As a significant topic in the energy field, electrochemical CO₂ reduction (ECO₂R) has emerged as a promising strategy for reducing carbon emissions and building carbon-neutral energy infrastructures [1–3]. In spite of the great potential advantages, effective electrocatalysts for direct conversion of CO₂ to more value-added multi-carbon products (e.g., ethylene, ethane, ethanol, and 1-propanol) remain elusive due to the high overpotential and low productivity [4]. By contrast, the technologies of ECO₂R to CO have been well developed, and a wealth of electrocatalysts can achieve a Faradaic efficiency (FE) larger than 90% [5]. In this regard, there is a growing demand for further advancing the electrochemical CO reduction (ECOR) performance towards desired multi-carbon products [6–12]. Consequently, the design of an efficient ECOR catalyst with high activity, excellent stability, and eminent selectivity is of paramount importance and greatly demanded.

Recently, single-atom catalysts (SACs) featuring isolated atom active sites dispersed on substrates have been developed as promising candidates for the ECOR due to the ultimate atom utilization efficiency and desirable electronic states [13–17]. However, because SACs usually have strong linear scaling relationships between the adsorption strength of reaction intermediates, they still face many challenges in simultaneously enhancing the productivity and FE of the ECOR [18]. Furthermore, it is commonly accepted that effective coadsorption of multiple CO and active CO coupling are of great importance for achieving the production of multi-carbon products [19–21]. On the basis of those considerations, the construction of double-atom catalysts (DACs) by loading two atoms onto the substrate provides a new avenue for the ECOR, which not only retains the advantages of SACs but can also have synergetic interatomic interactions and more active centers. For example, Li et al. [22] synthesized a dual Cu-Cu atomic catalyst, which exhibited an outstanding ECOR activity for C₂₊ products, with FE of 91% and current density > 90 mA/cm². Rong et al. [23] showed that few Cu atoms anchored on graphdiyne can efficiently catalyze CO reduction to acetate with an unprecedentedly high FE of 53.8%. Through density functional theory (DFT) calculations, a series of DACs, such as Cu+B@g-C₃N₄, Cu₂@g-C₆N₆, TM+B@BlackP, B+P@C₂N [24–29], have been theoretically designed, which have low limiting potentials and small CO coupling barriers for generating multi-carbon products. However, it remains a great challenge to readily and rationally construct DACs with high catalytic performance toward the ECOR.

In recent years, various novel two-dimensional (2D) carbon-nitrogen systems, including g-C₃N₄, C₂N and g-C₆N₆, have garnered significant attention in numerous fields of energy catalysis owing to the large surface area, high stability, and rich coordination environment [30–34]. As a member of the carbon-nitrogen family, C₉N₄ monolayer possesses several desirable characteristics, such as high thermal and dynamical stability, nodal-line semimetallic property, and periodic nanopores, which is conducive to single/dual-atom catalysis [35]. For instance, Zhou et al. [36] theoretically found that C₉N₄ monolayer anchored with a single Ni atom could perform as a promising water-splitting electrocatalyst with N and Ni atoms as catalytic active sites for the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER), respectively. Xue et al. [37] conducted high-throughput calculations on TM@C₉N₄ to screen the potential nitrogen reduction reaction (NRR) catalysts, and discovered that W@C₉N₄ exhibited the best catalytic activity with the limiting potential of −0.24 V. He et al. [38] reported the B-doped C₉N₄ monolayer possessed good catalytic performance for the oxygen reduction reaction (ORR) with an overpotential of 0.98 V. Sun et al. [39] systematically investigated the NRR performance of dual-metal TM₁TM₂@C₉N₄ electrocatalysts, and obtained five high activity catalysts with favorable limiting potentials of −0.33 V to −0.53 V. However, to the best of our knowledge, there have been no reports of the ECOR on C₉N₄ based bimetallic electrocatalysts.

Hence, in this work, we perform the first-principles theoretical calculations to systematically explore the potential of C₉N₄ monolayers anchored with four types of transition metal dual atoms (TM₂@C₉N₄; TM = Fe, Co, Ni, and Cu) as the ECOR electrocatalysts because of the high catalytic activity and cost-effectiveness. We firstly examine the structural stabilities, CO adsorption abilities and CO coupling process of the TM₂@C₉N₄ systems, and then calculate their free energy diagrams for the ECOR toward C₁ and C₂ products. According to the computed limiting potentials (−0.06~−0.46 V) and kinetic energy barriers for the CO coupling process (0.54–1.08 eV), our designed systems except for Ni₂@C₉N₄ are confirmed to have good catalytic activities and high selectivity toward C₂ products, compared with the competing C₁ products and HER. These findings demonstrate that the bimetallic TM₂@C₉N₄ systems can act as high-performance catalysts for electrochemical reducing CO to value-added multi-carbon products.

The spin-polarized DFT calculations in this study are performed using the Vienna ab initio Simulation Package (VASP) [40]. The electron-ion interactions are described using the projected augmented plane wave (PAW) method [41], and the exchange-correlation energies are calculated using the generalized gradient approximation theory proposed by Perdew, Burke, and Ernzerhof (GGA-PBE) [42]. Moreover, the van der Waals weak interaction between the adsorbates and substrate is further treated using Grimme's semiempirical correction scheme (DFT-D3) method [43]. In all our calculations, the plane wave cutoff energy is set to 500 eV A vacuum layer of 20.0 Å is introduced along the z-direction to minimize the interaction between two periodic units. The structure optimization and electronic property calculations are performed using Γ-centered 2 × 2 × 1 and 4 × 4 × 1 k-point grids according to the Monkhorst-Pack scheme [44]. The convergence criteria for the total energy and maximum force per atom are set as 10⁻⁵ eV and 0.02 eV/Å, respectively. To obtain the transition states and kinetic barriers for the CO coupling reaction, the climbing-image nudged elastic band (CI-NEB) method is employed [45]. More computational details about the binding energy and ECOR are given in the Supporting information.

Firstly, the geometric structures and stabilities of the proposed TM₂@C₉N₄ monolayers are investigated. We construct a 2 × 2 × 1 C₉N₄ supercell consisting of 72 carbon atoms and 32 nitrogen atoms (Fig. 1a). After the structural optimization using GGA-PBE, the calculated lattice constants (a = b = 19.28 Å), C—C bond lengths (d_C-C = 1.42 and 1.46 Å), and C—N bond lengths (d_C-N = 1.34 and 1.39 Å) are found to be consistent with previous reports [46–48]. The pristine C₉N₄ monolayer has an intrinsic central cavity surrounded by six sp² hybridized nitrogen atoms, with a diameter of approximately 5.52 Å Similar to the cases of g-C₃N₄ and C₂N, these nitrogen atoms with lone pair electrons also provide suitable coordination sites to anchor single or dual atoms. Hence, four typical transition metals, including Fe, Co, Ni, and Cu, are taken into consideration for constructing DACs. The most stable configurations for each bimetallic atoms are determined by comparing the calculated total energies, and are given in Fig. 1b. Clearly, there exist three types of anchoring sites: (1) S1 configuration, such as Fe₂@C₉N₄, in which two Fe atoms are bound to three adjacent N atoms with d_Fe-Fe = 2.01 Å, d_Fe-N = 1.92 and 2.21 Å; (2) S2 configuration, such as Cu₂@C₉N₄, in which each Cu atom is only bound to two nearest N atoms with d_Cu-Cu = 2.25 Å and d_Cu-N = 1.91 Å; (3) S3 configuration, in which two metal atoms are bound to two adjacent N atoms, but the two TM-N bond lengths are not equal, such as Co₂@C₉N₄ (d_Co-Co = 2.09 Å, d_Co-N = 1.89 and 2.02 Å) and Ni₂@C₉N₄ (d_Ni-Ni = 2.16 Å, d_Ni-N = 1.84 and 1.92 Å).

To verify the stability of TM₂@C₉N₄ systems, we calculate the binding energies (E_b) of the Fe₂, Co₂, Ni₂, and Cu₂ dual atoms on C₉N₄ monolayers, and compare them to the cohesive energies (E_coh) of the corresponding bulk material. As presented in Fig. 1c, the E_b values of these considered double TM atoms anchored on C₉N₄ monolayer range from −6.11 eV to −4.56 eV, close to those of other porous carbon nitride nanosheets anchored with dual metal atoms [49–52]. This is indicative of a strong interaction between the TM₂ dual atoms and C₉N₄ monolayer. It should be noted that due to the significantly increased surface free energy, the dual metal atoms are peculiarly prone to aggregation of small clusters rather than being singly dispersed on supports during the reaction [53,54]. Fortunately, we can see that the E_b of all TM₂@C₉N₄ systems are more negative than the E_coh of respective bulk material, suggesting that the metal dual atoms tend to bind more strongly with the TM₂@C₉N₄ monolayers than with the metal bulk, thus potentially avoiding forming metal clusters. To gain further insight into the thermodynamic stabilities of TM₂@C₉N₄, we perform ab initio molecular dynamics (AIMD) simulations with a duration of 10 ps at 500 K. As presented in Fig. S1 (Supporting information), the total energies of these TM₂@C₉N₄ systems fluctuate within small ranges and no discernible damage to the geometrical structures occur after the 10 ps simulation. Moreover, the constructed DACs supported on C₉N₄ monolayers exhibit high durability, sustaining up to 500 K beyond common experimental conditions. These results indicate that our designed bimetallic TM₂@C₉N₄ systems have excellent stability and high feasibility in terms of various energy applications.

The electronic characteristics of electrocatalysts are of significant importance for the catalysis reaction. Thus, the band structures of TM₂@C₉N₄ systems are calculated and illustrated in Fig. S2 (Supporting information). We can see that all the proposed systems display metallic properties, which are favorable for facilitating electrocatalytic reactions. According to the charge density differences and Bader charge analysis (Fig. S3 in Supporting information), we find that the charge transfer from TM atoms to C₉N₄ monolayer with an amount of 1.46 |e| for Fe₂@C₉N₄, 1.30 |e| for Co₂@C₉N₄, 1.13 |e| for Ni₂@C₉N₄ and 1.14 |e| for Cu₂@C₉N₄, leaving the double TM atoms and C₉N₄ monolayers with positive and negative charges, respectively. In light of this, these metal dual atoms are in positive oxidation state, which would be conducive to the subsequent CO adsorption and activation. Overall, these facts ensure these proposed TM₂@C₉N₄ systems can act as potential ECOR electrocatalysts.

The initial step of catalytic reactions is typically driven by the adsorption of reactants, which can greatly influence or even dictate the following reaction mechanism. Thus, we then explore the adsorption performance of CO molecule, where three adsorption modes of (e.g., O-end, C-end, and CO-side) in conjunction with various possible adsorption positions of TM₂@C₉N₄ are considered to obtain the most energetically stable adsorption structures. As illustrated in Fig. 2. we find that the bridging sites of double TM atoms exhibit strong affinity for CO molecule with the C-end configuration, forming TM-C bonds with the lengths of approximately 1.90, 1.88, 1.86, and 1.85 Å for TM = Fe, Co, Ni, and Cu, respectively. Notably, compared with the C—O bond length (1.13 Å) of a free CO molecule, the adsorbed CO molecules (*CO) have elongated bond lengths, i.e., 1.19 Å for Fe₂@C₉N₄, 1.19 Å for Co₂@C₉N₄, 1.18 Å for Ni₂@C₉N₄, and 1.18 Å for Cu₂@C₉N₄. Furthermore, the adsorption energies (ΔE_*CO) of the CO molecules adsorbed on TM₂@C₉N₄ (TM = Fe, Co, Ni, and Cu) are calculated to be −2.03, −2.57, −2.12, and −2.42 eV, respectively. When considering the corrections of zero-point energy (ZPE) and entropy, the corresponding adsorption free energies (ΔG_*CO) of these *CO species are found to be −1.52, −2.02, −1.53, and −1.81 eV, respectively. These results demonstrate that our designed TM₂@C₉N₄ systems interact with the CO molecules through chemisorption, which are sufficiently strong to capture and activate the CO molecules.

Additionally, we further compute and analyze the projected density of states (PDOS) and charge density differences to understand the mechanism of CO activation. As shown in Fig. 2b, there are pronounced hybridization between the C-2p and TM-3d orbitals, indicating a strong coupling interaction. After CO adsorption, the magnetic moment of the Cu dual atoms is 0 μ_B, while the magnetic moments of the Fe, Co and Ni dual atoms are 3.80, 3.60 and 1.20 μ_B, respectively. As shown in Fig. S4 (Supporting information), the magnetic moments on the Fe₂@C₉N₄, Co₂@C₉N₄, and Ni₂@C₉N₄ systems can be attributed to the asymmetrical distributions of their respective 3d orbitals. Moreover, the charge density differences in Figs. 2c—f show that when the CO molecules are chemisorbed on the TM₂@C₉N₄ systems, electrons accumulate (yellow color) around the C and O atoms while deplete (cyan color) between the C—O bonds. It is indicative that the double TM atoms receive the lone-pair electron from the 5σ orbital of CO and simultaneously back-donate their electrons to the antibonding π* orbital of CO, forming the accepting (σ)-donating (π) interaction. Bader charge analysis shows that 0.44 |e|, 0.38 |e|, 0.34 |e|, and 0.27 |e| are transferred to CO from the Fe₂, Co₂, Ni₂, and Cu₂ dual atoms, leading to the gathering of negative charges on the *CO species. These observed results rationalize the strong adsorption strength and sufficient activation of the CO molecules on TM₂@C₉N₄, which would be of benefit to the subsequent ECOR.

After the thorough examination of the adsorption properties of CO molecules on the TM₂@C₉N₄ systems, we continue to investigate the subsequent ECOR toward various C₁ products. On the basis of the possible reaction paths for generating C₁ products shown in Fig. 3a, we explore the lowest-energy reaction paths for these TM₂@C₉N₄ catalysts, which are regarded as the ones with the least Gibbs free energy change (ΔG) between any two elementary reaction steps. The limiting potential (U), widely accepted index to assess the ECOR activity, is defined as the applied voltage required to make the whole reaction along the lowest-energy reaction path proceeds spontaneously. Consequently, we optimize the geometries of all ECOR intermediates in each elementary step and calculate their Gibbs free energies, which can be applied to establish the Gibbs free energy diagrams for the ECOR to C₁ products along the considered paths. According to the predicted free energy diagrams, the energetically most favorable reaction path and the corresponding U are clearly elucidated (Fig. 3 and Fig. S5 in Supporting information).

The results indicate that the ECOR process on the Fe₂@C₉N₄, Co₂@C₉N₄, and Ni₂@C₉N₄ catalysts prefer to proceed along the same 6e⁻ reduction pathway, in which CH₄ is identified as the final product. While for Cu₂@C₉N₄, it prefers to proceed via a 4e⁻ reduction pathway to form CH₃OH. The U values of respective most favorable reduction pathway are calculated to be −0.70 eV for Fe₂@C₉N₄, −1.07 eV for Co₂@C₉N₄, −0.73 eV for Ni₂@C₉N₄, and −1.16 eV for Cu₂@C₉N₄. Taking the Fe₂@C₉N₄ as a representative, the most favorable free energy profile for the ECOR to C₁ products is given in Fig. 3c for further clarification. One can see that the chemisorbed *CO is preferentially hydrogenated to generate the *CHO intermediate in the initial step, which is endothermic with the free energy rise of 0.48 eV Subsequently, the *CHO intermediate reacts with another two (H⁺ + e⁻) pairs to produce the *CH₃O intermediate, in which the Gibbs free energies fall by 0.19 and 0.07 eV, respectively. Furthermore, the fourth protonation step results in the dissociation of *CH₃O into *O and CH₄. Remarkably, this CH₄-releasing step is exoergic with a negative ΔG value (−0.38 eV), which is lower than that of the formation of *CH₃OH intermediate (ΔG = 0.27 eV). After this, the remaining *O can be readily hydrogenated into *OH with a downhill free energy of 0.38 eV Finally, the hydrogenation of *OH to *H₂O takes place after the sixth (H⁺ + e⁻) pair attack, which requires an energy injection of 0.70 eV Based on these above ΔG values, the last hydrogenation step is found to be the potential-determining step (PDS) for the ECOR to CH₄, having the maximum ΔG of 0.70 eV among all the elementary reaction steps, thus the corresponding U is predicted to be −0.70 V. As these double TM atoms anchored on C₉N₄ monolayers can serve as possible dual-active sites, we wonder besides the C₁ products, whether more valuable C₂ products can be produced through the ECOR on these TM₂@C₉N₄ systems. Therefore, we examine the potentiality of the ECOR toward C₂ products by firstly studying the performance for coinstantaneous adsorption of two CO molecules on our designed TM₂@C₉N₄ catalysts. As shown in Fig. 4a, both of the CO molecules are separately adsorbed on the top sites of TM atoms, forming the 2*CO species with two TM-C bonds, in which the distances between the two C atoms are in the range of 2.63 - 2.71 Å, while the TM-C bond lengths are 1.80 Å for Fe₂@C₉N₄, 1.72 Å for Co₂@C₉N₄, 1.73 Å for Ni₂@C₉N₄, and 1.79 Å for Cu₂@C₉N₄. The calculated adsorption energies of the 2*CO species on TM₂@C₉N₄ (TM = Fe, Co, Ni, and Cu) are found to be −2.94, −3.12, −2.99, and −2.30 eV, respectively. These results mean that the TM₂@C₉N₄ systems have sufficient capability to simultaneously capture both of the two CO molecules, which would be conducive to the subsequent CO coupling process.

Following the formation of two separately adsorbed *CO, we further study the extremely challenging process of the CO coupling on these TM₂@C₉N₄ catalysts, which is widely recognized as the vital step in the generation of C₂ products. Hence, based on the CI-NEB method, we compute the kinetic energy barrier to check whether the two chemisorbed *CO species can be successfully dimerized. It can be seen from Fig. 4b that our designed four catalysts exhibit favorable kinetics for the reaction 2*CO → *CO-dimer, since the transition state barriers are predicted to be 0.54 eV for Fe₂@C₉N₄, 0.83 eV for Co₂@C₉N₄, 1.08 eV for Ni₂@C₉N₄, and 0.64 eV for Cu₂@C₉N₄. And more notably, the obtained values are comparable to those of previous reported ECOR electrocatalysts, for instance, Cu+B@g-C₃N₄ (0.63 eV) [24], TM+B@BlackP (0.52 - 0.91 eV) [26], Cu₂@g-CN (1.16 eV) [29], Cu (211) surface (1.62 eV) [55], B@C₂N (0.73 eV) [56], Cu₂B₂ monolayer (0.41 eV) [57], and Cu₂Zn₂@C₅N₂H₂ (0.4 eV) [58]. Moreover, the conversions of two CO molecules into the *CO-dimer on TM₂@C₉N₄ (TM = Fe, Co, Ni, and Cu) are highly exothermic by 3.41, 3.04, 2.54 and 1.56 eV, respectively. Furthermore, as shown in Fig. 4a, there appear new C—C bonds in the *CO-dimer with the lengths of 1.41 Å for Fe₂@C₉N₄, 1.40 Å for Co₂@C₉N₄, 1.40 Å for Ni₂@C₉N₄ and 1.36 Å for Cu₂@C₉N₄, which are shorter than not only the C—C distances in the 2*CO species but also that of ethane (1.54 Å). These observations provide further evidence for the high feasibility and stability of the *CO-dimer formed through CO coupling. Consequently, our designed TM₂@C₉N₄ catalysts demonstrate the impressive potential to facilitate the generation of C₂ products, highlighting their prospects in sustainable energy conversion and storage systems.

To gain a more comprehensive understanding of the CO coupling, we further calculate the charge density differences and the charge transfer of the 2*CO and *CO-dimer on the TM₂@C₉N₄ systems. As shown in Fig. 4a, there is apparent charge accumulation (yellow region) on the 2*CO, while the charge depletion (cyan region) occurs on the substrate. Bader charge analysis shows that 2*CO on TM₂@C₉N₄ (TM = Fe, Co, Ni, and Cu) catalysts gain 0.51 |e|, 0.47 |e|, 0.38 |e|, and 0.32 |e|, respectively. After forming the *CO-dimer via the CO coupling, we find that not only TM₂@C₉N₄ donate electrons to the *CO-dimer but also the *CO-dimer back-donate electrons to the TM₂@C₉N₄, and there is more significant charge transfer (about 0.78 |e|, 0.70 |e|, 0.60 |e|, and 0.62 |e|, respectively), implying the strong interaction between the *CO-dimer and TM₂@C₉N₄ catalysts. These results rationalize the high possibility of forming stable *CO-dimer on the TM₂@C₉N₄ systems, which would have a significant influence on the formation of multi-carbon products.

Next, we explore the energetically most favorable reaction pathway for the ECOR toward C₂ products on these TM₂@C₉N₄ systems, as illustrated by the free energy diagram in Fig. 5, in which the ΔG between any two elementary reaction steps along the reduction path has the least value. When the C₉N₄ monolayers anchored Fe₂, Co₂, and Ni₂ dual atoms, the CO molecule can be ultimately reduced to C₂H₅OH product through a successive protonation of *CO-dimer, which proceeds along the following pathway: 2CO(g) → *CO-dimer → *COCHO → *CHOCHO → *CHOCHOH → *CHOCH₂OH → *CH₂CHO → *CH₃CHO → *CH₃CH₂O → *CH₃CH₂OH. It should be pointed out that the C₂H₅OH and C₂H₄ pathways differ at the step of *CH₂CHO hydrogenation, and their free energy changes are −0.19 vs. 0.20 eV for Fe₂@C₉N₄, −0.31 vs. 0.45 eV for Co₂@C₉N₄, and −0.23 vs. 0.67 eV for Ni₂@C₉N₄. Consequently, the C₂H₅OH pathway is followed. The PDS for all the Fe₂@C₉N₄, Co₂@C₉N₄, and Ni₂@C₉N₄ systems are the final protonation step of *CH₃CH₂O + (H⁺ + e⁻) → *CH₃CH₂OH, which are uphill by 0.06, 0.26, and 0.31 eV in their free energy profiles, respectively. Thus, the corresponding U values are calculated to be −0.06, −0.26, and −0.31 V, respectively.

As for the Cu₂@C₉N₄ system, the first protonation of the adsorbed *CO-dimer is more inclined to generate *CO*COH and the C₂H₅OH and C₂H₄ pathways differ at the step of *CH₂CHOH hydrogenation with the free energy changes of −0.47 and 0.11 eV, respectively. Hence, the C₂H₄ product is more likely to be produced via subsequent continuous protonation plus reduction process. And the whole reaction path is as follows: 2CO(g) → *CO-dimer → *COCOH → *CCO → *CHCO → *CH₂CO → *CH₂CHO → *CH₂CHOH → *CH₂CH → *CH₂CH₂. Based on the calculated reaction free energy diagram of the most favorable pathway, the PDS is identified to be *CHCO + (H⁺ + e⁻) → *CH₂CO, and the whole ECOR process only demands an energy injection of 0.46 eV That is to say, the U for generating C₂H₄ product on Cu₂@C₉N₄ is −0.46 V.

As discussed above, our proposed bimetallic TM₂@C₉N₄ systems have suitable U to achieve the electrochemical reduction of CO molecules into more valuable C₂ products, which are comparable with or even lower than those of the electrocatalysts reported in previous studies (Table 1), such as Ni+B@BlackP (−0.28 V for C₂H₆) [26], B + P@C₂N (−0.44 V for C₂H₄) [27], Cu₂@g-CN (−0.47 V for C₂H₄) [29], Cu₂B₂ (−0.59 V for C₂H₅OH) [57], Cu₂Zn₂@C₅N₂H₂ (−0.46 V for C₂H₅OH)^[58], and Cu@MoSi₂N₄ (−0.44 V for C₂H₆, −0.47 V for C₂H₅OH and −0.47 V for C₂H₄) [59]. These observations suggest that our proposed four electrocatalysts would possess high ECOR activities for producing valuable C₂ products rather than C₁ products.

As is well known, the HER would compete with the ECOR as the major side reaction and result in low FE, once the (H⁺ + e⁻) pairs are apt to participate in hydrogen evolution rather than CO reduction. In order to examine the ECOR selectivity, the hydrogen adsorption Gibbs free energy (ΔG_*H) of these TM₂@C₉N₄ systems are further investigated. As illustrated in Fig. 6a, the ΔG_*H values of the TM₂@C₉N₄ systems are calculated to be −0.24, 0.46, −0.20, and −0.77 eV for TM = Fe, Co, Ni, and Cu, respectively, and thus the corresponding U required to initiate the HER are −0.24, −0.46, −0.20 and −0.77 V, respectively. Noted that the ΔG_*H at other possible sites shown in Fig. S6 (Supporting information) have many positive values (Table S1 in Supporting information), indicating that the protons are unstably adsorbed. Furthermore, as shown in Fig. 6b, the differences between the U required for the ECOR and HER, i.e., U(ECOR) - U(HER), are found to be 0.18 V for Fe₂@C₉N₄, 0.20 V for Co₂@C₉N₄, −0.11 V for Ni₂@C₉N₄, and 0.31 V for Cu₂@C₉N₄, indicating an excellent suppressing effect on HER during the ECOR process except for Ni₂@C₉N₄. These results suggest that the C₉N₄ monolayers anchored with Fe_2, Co₂ and Cu₂ dual atoms would possess great potential as ECOR electrocatalysts for C₂ products with both of superior catalytic activity and high selectivity.

In conclusion, we devise four kinds of C₉N₄ based bimetallic electrocatalysts, designated as TM₂@C₉N₄ (TM = Fe, Co, Ni, and Cu), and subsequently evaluate their potentialities of ECOR toward valuable C₁ and C₂ products through the extensive DFT calculations. Our calculations show that the synergy effect of transition-metal dual atoms anchored on C₉N₄ monolayers enables the two chemisorbed CO molecules to couple into the crucial *CO-dimer intermediates both thermodynamically and kinetically. Moreover, based on the Gibbs free energy diagrams, all these TM₂@C₉N₄ systems are identified to have high ECOR activities toward C₂ products rather than C₁ products. C₂H₅OH formation is preferred on the Fe₂@C₉N₄, Co₂@C₉N₄, and Ni₂@C₉N₄ systems with rather low U of −0.06, −0.26, and −0.31 V, respectively, while the Cu₂@C₉N₄ system can catalyze the conversion of CO for C₂H₄ product with the U of −0.46 V. Meanwhile, these proposed TM₂@C₉N₄ electrocatalysts show well-suppressed HER ability except for Ni₂@C₉N₄, indicating the excellent selectivity for C₂ products. These findings not only accelerate the discovery of effective ECOR electrocatalysts for more valuable multi-carbon products but also contribute to developing C₉N₄-based DACs in the future.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the Science and Technology Research Project of Hubei Provincial Department of Education (No. D20212603), Hubei University of Arts and Science (No. 2020kypytd002), National Natural Science Foundation of China (No. 22303098) and Natural Science Foundation of Hubei Province (No. 2022CFC030).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2023.109050.