalt phthalocyanine (Co-Pc) has been widely studied for ECR to CO in both immobilized and homogeneous forms and was also used in an MEA system with FE of CO >90% over 8 h at 50 mA cm −2 [165]. Gabardo et al. designed an MEA electrolyzer that achieved an FE of~80% for higher carbon products such as ethylene, ethanol, acetate and 1-propanol with current densities >100 mA cm −2 for 100 h [164].

Conclusions and Perspectives
Electrochemical reduction reactions are a promising strategy to convert CO 2 into valuable fuels and chemical products, which can offer potential technological solutions for energy problems and greenhouse effects. An effective electrocatalyst is the key to the potential deployment of commercial ECR processes. In this contribution, we review a series of 2D materials, including metal, graphene-based materials, TMDs, MOFs materials, metal oxide nanosheets and 2D materials incorporated with single atoms as ECR catalysts. Compared to the conventional metal-based electrocatalysts, 2D materials show unique physicochemical properties and electronic structures benefiting ECR activity.
Nevertheless, there are still challenges in the design and synthesis of 2D electrocatalysts for ECR. To date, most ECR catalysts can only produce C 1 products, especially CO, while producing more valuable products, such as C 2 and C 3 products, is desired. In general, the intrinsic activity can be modified in various ways such as doping, constructing special structures and morphologies, etc. Meanwhile, the choices of different classes of 2D materials are important for producing specific products. Strategies, such as constructing composite electrocatalyst materials and surface engineering of electrocatalysts, could be employed. Composite electrocatalyst materials could have different properties and catalytic performance in comparison with individual components due to synergistic effects. They could possess excellent properties for ECR such as better conductivity, high specific surface area, more active sites, better transport property and high durability. Another strategy is surface engineering of electrocatalysts by synthesizing nanotubes, nanosheets, nanospheres, isolated single atoms and defects. These properties are beneficial for effective contact for electrolyte ions and adsorption and desorption of intermediates. It is worth noting that single metal atoms anchored on the surface of 2D materials as SACs have rather high efficiency in ECR. However, the synthesis of SAC loaded on 2D materials is still a challenge, and key aspects, including suitable mononuclear metal complex precursors and stabilization of single metal atoms on the 2D materials, should be considered. Future research opportunities could be to develop new 2D materials such as MXenes and COFs.
Combining theoretical calculation and experimental work is necessary to obtain a fundamental understanding on the design and synthesis of 2D electrocatalysts. Although theoretical calculations are conducted with simplified models under idealized conditions, they could always be beneficial for revealing the reaction mechanism and providing guid-ance for design principles. For instance, breaking the scaling relationships between adsorption energies of different intermediates is of great significance to improve the activity of catalysts, which can be revealed from theoretical calculations. Similarly, to mitigate catalyst deactivation in ECR, it is necessary to understand the deactivation mechanisms. Theoretical models could be constructed to predict the intermediates which affect the stability of electrocatalysts. Finally, rapid development of characterization techniques, especially in situ spectroscopy, can greatly promote the development of catalysts of ECR. In situ spectroscopy can investigate catalytic reactions under reaction conditions and provide insights into the true active sites and reaction mechanisms. Through in situ techniques, such as infrared spectroscopy, Raman spectroscopy, X-ray absorption spectroscopy, X-ray photoelectron spectroscopy, mass spectrometry, etc., crucial issues in ECR, including identifying reaction intermediates, active sites and reaction pathways, can be resolved.
H-type cells are widely used in ECR. The current density of an H-type cell is limited to tens of milliamps, which is far from industrial application due to limited CO 2 solubility and energy loss. The design of microfluidic flow cells has made great progress recently. For example, the GDE and MEA electrolyzers can effectively tune reaction conditions and local environments in ECR. GDEs utilize a porous catalyst layer along with diffusion media, which can promote reactant transport. Thus, high current density can be realized, and transport losses can be effectively avoided. In the flow cell, higher carbon products are favored at high pH values, which can be hardly realized in an H-type cell. Flow cells and MEA make the separation of products convenient. These advantages could effectively accelerate the large-scale application of ECR at the industrial level.
It has been demonstrated that 2D nanomaterials possess great potential for ECR. Nevertheless, most of the 2D materials discussed in this review are still far from practical application. It is still demanding to synthesize 2D electrocatalysts with excellent performance at a large scale. Therefore, it is important to narrow the gap between laboratoryand commercial-scale production. Currently, some techniques, such as strip and salttemperature, have been applied in the large-scale synthesis of 2D materials, but scalable routes to prepare ultrathin nanosheets remain to be developed.