Electrocatalytic Reduction of CO₂ to C₁ Compounds by Zn-Based Monatomic Alloys: A DFT Calculation

Abstract

Electrocatalytic reduction of carbon dioxide to produce usable products and fuels such as alkanes, alkenes, and alcohols, is a very promising strategy. Recent experiments have witnessed great advances in precisely controlling the synthesis of single atom alloys (SAAs), which exhibit unique catalytic properties different from alloys and nanoparticles. However, only certain precious metals, such as Pd or Au, can achieve this transformation. Here, the density functional theory (DFT) calculations were performed to show that Zn-based SAAs are promising electrocatalysts for the reduction of CO₂ to C₁ hydrocarbons. We assume that CO₂ reduction in Zn-based SAAs follows a two-step continuous reaction: first Zn reduces CO₂ to CO, and then newly generated CO is captured by M and further reduced to C₁ products such as methane or methanol. This work screens seven stable alloys from 16 SAAs (M = Fe, Ru, Os, Co, Rh, Ir, Ni, Pd, Pt, Cu, Ag, Au, V, Mo, Ti, Cr). Among them, Pd@Zn (101) and Cu@Zn (101) are promising catalysts for CO₂ reduction. The reaction mechanisms of these two SAAs are discussed in detail. Both of them convert CO₂ into methane via the same pathway. They are reduced by the pathway: *CO₂ → *COOH → *CO + H₂O; *CO → *CHO → *CH₂O → *CH₃O → *O + CH₄ → *OH + CH₄ → H₂O + CH₄. However, their potential determination steps are different, i.e., *CO₂ → *COOH (ΔG = 0.70 eV) for Cu@Zn (101) and *CO → *CHO (ΔG = 0.72 eV) for Pd@Zn, respectively. This suggests that Zn-based SAAs can reduce CO₂ to methane with a small overpotential. The solvation effect is simulated by the implicit solvation model, and it is found that H₂O is beneficial to CO₂ reduction. These computational results show an effective monatomic material to form hydrocarbons, which can stimulate experimental efforts to explore the use of SAAs to catalyze CO₂ electrochemical reduction to hydrocarbons.

1. Introduction

A large amount of fossil energy is constantly consumed and the emission of carbon dioxide into the atmosphere increases dramatically, resulting in environmental problems and energy crises. Converting CO₂ into value-added products is an effective response to solve these problems. CO₂ reduction reaction (CO₂RR) achieves the purpose of energy storage through electricity, light, and heat catalysis, in which electrocatalytic reduction of CO₂ has attracted widely attention due to the advantages of accelerating the reaction rate and regulating the reaction process [1,2,3]. It is crucial to find cheap (easy to synthesize and rich in content), highly efficient (low voltage), highly selective (lower Faradaic efficiency of hydrogen evolution reaction (HER)), and stable CO₂ reduction electrocatalysts.

Since Sykes and co-workers first reported that the low energy barrier of adsorption and desorption for hydrogen on Cu by isolated Pd atoms led to a higher selectivity of Pd@Zn hydrogenation than pure metal in 2009 [4], single atom alloys (SAAs) have attracted extensive attention both in experimental and theoretical fields [5,6,7,8]. Different from nanoparticles and metal clusters, SAAs can produce many new characteristics. Sykes et al. [9] demonstrated that the SAAs strategy applied to Pt reduced the binding strength towards CO while maintaining the catalytic performance. SAAs showed high selective hydrogenation reactivity [6,10,11,12,13]. For example, ethylene is selectively hydrogenated rather than desorbed on Cu-based alloys [11]. The presence of Pd promoted the selective acetylene hydrogenation to form oligomers on Cu (111) [13]. The above experiments results and DFT calculations have proved the special activity of SAAs. However, the reaction sites are generally on the surface of SAAs, resulting in low utilization of noble metal substrate. Zn (101) is known for its reducing CO₂ to CO with high Faradaic efficiencies, rather than HER [14,15,16], which is the main basis for our proposal that the major component of the catalyst is Zn. Zn alone exhibits poor activity for producing methane by direct CO hydrogenation, but the activity is greatly enhanced when it is alloyed with other transition metals [17,18,19]. Therefore, we expect that CO₂ will first be reduced to CO by the surrounding Zn. While Zn cannot further reduce CO because of its much stronger binding with M than with Zn, the newly formed CO will be preferentially bound to the single atom alloy site. CO may be further reduced to hydrocarbons, if the CO reduction reaction can be favored versus HER.

At present, various efforts have been made to explore the selectivity of catalysts for CO₂ reduction. Egill Skúlason et al. [20] explained the principle that Cu electrodes gave a significant yield for particular hydrocarbons and alcohols, and indicated the selectivity of forming other products in the electroreduction process, by combining the DFT and rate theory. Through DFT and scanning tunneling microscopy (STM), Sykes and co-workers [21] used Cu as an example to explain the mechanistic differences in the selectivity of chemical products by different surface carving structures and alloy mechanisms. Inspiration from the calculations is used to identify new and improved catalysts for electrochemical CO₂ reduction. For highly reduced C₁ products such as methane and methanol, the catalyst selectivity can be effectively explained by analyzing the thermodynamics and kinetics of the *CO protonation step [22,23,24,25]. Often the hydrogenation of *CO (*CO → *CHO/*COH) was considered as one of the potential-limiting steps of CO/CO₂ reduction to methane [23,26,27]. Bell et al. [8] revealed that CO protonation was a potential determining step (PDS, the step of overcoming the maximum Gibbs free energy in thermodynamics) on M-doped silver or gold catalysts, and Rh@Au (100) and Rh@Ag (100) were more favorable for methane formation.

In this study, we performed a DFT-based study in which we screened 16 SAAs as electrochemical catalysts for CO₂RR. Those catalysts are composed of isolated single atoms embedded into the Zn (101), and these catalysts are denoted as M@Zn (101). The proposed electrocatalysts mainly consist of an alloy of Zn (101) and a small amount of M. The feedstock CO₂ is first reduced to CO by Zn (101), and is subsequently captured and further reduced to C₁ products by M. Beginning with the computational screening of 16 Zn-based SAAs stability, we search for alloys that exhibit high selectivity for CORR versus HER, as well as a relatively low application bias, enabling absorption CO to overcome the critical first reduction step. We then perform a detailed mechanical examination for the two of the most promising alloys, showing that both of them appear to be selective for methane formation. It gives an extended theoretical prediction of the selectivity roadmap for a wide range of Zn-based alloys. The inherent properties of the catalyst materials are closely related to their selectivity, providing a feasible design strategy for highly selective CO₂RR electrocatalysts.

2. Results and Discussion

2.1. Screening Stable SAAs

There are two kinds of doping sites on Zn (101), named “Low” and “High”, respectively. The formation enthalpy and surface energy of two SAAs sites are shown in Figure 1. The results show that IB elements is insensitive to the doping sites, while others tend to be doped at the “low” one.

Before studying catalytic performance, we check the structures of 16 Zn-based SAAs. The main substrate of the SAAs is Zn (101), in combination with an isolated single atom, M (M = V, Mo, Ti, Cr, Fe, Ru, Os, Co, Rh, Ir, Ni, Pd, Pt, Cu, Ag, Au,), replacing the surface Zn atoms. After structure relaxation, about seven elements (M = Pd, Pt, Cu, Ag, Au, Ir, Rh) are stably doped at the “low” site. To describe the electrochemical stability, the dissolution potential is calculated. When U_diss < 0 V, it shows that the catalyst is relatively unstable in realistic electrochemical environments. As shown in Figure 2, the negative H_f of SAAs (M = Pd, Pt, Cu, Ag, Au, Ir, Rh, Ni) indicate the high thermodynamic stability of these SAAs. It is noted that the calculated U_diss of Ni@Zn (101) is negative, suggesting it tends to dissolve in liquid electrolyte. Therefore, seven SAAs (M = Pd, Pt, Cu, Ag, Au, Ir, Rh) which meet the criteria of electrochemical and thermodynamic stability are further investigated on CO₂RR.

According to the definition of formation enthalpy, the stability of SAAs is related to the following three factors: the difficulty of dissociating the doped atom from the block, the dissociating energy of a Zn atom forming block and defect surface from the Zn (101) surface, and the possibility of binding doped atom to the defect surface. To gain insight into the stability of M@Zn (101), the Bader charge of M@Zn (101) is calculated. We find that charge transfer has little effect on ΔH_f, but the electron gain of M is a prerequisite for ΔH_f < 0 (Table S5). As shown in Figure 3, the formation enthalpy of VIIIB elements does not decrease significantly with the increase of the transferred charge, possibly because of the lanthanide elements [28], which makes it more difficult to dissociate an atom in sixth period than those in fifth period.

2.2. Identification of Catalysts with High Selectivity for CORR

The premise that a catalyst is selective to CORR but not to HER is ΔG_*CO < ΔG_*H [29,30], which means that the binding of the catalyst to *CO is tighter than to *H, *CO can be further reduced on the catalyst. The main attention is that blindly requiring small ΔG_*CO will lead to catalyst poisoning, which is not good for CORR. Therefore, as the second step in our investigation, ΔG_*CO, ΔG_*H, E_ads*CO, and E_ads*H are calculated for all 7 SAAs. As demonstrated in Figure 4a, it can be found that all SAAs is favorable for CORR. It is worth noting that Au and Ag cannot adsorb CO well as reported [31], so they will not be considered as potential CO₂RR catalysts. Figure 4b shows that when M is Pd, Cu, and Pt, the SAAs have moderate adsorption energy. Too strong adsorption means difficulty to desorb (such as Rh, Ir), while too weak to proceed to the next reaction (such as Au, Ag). Considering that the ΔG_*H of Pt@Zn (101) is close to zero, the following work will take Cu@Zn (101) and Pd@Zn (101) as examples to explore the mechanism of CO₂RR reaction in terms of thermodynamics.

To deeply understand the behavior of CO adsorption, the differential charge density diagrams of CO-Cu@Zn (101), CO-Pd@Zn (101) and Zn (101) in 3D forms are provided. As shown in Figure 5b–d, there are two parts with a significant increase in charge density, M–C and CO–2π*, suggesting that there is a strong interaction between M and CO. It is obvious that CO gains electrons when it binds to M, which is in consistent with the Bader charge calculation results (Figure 5a). It is also clear that for electron transfer Pd@Zn (101) > Cu@Zn (101), which corresponds for the adsorption of CO on Pd@Zn is stronger than that on Cu@Zn, while the adsorption capacity of CO on Zn (101) is weaker (Figure 4). The similar trend can also be found in other SAAs (Table S1). According to Bader charge analysis, when CO is adsorbed on M@Zn (101) (M = Rh, Ir, Pd, Pt, Cu), the electron transfer from M to CO is about 0.13–0.23 e. Comparing with clean SAAs, the binding with CO makes M losses about 0.2–0.33 e, and Zn gains electrons from a single doped atom to maintain stability. Based on the relationship between electron transfer and CO adsorption capacity, we find that with the enhancement of the adsorption ability of the SAAs to CO, the transferred electrons increase, which affects the catalytic performance of the CORR. This indicates that the presence of CO induces the charge transfer between M, Zn, and CO, some electrons donating to CO, and others stabilizing SAAs. SAAs provide electrons and act as electron storage when the CO₂ reduction reaction occurs.

To understand the reason why M@Zn (101) (M = Cu, Pd, Pt, Rh, Ir) favors CORR over the HER, we plot the scaling relations between E_*H–E_*CHO and E_*H–E_*COH. Figure 6 shows that the slopes for ΔG_*CHO–ΔG_*H and ΔG_*COH–ΔG_*H are 1.49 and 1.77, respectively. This indicates that when alloys bind strongly with H, they would bind with CHO and COH even more strongly, making CORR the major pathway. The larger slope for E_*COH/E_*H (1.77), compared to that for E_*CHO/E_*H (1.49), can be rationalized by the number of bonds formed between the COH and CHO and the surface; COH can form two bonds to the surface, while CHO can only form one [32].

2.3. Mechanisms of CO Reduction to Methane Catalyst by Cu@Zn (101) and Pd@Zn (101)

After identifying several promising SAAs for CORR, we then select two of them to investigate the reaction mechanism for further reduction to C₁ products. Cu@Zn (101) and Pd@Zn (101) are chosen from the Zn-based alloys for three reasons. First of all, the adsorption energy of *CO on these catalysts is 0.35 eV and 0.44 eV higher than that of H on them, which ensures most of the electrons flowing to CORR rather than HER, subsequently high selectivity for CORR. Secondly, the adsorption energy of *CO on the catalyst is relatively moderate, which is beneficial to the subsequent reaction. Finally, these two SAAs require a relatively small ΔG for the reaction *CO → *CHO. In addition, although the ΔG (step *CO → *CHO) of Pt@Zn is lower than that of Pd@Zn, the ΔG_*H of Pt is only 0.18 eV, and the competition of side reaction is fierce. Moreover, the synthesis method of Pd@Zn is known [33].

There are many possible CORR paths for generating CH₃OH or CH₄ from *CO. There are two ways to hydrogenate C and O in different environments. The first hydrogenation products are *COH and *CHO, while the second hydrogenation products may be *CHOH, *CH₂O (further formaldehyde formation), and *C + H₂O. The third hydrogenation products can probably be *CH₂OH, *OCH₃, and *CH + H₂O, and the next hydrogenation products perhaps are *CH₃OH (further methanol formation), *O + CH₄, and *CH₂ + H₂O, and the fifth hydrogenation products may be *CH₃ + H₂O and *OH + CH₄. The sixth hydrogenation product is likely CH₄ + H₂O. All possible pathways for Cu@Zn (101) and Pd@Zn (101) are taken into account. A large number of DFT studies have been devoted to the study of the mechanisms of CO₂RR catalyzed by SAAs [8,34,35,36,37]. Many pathways have been found in previous studies, but the difference is that the reaction intermediate formed by the *CO reduction are different. For example, An and co-workers [34] proposed that the CO₂RR pathway toward CH₄ on Fe@Cu (211) is via a *CO₂ → *COOH → *CO → *CHO → *CH₂O → *OCH₃ → *O + CH₄ → *OH + CH₄ → *H₂O+ CH₄ route, which is the same as the one supposed by Nørskov [23], and *CO → *CHO is the electric potential determining step as reported [27,38], while Nie supported *CO₂ → *COOH → *CO → *COH → *C → *CH → *CH₂ → *CH₃ → *CH₄ route on Cu (111) [39].

Our results show that CORR catalyzed by Cu@Zn (101) take the same pathway as the one proposed by Nørskov et al. in Cu (211) surface. The first (H⁺ + e⁻) pair is added to the carbon of *CO to form *CHO, the same step is followed by further hydrogenation to *CH₂O and *CH₃O. The fourth one is added to the carbon to liberate *CH₄, leading to the formation of *O, and the last two are added to the oxygen to form H₂O. (The free energy of all possible reducing intermediates is provided in Figure 7 and Table S2). It is worth mentioning, the ΔG of *CH₃O → *CH₃OH is 0.37 eV/0.36 eV, the ΔG of *CHOH → HCHO is 0.5 eV/0.7 eV, while *CH₂O → *CH₃O → *CH₄O has negative ΔG. Therefore, CH₄ as the C₁ product exhibits high selectivity. Figure 7 shows that for the *CO and *CHO intermediates, the adsorbates are more likely to bind to the top site of Cu. In contrast, *CH₂O and *CH₃O are more likely to bind to the Zn bridge site via O, *O prefers to combine in the hollow site formed by four Zn, and *OH prefers to bind to the top site of Zn. Analyzing the Gibbs free-energy diagram, it is found that the most uphill step (the potential-determining step) is the initial reduction step: *CO₂ → *COOH with ΔG = 0.70 eV. Each step involves only one single electron transfer, and when the applied bias is −0.70 V_RHE, all the steps become spontaneous. PdZn can exist stably, in which Pd, like a Cu pseudoelement, acts as an electron reservoir and has good performance in CO₂ reduction reaction [40]. The most favorable pathway for catalyzing CORR along Pd@Zn (101) is the same as that of Cu@Zn (101) and ends with the same C₁ product—methane. Moreover, the determining potential is −0.72 V_RHE. However, for the *CH₂O intermediates, the adsorbate is more likely to bind to the bridge site of Pd and Zn, and *OH prefers to bind to the top site of three Zn atoms. The highest step of Pd@Zn (101) is *CO → *CHO with ΔG = 0.70 eV from Figure 7, where the step of Cu@Zn (101) is only 0.42 eV. An implicit model is used to simulate the solvation effect of H₂O (Table S6). Solvation effect plays a critical role in stabilizing CO adsorption. It is found that PDS on Pd@Zn (101) and Cu@Zn (101) are the same as that in vacuum, but ΔG decreases to 0.67 eV and 0.44 eV, respectively, resulting in the CO₂RR is easier in solution. Moreover, ΔG of *O → H₂O also becomes smaller (Cu@Zn (101): −0.06 eV → −0.21 eV; Pd@Zn (101): 0.25 eV → 0.12 eV), indicating that *O is more likely to release the catalyst reaction site for the next cycle. This may enlighten the work in the CO₂RR experiments.

2.4. The Source of Catalytic Activity

In order to evaluate the M-C binding interaction from CO adsorbed on the SAAs, integrated-crystal orbital Hamilton population (ICOHP) analysis is performed by integrating the band states up to the highest occupied energy level to measure the bond strength. Moreover, the more negative the value of ICOHP, the stronger the bonding interaction. The results show that the ICOHP values change as follows: Ir (−4.07) < Rh (−3.52) < Pd (−2.85) < Cu (−2.46), and the bonding strength follows the order of Cu-O < Pd-O < Rh-O < Ir-O. The E_ads of *CO is positively correlated with the ICOHP values of the M-C. In other words, the smaller the ICOHP value, the smaller the E_ads value of *CO. On the one hand, this explains why there are obvious differences among different catalysts in the first step of activation. On the other hand, the ICOHP value can be used to describe the adsorption strength of the adsorbent on SAAs. To further gain insight into the catalytic performance of M@Zn (101) (M = Cu, Pd), the partial density of states (PDOS) of CO adsorbed on the catalyst was calculated and is shown in Figure 8.

It can be seen that the two catalysts have similar activation mechanisms for CO, and 5d atomic orbitals of Cu and Pd overlap with 5σ and 2π* molecular orbitals of CO, which makes the electrons transfer from Pd-5d to 2π*. This back-donation weakens the C≡O due to the existence of antibonding electrons. In addition, in the CO activated structure, the bond of C-O becomes longer (Table S3). Combined with COHP analysis, it is found that the orbitals in the z direction play a major role in the M-C adsorption: s/${\mathrm{d}}_{z^2}$/p_z of M atom and 2s/2p_z of C atom; the second is d_xy/d_yz of M atom and 2p_x/2p_y of C atom in the xy plane (Figure S3).

To further illustrate the special performance of single atoms compared with the slab, the PDOS of CO-Pd (111), CO-Pd@Zn (101), and CO-Zn (101) is calculated and illustrated in Figure 9. It can be seen that the main contribution of bonding is the interaction between the d orbital of Pd and the π* orbital of the CO. The overlap between Pd@Zn and CO is less than that between Pd and CO, but more than that on Zn (101). In combination with the ICOHP result, Pd (111) (−3.608) < Pd@Zn (101) (−2.846) < Zn (101) (−0.001), and the bond between single-atom Pd and C atom is weaker than that of Pd (111) and C. By calculating the adsorption energy, it is found that CO is strongly adsorbed on Pd (111), and the adsorption energy of −1.48 eV is stronger than that of −0.58 eV for Pd@Zn (101). However, the adsorption energy of 0.25 eV for Zn (101) can hardly adsorb and activate CO. This clearly demonstrates that a Pd single atom catalyst can alleviate the poisoning of Pd metal and enhance the CO adsorption of Zn (101). Therefore, the CORR activity of Pd@Zn (101) will be better than pure metal.

To explore the role of M@Zn (101) in the electrochemical CORR, taking Cu@Zn (101) and Pd@Zn (101) as examples, the charge transfer between the reaction intermediates and the catalyst is analyzed through Bader charge analysis along the reaction pathway. The studied system is divided into two parts including adsorbed C_xH_yO species and M@Zn (101) (M = Cu, Pd). As shown in Figure 10, it is found that the electrons gaining of intermediates is not quite consistent, due to the influence of the adsorption configurations and the type of M. Firstly, CO gains 0.1–0.2 e by adsorption on M@Zn (101), and SAAs lose the same number of electrons. As the reaction goes on, M@Zn (101) continuously donates electrons during the electrochemical CORR, behaving as an “electron provider”. After the formation of *CH₃O intermediate, the O atom prefers to be adsorbed on Zn, electrons transfer trend of Cu@Zn (101) is the same as Pd@Zn (101). Therefore, M@Zn (101) acts as an electron reservoir in the electrochemical CORR.

3. Computational Details

The DFT calculations is performed by the Vienna ab initio Simulation Package (VASP) [41,42]. The projector augmented wave pseudo-potentials and Perdew-Burke-Ernzerhof (PBE) [43] functional are used for the calculations. Using this functional, the calculated lattice constant of Zn is 2.64 Å, which is in agreement with the experimental value of 2.66 Å. The electronic wave function is converged to 10⁻⁵ eV, and the force is relaxed to less than 0.03 eV·Å⁻¹. The energy cutoff is set to 450 eV. The spin-polarization and z-direction dipole correction is used. A Gamma-centered grid of 3 × 3 × 1 is chosen to sample the reciprocal space for the slab calculations.

To prevent interactions between the periodic replicas along the z-direction, a vacuum separation of 15 Å between adjacent images is used. Supercells (4 × 4) of the slab configurations, with a thickness of five layers, are used to study the catalytic properties, where the top two layers were relaxed in the calculations.

The formation enthalpy, H_f, defined as ΔH_f(α) = E_M@Zn-slab − E_Zn-slab + n_Zn E_Zn − n_M E_M, is calculated, where E_M@Zn-slab and E_Zn-slab represent the total energies of M@Zn (101) and Zn (101), while E_Zn and E_M are the energy of one metal atom in the bulk state, respectively, and n is the number of atoms in the plate. The surface energy, γ = (E _M@slab − ∑n_slab E_bulk)/2A, is calculated. Here, A is the surface area, and n_slab E_bulk is the product of the metal atoms number in the SAAs and the atom energy in the block. The adsorption energy (E_ads) is defined as E_abs = E_TM-slab − E_TM − E_slab. Here, E_TM-ads, E_TM, and E_slab is the total energy of the adsorbate-substrate (TM-ads) complex, a gas molecule (TM), and the energy of the catalyst surface (slab), respectively. The dissolution potential, U_diss = U^θ − H_f/|ne|, is used to describe stability, where U^θ denotes the standard reduction potential of the bulk metal, H_f is the formation enthalpy, and n is the number of electrons transferred during dissolution. The Gibbs free energy change (ΔG) is calculated using the calculating hydrogen electrode (CHE) model developed by Professor Nørskov et al. [23], ignoring the coupling kinetic energy barrier between H⁺ and e⁻. For this work, they are calculated as:

$${\Delta \mathrm{G}=\Delta \mathrm{E}}_{\mathrm{DFT}}{+\Delta \mathrm{E}}_{\mathrm{ZPE}}+\int {\mathrm{C}}_{\mathrm{p}}\mathrm{d}T-T{\Delta \mathrm{S}+\Delta \mathrm{G}}_{\mathrm{pH}}$$

where ΔE_DFT is the reaction energy, which can be directly obtained from DFT calculations. ΔE_ZPE and ΔS are the free energy correction and entropy, respectively, and T is 298.15 K. Only the adsorbed intermediates take into account the vibrational entropy, which is calculated from the vibrational frequencies. Based on the calculated vibrational frequencies, the zero-point vibrational energy (ZPVE), and vibrational contributions to the internal energy and entropy at 298.15 K are calculated. ΔG_pH only depends on pH and cannot change PDS. To compare the catalytic performance of SAAs conveniently, pH = 0, ΔG_pH = k_BT × ln10 × pH = 0. In order to further analyze the influence of surrounding environment on the thermodynamic pathway, implicit model in VASPsol is used to deal with the solvation effect [44,45]. The dielectric constant of water is 78.4 in the calculations.

To evaluate the performance of the DFT method and the rationality of the used model, we first calculate the reaction of CO₂ reduction to CO on Zn (101) and Zn (002) crystal planes (Figure S1). The calculated CO reduction equilibrium potentials of CO reduction are −0.70 V and −1.22 V, respectively. These values match the those of −0.71 and −1.14 V obtained by previous works [14].

4. Conclusions

The Zn-based single-atom alloys (SAAs) are studied and evaluated as catalysts for the electrochemical reduction of CO₂ to high value-added products, such as alkanes and alcohols, using density functional theory (DFT) calculations. SAAs as a surface monatomic replacement of Zn (101) must contain a single atomic surface site of M (M = Fe, Ru, Os, Co, Rh, Ir, Ni, Pd, Pt, Cu, Ag, Au, V, Mo, Ti, Cr). Firstly, CO₂ is reduced to CO occurs on Zn host, and then CO preferentially binds M. We find that only seven of 16 SAAs are stable in electrochemical environments, and compared to Zn, the M prefers to bind to *CO rather than to *H, implying that that CO adsorption is more preferential than H adsorption for those SAAs. This suggests that, for these catalysts, CO₂ reduction is the main pathway and inhibits the competitive evolution of hydrogen. We also perform detailed thermodynamic studies of the reaction mechanism for the two most promising SAAs: Cu@Zn (101) and Pd@Zn (101). We find that both of them can reduce CO₂ to CH₄, and the reduction occurs via the same pathways. For Cu@Zn (101) and Pd@Zn (101), the reaction proceeds through the path: *CO₂ → *COOH → *CO(Zn) + H₂O → *CO(M) → *CHO → *CH₂O → *OCH₃ → *O + CH₄ → *OH + CH₄ → *H₂O+ CH₄. The potential determination step needs to overcome ΔG of 0.70 eV and 0.72 eV for Cu@Zn (101) and Pd@Zn (101), respectively. This suggests that Zn-based SAAs can efficiently reduce CO₂ to methane. It is worth mentioning that the potential determining step of Cu@Zn (101) is not *CO → *CHO as previously thought, but the first protonation of CO₂, which means that the presence of Cu greatly changes the CO₂ reduction properties of Zn materials. The solvation effect does not change the potential determination step, rather it reduces the ΔG of Cu@Zn (101) and Pd@Zn (101) to 0.44 eV and 0.67 eV, respectively, which is conducive to CO₂ reduction.

In summary, we forecast the Zn-based single-atom alloys, which selectively catalyze the CO₂ reduction to methane with high activity and selectivity. This indicates that the catalytic performance of CO₂ reduction reaction can be improved by introducing single atoms on the surface of Zn (101). Therefore, the characteristics of M can determine the performance of SAAs catalysts more than the surrounding metal, which inspires the design of CO₂ reduction materials. In addition, the limited ability of an M atom to bind CO molecule means that the SAAs will have better selectivity for C₁ product species but not forming multi-carbon compounds. We believe that this computational work will provide ideas for the design of CO₂ reduction electrocatalysts and encourage the exploration of SAAs in the field of CO₂ reduction.