Quasi‐Copper‐Mers Enable High‐Performance Catalysis for CO2 Reduction

Abstract

As the atmospheric carbon dioxide (CO₂) level keeps hitting the new record, humanity is facing an ever‐daunting challenge to efficiently mitigate CO₂ from the atmosphere. Though electrochemical CO₂ reduction presents a promising pathway to convert CO₂ to valuable fuels and chemicals, the general lack of suitable electrocatalysts with high activity and selectivity severely constrains this approach. Herein, a novel class of electrocatalysts is investigated, the quasi‐copper‐mers, in which the CuN₄ rather than Cu atom itself serve as the basic building block. The respective quasi‐copper‐monomers, ‐dimers, and ‐trimers hosted in a graphene‐like substrate are first synthesized and then performed both experimental characterization and density functional theory (DFT) calculations to examine their atomic structures, evaluate their electrocatalytical performance and understand their underlying mechanisms. The experimental results show that the quasi‐copper‐trimers not only outperform the quasi‐copper‐dimer and quasi‐copper‐monomer when catalyzing CO₂ to CO, it also shows a superior selectivity against the competing hydrogen evolution reaction (HER). The DFT calculations not only support the experimental observations, but also reveal the volcano curve and the physical origin for the qausi‐copper‐trimer superiority. The present work thus presents a new strategy in the design of high‐performance electrocatalysts with high activity and selectivity.

Introduction

Over the past two centuries, the global atmospheric average carbon dioxide (CO₂) concentration has been steadily rising, and it has reached an unprecedented level of 420 ppm in April 2022. This has caused the climate change and even-worsening environment, which are among the most threatening challenges faced by humanity in the 21st century.^[1] Developing a cost-effective catalytic process for electrochemical CO₂ reduction reaction (CO₂RR) has been considered as a feasible solution to overcome this big challenge. Cu and Cu-based materials stand out as the most promising electrochemical catalysts for CO₂RR, because of their unique capability in producing a broad range of hydrocarbon fuels and chemicals at significant rates.^[2] Cu was among the first electrocatalysts reported that can catalyze CO₂ to CH₄ and C₂H₄.^[3] However, unmodified Cu exhibits rather poor selectivity and activity for formate production. Both entity size and morphology were found to play significant roles. For example, Cu nanocubes can substantially enhance the catalytic activity and selectivity for CO₂ reduction, compared to Cu nanospheres of similar particle sizes, and remarkably they can reach a Faradaic efficiency of 60% and a partial current density of 144 mA cm⁻² toward C₂H₄ production.^[4] Cu nanoparticles of sizes below 5 nm have shown a dramatic increase in the overall catalytic activity and the selectivity for certain hydrocarbons (methane and ethylene).^[5] Owing to the unique electronic structure and high atom utilization, single-atom catalysts (SACs) have recently demonstrated unprecedented activity and selectivity for CO₂RR, which holds great promise in reducing carbon emission and storing renewable energy.

The pioneer work on electrochemical CO₂ reduction by SACs can be traced back to 1974 when Meshitsuka et al. reported that Co and Ni phthalocyanines on graphene substrate were active in CO₂ electroreduction.^[6] Since then, metal–organic complexes, normally featured as M─N₄ sites, have been widely investigated as a catalyst for electrocatalytic CO₂ reduction with enhanced activity and selectivity.^[7] The M─N₄ CO₂RR catalysts were first reported in 2015, by Strasser et al. who demonstrated that Ni and/or Mn SACs catalysts were active and highly CO-selective in CO₂RR to CO/H₂ mixtures, outperforming a low-area polycrystalline gold benchmark.^[7b] Following on, this research field has been expanded rapidly, and several new single metal sites (such as Sn,^[8] Sb,^[9] Bi,^[10] Mo,^[11] Cu^[12] et al.) have been identified and different reduction products have been achieved, for example including CO,^[8,10] formate,^[11] methanol,^[12] ethanol^[13] et al. These atomically dispersed metal sites usually act as the main active centers in the reported electrocatalysts; the coordinated atoms (mostly C and N) around the metal sites, on the other hand, are believed to facilitate the CO₂ activation or the dissociation of the intermediates.

Although SACs possess impressive CO₂RR-catalyzing ability and selectivity, their further modifications are needed in order to achieve an even higher CO₂RR performance toward practical applications in terms of activity, selectivity, energy efficiency, as well as long-term stability. Motivated by the effectiveness of the paired or quasi-paired metal atoms that were found to significantly enhance the electrocatalytical performance,^[14] herein we extended the Cu SAC to the quasi-copper-mer catalysts, in which the CuN₄ rather than Cu atom serves as the basic building blocks. Cu atoms are selected in the present study because the Cu is one of the most demonstrated SAC for CO₂RR, and has been proven to be able to produce C1, C2, and C3 products.^[13,15] N atoms around Cu atoms are believed to facilitate the CO₂ activation or the dissociation of the intermediates. We first synthesize three different types of quasi-copper-mer catalysts: quasi-copper-monomers, quasi-copper-dimers, and quasi-copper-trimers hosted in a graphene-like substrate and perform experiment characterization and then the DFT calculations to examine their atomic structures, evaluate the catalytical performance and understand their underlying physical mechanisms. Our results show that the quasi-copper-trimers outperform quasi-copper-monomer and quasi-copper-dimer such that they give the best activity for CO₂RR to CO and the best selectivity against the competing hydrogen evolution reaction (HER). By combining the DFT calculation and experimental results, we not only show a promising catalyst for CO₂RR, but also demonstrate a practically viable route in the design of new catalysts going beyond of single-atom catalysts.

Results and Discussion

Structure Characterization

In order to synthesize the representative quasi-copper-mer samples for the investigation of their CO₂RR performance, an appropriate substrate is essential in order to provide high population of neighboring anchor sites for Cu atoms, stabilize the Cu atoms, and prevent their aggregation. Therefore, a heavily N-doped carbon substrate was purposely synthesized through the carbonization of C₃N₄. The as-synthesized C₃N₄ exhibits a twisted thin sheet morphology and its X-ray diffraction (XRD) peaks can be well indexed to the (100) and (002) planes of g─C₃N₄,^[16] which proves the successful synthesis of g─C₃N₄ (Figure S1a,b, Supporting Information). Upon the heat treatment, the corresponding XRD peaks of C₃N₄ disappeared. Instead, a widened carbon peak at ≈2θ of 25° occurred in all three samples. This suggests that C₃N₄ had been fully transferred to N-doped carbon. The CHNS elemental analyzer results also show C₃N₄ consists of 34 wt.% C and 61 wt.% N, while the C content in N-doped carbon is ≈57 wt.% and N content is ≈31 wt.%, which are due to the decomposition of C₃N₄ to N-doped carbon and N₂ at high temperature. In the corresponding Raman spectrum of the final products (Figure 1a), all three samples exhibited a 2D Raman signal at ≈2670 cm⁻¹, which is characteristic of graphitic sp² hybridized carbon.^[17] The I_D/I_G ratios of the three samples are 1.24, 1.28, and 1.30, respectively, suggesting that the carbon contains numerous defects, which would benefit the stabilization of quasi-copper-mers. The abundant nitrogen atoms and defects can help anchor and stabilize Cu atoms to assist the formation of quasi-copper-mers.

[IMAGE OMITTED. SEE PDF]

Cu ions were introduced into C₃N₄ substrate by the impregnation method, and then converted to atomic Cu species upon the heat treatment of C₃N₄ and acid leaching. By tuning the loading of the copper precursors during the impregnation, the spacing between neighboring Cu atoms is controlled. Hence, by raising the loading level, the quasi-copper-monomers (1Cu@NC), quasi-copper-dimers (2Cu@NC), and quasi-copper-trimers (3Cu@NC) can be achieved. In this work, the loading of Cu in the as-prepared three samples were determined by Inductively Coupled Plasma (ICP) to be 2.43, 3.13, and 3.41 wt.%, respectively. Before the acid leaching, the 3Cu@NC sample shows small Cu peaks and the scanning electron microscopy (SEM) image shows the clear presence of copper particles (Figure S1b,c, Supporting Information), while after acid leaching, XRD results of all three samples only show a carbon peak and no copper metal peak is detected (Figure 1b), suggesting that no crystalline Cu is formed. Copper particles cannot be observed as well in all three samples, as shown in Figure 1c–e. These results indicate that copper metal particles have been leached out and copper atoms are dispersed into the N-doped carbon matrix without obvious agglomeration. Further investigation of the samples through HAADF-STEM has revealed the distribution of Cu atoms in the N-doped carbon substrate. As shown in Figure 1f–h, 1Cu@NC has few Cu atoms, and the atoms are separated from each other. With increasing Cu content, more 2Cu@NCs are formed, and the inter-spacings between Cu atoms are shorter, and 2Cu sites can be observed on the carbon matrix. Further increasing the Cu content, 3Cu sites can be found in the 3Cu@NC sample, which is indicated by the red circles. This suggests that the number of quasi-copper-dimer and quasi-copper-trimer sites become increasingly higher when the loading of Cu increases.

The extended X-ray absorption fine structure (EXAFS) results demonstrate more clearly the local structures of the Cu atoms. More specifically, they reveal that the three samples after acid washing present atomically dispersed Cu, whose main coordination structure can be fitted to Cu─N₄, as shown in Figure 2a and Figure S2 (Supporting Information). The fitting parameters are shown in Table S1 (Supporting Information). It is seen that even with the increase of Cu loading, each Cu atom still binds with four N atoms and is separated from other Cu atoms. The Cu 2p X-ray photoelectron spectroscopy (XPS) spectrum shows a characteristic Cu²⁺ peak and its corresponding satellite peak. The X-ray absorption near-edge structure (XANES) spectrum also suggests the oxidation state of Cu in the three samples is similar to that of CuO (Figure 2b,c). These results prove again that there is no Cu─Cu direct bond formed and the Cu─N₄ coordination environment is verified. Further investigation by using high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) revealed the distribution of Cu atoms on the N-doped carbon. As shown in Figure 1f–h, there are few Cu atoms, and there are separated from each other. With increasing Cu content, more Cu atoms appear, and the spacing between Cu atoms become closer. 2Cu sites can well be observed on the carbon matrix. Further increasing the Cu content, 3Cu sites can be found in 3Cu@NC sample, which is indicated by the red circles. The structures of the three samples therefore are in agreement with the quasi-copper-mers model and will be used subsequently to evaluate their CO₂RR performance.

[IMAGE OMITTED. SEE PDF]

In addition, the preferred formation of quasi-copper-mer structures are further supported by our DFT simulations. Figure S3 (Supporting Information) shows the possible geometries of the different copper-mers models in the 1Cu, 2Cu, and 3Cu@NC samples. The relationship between the neighboring Cu atoms can be classified into two types – direct Cu─Cu interaction and indirect (i.e., quasi-type) Cu─Cu interaction. As suggested by the EXAFS results discussed above, the coordination structure of Cu single-atoms is determined to be Cu─N₄ as shown in Figure 2d, which is also well documented in the available literatures.^[7a,b] For the quasi-copper-dimers, two different geometries are proposed in Figure S3 (Supporting Information). Cu₂N₈ structure shows two Cu─N₄ that are close to each other while Cu₂N₇ shows two Cu─N₄ sharing one N atom. The Cu─Cu distance of 4.05 and 3.65 Å are much longer than the Cu─Cu distance of 2.34 Å, suggesting that these two Cu atoms do not interact directly, which highlights the uniqueness of the quasi-copper-dimers. The other Cu dimer of Cu₂N_6, as shown in Figure S3 (Supporting Information), demonstrates a Cu─Cu distance of 2.40 Å, which is close to the experimental 2.34 Å, suggesting that these two neighboring Cu atoms are directly bonded. As for the Cu trimer, seven possible structures were considered with three of them containing indirect Cu─Cu interaction solely (Cu₃N₁₁, Cu₃N₁₂-a, and Cu₃N₁₂-b, as shown in Figure S3, Supporting Information), two of them only contain direct Cu─Cu interaction (Cu₃N₇ and Cu₃N₈), and two of them contain a mixture of direct and indirect Cu─Cu interaction (Cu₃N₁₁ -a and Cu₃N₁₀). The formation energy was calculated for the final Cu adsorption to form Cu dimers and trimers. The results are summarized in Figure S4 (Supporting Information), in which the adsorption energy of the Cu to form the quasi-dimer/trimer is lower than that of the direct dimers/trimers, suggesting that the quasi-copper-mer geometries are more stable than those of the direct Cu─Cu interaction geometries. The formation energy calculation thus well supports the EXAFS and XANES results that only quasi-copper-mers are detected because they are energetically more favorable. Thus, as shown in Figure 2d–i, only the models with indirect Cu─Cu interaction are selected for further Gibbs free energy calculations.

Catalytical Performance

In the present work, we have successfully synthesized an appropriate substrate with a high population of anchoring sites, together with abundant chemical N doping, to stabilize the Cu quasi-copper-mers and prevent their aggregation. We have performed both first-principles calculations and EXAFS and XANES characterization to support the stability of the quasi-copper-mers atomic configurations. Both our theoretical and experimental results provide compelling evidence supporting the stability of the quasi-copper-mers. The electrocatalytic activity of the three samples for CO₂RR was thus performed in an H-type cell with CO₂ saturated 0.5 m KHCO₃ solution as the electrolyte. The gas products were analyzed by gas chromatography analysis. It suggests the gas-phase products catalyzed by all three samples are only CO and H₂; both methane and ethylene are not detected in the measurements. The total faradic efficiency for the two products exceeds 90% at all potential. The rest products should be different liquid compounds, which are not focused in this work and termed as “liquid product” to denote all the potential liquid products. Compared to the onset potential of −0.49 V for 1Cu@NC, the 2Cu, and 3Cu@NC exhibit a less cathodic value of −0.43 and −0.38 V, respectively (Figure 3a). In addition, the mass-specific partial current density has been calculated using the weight ratio of Cu determined by ICP for the three samples before the CO₂RR measurements, which is similar as the previous literatures.^[18] 3Cu@NC shows a much higher mass-specific partial current density of CO as compared to that of the 1Cu@NC sample at the potential of −0.7 to −1 V, and reaches a 100% higher value at −0.8 V. Meanwhile, the mass-specific partial current density of CO for 3Cu@NC is also 29% higher than that for 2Cu@NC. These measurement results indicate a higher CO₂ to CO catalytic ability of 3Cu@NC (Figure 3b). Concerning the competition between HER and CO₂RR, the Faradic efficiency of CO and H₂ is analyzed and shown in Figure 3c. At the potential of −0.8 V, 3Cu@NC exhibits the lowest HER contribution as compared to the other two samples. Its Faradic efficiency of CO reaches ≈56%, while the values for 1Cu@NC and 2Cu@NC are only 41% and 47%, respectively. When the potential is < −1 V, the majority of the gas product is H₂ for all three samples, and the faradaic efficiency of CO becomes less than 10%. The electrocatalytic performance results suggest that the selectivity and activity of the 3Cu@NC are better than those of 2Cu@NC and the two samples are both better than 1Cu@NC. Hence, our results suggest that by increasing the number of CuN₄ units, the catalytic activity, and selectivity of the catalyst can be both enhanced.

[IMAGE OMITTED. SEE PDF]

To understand the mechanism of such improvement in electrocatalytic performance, DFT calculations have been carried out. Both DFT calculations and experimental techniques have already been extensively applied in studying the CO₂ reduction to CO on crystals, nanoparticles/clusters, and atomic Cu.^[19] Previous studies generally agree that *CO₂ is reduced to ^*CO via ^*COOH as the intermediate. Herein, we adopt this reaction pathway in our DFT calculations to explore the catalytical energy landscapes of our quasi-copper-mer catalysts.

The Gibbs free energy change (ΔG) for CO₂ reduction to CO catalyzed by quasi-copper-mers are summarized in Figure 4a and Table S2 (Supporting Information). For 1Cu@NC, the CO₂ weakly adsorbs on the substrate with E_ads = −0.13 eV, which is still stronger than the −0.02 to −0.05 eV on Cu surface^[20] and −0.03 eV on graphene.^[21] The relatively stronger adsorption energy of CO₂ would be beneficial for atomic Cu to fix and activate CO₂, thereby making the reduction reaction more favorable. Here the ΔG for ^*CO₂ protonation to ^*COOH is calculated to be 1.31 eV (as shown in Figure 4a, black line and Table S2, Supporting Information), which is in good agreement with the literature value.^[22] Then, it will undergo an exothermic reaction to ^*CO with ΔG of −0.54 eV. There are two different 2Cu@NCs, namely Cu₂N₇ and Cu₂N₈, as shown in Figure S3 (Supporting Information). The 2Cu@NC Cu₂N₇ performs differently from 1Cu@NC (Figure 4a, red line vs black line) that on top of Cu₂N₇, ^*CO₂ is reduced to ^*COOH with an exothermic ΔG of −0.03 eV and then ^*COOH is reduced to ^*CO with an endothermic ΔG of 0.79 eV. The limiting potential (U_e) is 0.79 eV, which is much lower than the U_e of 1.33 eV for the 1Cu@NC. We expect that 2Cu@NC Cu₂N₇ is more active than 1Cu@NC for CO₂RR to CO. The 2Cu@NC Cu₂N₈ performs similarly with 1Cu@NC with ΔG of 1.33 and −0.88 eV, when catalyzing CO₂RR to CO. It is difficult to determine the quantity of these two 2Cu@NCs, considering that Cu₂N₇ is more active than 1Cu@NC and Cu₂N₈ performs similarly to 1Cu@NC. We, therefore, conclude that the 2Cu@NC presents a higher activity than that of 1Cu@NC due to the combined effect of Cu₂N₇ and Cu₂N₈. The catalytical performance for 3Cu@NCs is remarkable. As shown in Figure 3a and Table S2 (Supporting Information), the 3Cu@NC Cu₃N₁₁ demonstrates an exothermic ΔG of −0.06 eV when catalyzing ^*CO₂ to ^*CO, and then to ^*COOH on Cu₃N₁₁ by further releasing an energy of 0.16 eV to obtain the final product of ^*CO. Both reaction steps are exothermic, which indicate that the CO₂RR to CO on Cu₃N₁₁ is thermodynamically favorable. Cu₃N₁₂-a demonstrates an exothermic ΔG as low as −0.75 eV for ^*CO₂ to ^*COOH and then needs 0.24 eV to the final product of ^*CO. The ΔG for CO2RR to CO on Cu₃N₁₂-b via ^*COOH is 0.72 and −1.22 eV. The limiting potential is 0.72 eV, which is lower than the 0.79 and 1.33 eV for the two 2Cu@NCs as well as the 1.31 eV for 1Cu@NC.

[IMAGE OMITTED. SEE PDF]

Our DFT calculations clearly show that the quasi-copper-trimer outperforms the quasi-copper-monomer and quasi-copper-dimer when catalyzing CO₂RR to CO. It is of interest to note that the 3rd Cu atom is of significant importance in enhancing the activity. The C₂N₇ (Figure 2d) gives a Ue of 0.79 eV while the 3rd Cu (as in C₃N₁₁,) makes the whole process thermodynamically favorable. Similarly, the C₂N₈ gives a Ue of 1.33 eV while the 3rd Cu (as in C₃N₁₂-b) effectively lowers the Ue to 0.72 eV. To further reveal the underlying mechanism responsible for the changes, we have performed p-band center calculations on the active sites of all the catalysts (Figure S5, Supporting Information). Interestingly, a volcano curve is revealed, which can explain the activity trend of all these Cu mers catalysts based on the understanding that the one with its p-band center closer to the optimum value gives a better activity. Importantly, the role of the 3rd Cu atom is also revealed: The 3rd Cu atom is able to move the active site's p-band center toward the optimum value, thus giving rise a better activity.

The competition between CO₂RR and HER is another major concern when designing highly efficient catalysts for CO₂RR, as HER can lower the CO₂RR selectivity and efficiency.^[23] Therefore, a good catalyst should exhibit appropriate electronic properties which not only favor the CO₂RR but also dramatically suppress the competitive HER. Herein, we have compared the selectivity performance of these quasi-copper-mer catalysts against HER. As shown in Figure 4b and Table S3 (Supporting Information), the 1Cu@NC demonstrates a positive ΔG of 0.46 eV, while 2Cu@NCs Cu₂N₇ and Cu₂N₈ demonstrate a ΔG of −0.59 and −0.24 eV, respectively. On the one hand, it is hard to conclude whether the 2Cu@NC is more active than 1Cu@NC for HER as it is dependent on the geometry of 2Cu@NC (Cu₂N₇ or Cu₂N₈). On the other hand, 3Cu@NCs Cu₃N₁₁ and Cu₃N₁₂-a exhibit a high ΔG of −0.84 and −1.29 eV, respectively. Such low ΔG values indicate that they are not active in HER reaction, thus be beneficial for CO₂RR. Hence, the 3Cu@NCs Cu₃N₁₁ and Cu₃N₁₂-a are expected to possess a good selectivity against HER. However, the 3Cu@NC Cu₃N₁₂-b demonstrates a ΔG of −0.14 eV, which is the closest to the zero line, indicating that it is active in HER. In summary, 3Cu@NCs Cu₃N₁₁ and Cu₃N₁₂-a demonstrate their robustness against HER because they exhibit a high ΔG (−0.84 and −1.29 eV, respectively). And the small ΔG (−0.14 eV) suggests that 3Cu@NC Cu₃N₁₂-b is active in HER. Since in the real experiments, the quasi-copper-trimer sample can be a mixture of all these different geometries, we can still expect that the 3Cu group demonstrates a better selectivity than 1Cu@NC and 2Cu@NC.

Both the DFT calculations and the experimental performance results show that 3Cu@NCs have both higher activity and selectivity to CO than 2Cu@NCs and 1Cu@NC. This demonstrates that by increasing the number of CuN₄ building blocks, the CO₂RR performance can be enhanced. Hence, quasi-metal-mers present a great potential to improve the current catalytical performance. In the present study, we have studied the quasi-copper-mers up to three CuN₄ units. Clearly, the properties of the sample with more CuN₄ units are worth being investigated in the future. In addition, the atomic arrangement of quasi-copper-mers also plays an important role in the electrocatalytic CO₂RR performance. Establishing a precise control method to specifically synthesize one type of quasi-metal-mers will be another interesting topic worth further investigation.

Conclusion

Going beyond the recently established SACs, where single atomic entity functions, we proposed a new class of electrocatalysts: quasi-copper-mers embedded in an N-doped graphene-like substrate, and investigated their atomic structures and electrocatalytic performance for CO₂RR to CO by both experimental characterizations and DFT calculations. First, we synthesized the quasi-copper-mer catalysts and demonstrated that the effectiveness of the indirect Cu─Cu quasi-copper-mers in electrocatalysis, including both quasi-copper-dimers and quasi-copper-trimers. We then performed DFT calculations on various potential structures with both direct and indirect Cu─Cu interactions, and shown that the indirect Cu─Cu quasi-copper-mer structures are thermodynamically more favorable, thus supporting our experimental observations. Our performance characterization results showed that the quasi-copper-trimer outperforms quasi-copper-monomer and quasi-copper-dimer when catalyzing CO₂ to CO, which is consistent with the energy landscape of CO₂RR from our DFT calculations. More interestingly, quasi-copper-trimers exhibit a 2-times higher partial current density of CO than that of quasi-copper-monomer. Meanwhile, the ratio of Faradic efficiency of CO to H₂ for quasi-copper-trimer sample is also the highest among all three samples, at −0.8 V. Hence, we conclude that Cu quasi-trimers outperform both quais-copper-monomer and quasi-copper-dimer when electrocatalyzing CO₂RR to CO. The p-band center calculations explain the activity trend based on the understanding that the one with its p-band center closer to the optimum value gives a better activity. It also highlights the important role of the 3rd Cu atom that can move the p-band center of the active site toward the optimum value, thus giving rise to a better activity. The present work presents a novel route in the design of new electrocatalysts for CO₂RR to CO with high activity, selectivity, and stability.

Experimental Section

Density Functional Theory (DFT)

All the calculations were performed based on DFT with the Perdew–Burke–Ernzerhof functional under the generalized gradient approximation^[24] for the exchange-correlation interaction, as implemented in the Vienna Ab initio Simulation Package (VASP).^[25] A cutoff kinetic energy of 500 eV was applied to expand the electronic wave functions and the projector augmented-wave method was adopted to describe the electron-core interaction. A Gamma centered 3 × 3 × 1 k-mesh was used for the structural optimization. The Gibbs free energy change (ΔG) at each electrochemical step involving a proton–electron transfer was computed based on computational hydrogen electrode (CHE) model, in which the free energy of (H⁺ + 𝑒⁻) equals to $$\frac{1}{2}{H}_2( g )$$ for standard hydrogen electrode (SHE).^[26] So the ΔG of each reaction step is defined as: 1 $$$\begin{equation} \Delta G = \Delta E_{adsorp} - \Delta E_{ZPE} - T \Delta S\end{equation}$$$where ΔE_adsorp is the adsorption energy differences between the product adsorbate and reactant adsorbate, ΔE_ZPE is the difference in zero-point energy, T is the temperature (300 K) and ΔS is the entropy difference between the adsorbed adsorbate and non-adsorbed gas-phase adsorbate. ΔGdescribes the energy needed or released for a reaction to occur. A positive ΔG suggests an endothermic reaction, while a negative ΔG suggests an exothermic reaction. It is understood that the higher the ΔG is, the more energy is needed for a reaction to take place. To better mimic the real experiments, solvation effects have been included by using an implicit model.^[27]

Preparation of g─C₃N₄

A covered ceramic crucible filled with urea was heated in a muffle furnace at 525 °C for 4 h at the heating rate of 5 °C min⁻¹. The yellowish g─C₃N₄ powder was then obtained and grinded.

Preparation of 1Cu, 2Cu, and 3Cu@N-Carbon

g─C₃N₄ (4.4 g) and 3.6 g of Pluronic F127 were added into 360 mL of deionized (DI) water. The mixture was then sonicated for 1.5 h and stirred for 1.5 h to achieve uniform dispersion. Afterward, 20 mL of 0.2 m, 0.5 m, and 1 m CuCl₂ aqueous solution were added drop-by-drop into the above mixture respectively and stirred overnight. The dispersion was collected and washed using DI water three times by centrifugation at 8000 rpm for 5 min each. After drying, the powder thus-obtained was calcinated at 550 °C for 2 h and 800 °C for 1 h at the heating rate of 3 °C min⁻¹ in Argon atmosphere. Afterward, 10 mg of the obtained product was immersed in 20 mL of 2 m HCl solution for 6 h with stirring to leach out the copper particles and clusters. 1Cu, 2Cu, and 3Cu@N-carbon were then obtained after filtering, washing, and drying.

Characterization

The morphology and structure of the prepared samples were characterized by using scanning electron microscopy (Zeiss Supra 40) and scanning transmission electron microscopy (JEOL ARM200F). X-ray diffraction analysis studies were operated using Bruker D8 diffractor at 40 kV and 40 mA with Cu K radiation (0.15 406 nm). Raman spectroscopy was performed using HORIBA LabRAM HR Evolution Raman microscopes with an Argon laser (λ = 514 nm, National Laser Model 800AL) as the excitation line. The X-ray photoelectron spectroscopy tests were conducted using Kratos Analytical Axis Ultra DLD UHV. The results were callibrated by alinging the carbon 1s peak to 284.6 eV. The elemental composition were determined by using the ThermoFisher Scientific FlashSmart CHNS Elemental Analyzer and ICP (Perkin Elmer Avio 500). The X-ray absorption spectra (XAS) including X-ray absorption near-edge structure and extended X-ray absorption fine structure of the samples at Cu K-edge were collected at the XAFCA beam line of the Singapore Synchrotron Light Source (SSLS), where a pair of Si (111) crystals was used in the monochromator. The XAS data were recorded in a transmission mode. Cu foil, Cu, and CuO were used as references. The storage ring was working at the energy of 700 M eV with an average electron current of 200 mA.

Electrochemical Measurements

The electrochemical studies were performed in an H-type cell using the electrochemical workstation (CHI 760E). Pt mesh and Ag/AgCl electrode (3.5 m KCl) were used as the counter electrode and reference electrode, respectively. 4.5 mg of catalysts were suspended in 440 µL ethanol solution with 10 µL Nafion added. Then, 100 µL of catalyst ink was drop cast onto a carbon paper (1 × 1 cm²). The mass loading of the catalyst on working electrodes was 1 mg cm⁻². A CO₂-saturated 0.5 m KHCO₃ (pH ≈7.22) was used as the electrolyte with continuous CO₂ supply at the flow rate of 23 mL min⁻¹. The volume of electrolyte was 30 mL for both anode and cathode chambers in the H-type cell. All potentials measured were calibrated to the reversible hydrogen electrode (RHE) reference scale using E_RHE = E_Ag/AgCl + 0.0591 × pH + 0.2046. Gas products were analyzed by using the online gas chromatograph (Shimadzu, 2014C). H₂ was detected by a thermal conductivity detector, and CO was detected by a flame ionization detector. The Faradaic efficiency was calculated based on the equation below, 2 $$$\begin{equation} FE = \frac{{nvpVF}}{{IRT}} \times 100 \%\end{equation}$$$where n is the electron transfer number, v is the volume concentration of examined gas in the outlet gas, p is the atmospheric pressure (1.013 × 10⁵ Pa), V is the gas flow rate, F is the Faradic constant, I is the steady-state total current density, R is the ideal gas constant, T is the room temperature (298.15 K). The test at potential of −0.8 V has been repeated for three times for calculating the error bar.

Acknowledgements

J.Y. and X.L. contributed equally to this work. Y.W.Z. and team acknowledge the support from the Italy-Singapore Science and Technology Cooperation (Grant no. R23101R040) and the Singapore A*STAR SERC CRF Award and the use of computing resources at the A*STAR Computational Resource Centre and National Supercomputer Centre, Singapore. J.W. and team thank the support of the Singapore Ministry of Education (Tier 1, A-8000186-01-00) and the Singapore National Research Foundation (NRF-CRP26-2021-0003), for research conducted at the National University of Singapore. The authors acknowledge the XAFCA beamline at the Singapore Synchrotron Light Source (SSLS) for providing the facilities necessary for conducting the XAFS measurements.

Conflict of Interest

The authors declare no conflict of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

a) W. Ju, A. Bagger, G.‐P. Hao, A. S. Varela, I. Sinev, V. Bon, B. Roldan Cuenya, S. Kaskel, J. Rossmeisl, P. Strasser, Nat. Commun. 2017, 8, 944;

b) W. Liu, P. Zhai, A. Li, B. Wei, K. Si, Y. Wei, X. Wang, G. Zhu, Q. Chen, X. Gu, R. Zhang, W. Zhou, Y. Gong, Nat. Commun. 2022, 13, 1877.

a) K. P. Kuhl, T. Hatsukade, E. R. Cave, D. N. Abram, J. Kibsgaard, T. F. Jaramillo, J. Am. Chem. Soc. 2014, 136, [eLocator: 14107];

b) D. Raciti, C. Wang, ACS Energy Letters 2018, 3, 1545.

H. Yoshio, K. Katsuhei, S. Shin, Chem. Lett. 1985, 14, 1695.

Y. Wang, H. Shen, K. J. T. Livi, D. Raciti, H. Zong, J. Gregg, M. Onadeko, Y. Wan, A. Watson, C. Wang, Nano Lett. 2019, 19, 8461.

R. Reske, H. Mistry, F. Behafarid, B. Roldan Cuenya, P. Strasser, J. Am. Chem. Soc. 2014, 136, 6978.

S. Meshitsuka, M. Ichikawa, K. Tamaru, J. Chem. Soc., Chem. Commun. 1974, 5, 158.

a) Y. Cheng, S. Yang, S. P. Jiang, S. Wang, Small Methods 2019, 3, [eLocator: 1800440];

b) A. S. Varela, N. Ranjbar Sahraie, J. Steinberg, W. Ju, H.‐S. Oh, P. Strasser, Angew. Chem., Int. Ed. 2015, 54, [eLocator: 10758];

c) J.‐H. Liu, L.‐M. Yang, E. Ganz, ACS Sustainable Chem. Eng. 2018, 6, [eLocator: 15494].

Y. Zhao, J. Liang, C. Wang, J. Ma, G. G. Wallace, Adv. Energy Mater. 2018, 8, [eLocator: 1702524].

M. Jia, S. Hong, T.‐S. Wu, X. Li, Y.‐L. Soo, Z. Sun, Chem. Commun. 2019, 55, [eLocator: 12024].

E. Zhang, T. Wang, K. Yu, J. Liu, W. Chen, A. Li, H. Rong, R. Lin, S. Ji, X. Zheng, Y. Wang, L. Zheng, C. Chen, D. Wang, J. Zhang, Y. Li, J. Am. Chem. Soc. 2019, 141, [eLocator: 16569].

P. Huang, M. Cheng, H. Zhang, M. Zuo, C. Xiao, Y. Xie, Nano Energy 2019, 61, 428.

H. Yang, Y. Wu, G. Li, Q. Lin, Q. Hu, Q. Zhang, J. Liu, C. He, J. Am. Chem. Soc. 2019, 141, [eLocator: 12717].

D. Karapinar, N. T. Huan, N. Ranjbar Sahraie, J. Li, D. Wakerley, N. Touati, S. Zanna, D. Taverna, L. H. Galvão Tizei, A. Zitolo, F. Jaouen, V. Mougel, M. Fontecave, Angew. Chem., Int. Ed. 2019, 58, [eLocator: 15098].

L. Zhang, T. Yang, W. Zang, Z. Kou, Y. Ma, M. Waqar, X. Liu, L. Zheng, S. J. Pennycook, Z. Liu, X. J. Loh, L. Shen, J. Wang, Adv. Sci. 2021, 8, [eLocator: 2101344].

a) K. P. Kuhl, E. R. Cave, D. N. Abram, T. F. Jaramillo, Energy Environ. Sci. 2012, 5, 7050;

b) L. Zhang, Z.‐J. Zhao, J. Gong, Angew. Chem., Int. Ed. 2017, 56, [eLocator: 11326].

P. Qiu, H. Chen, C. Xu, N. Zhou, F. Jiang, X. Wang, Y. Fu, J. Mater. Chem. A 2015, 3, [eLocator: 24237].

D. B. Schuepfer, F. Badaczewski, J. M. Guerra‐Castro, D. M. Hofmann, C. Heiliger, B. Smarsly, P. J. Klar, Carbon 2020, 161, 359.

a) Q. He, J. H. Lee, D. Liu, Y. Liu, Z. Lin, Z. Xie, S. Hwang, S. Kattel, L. Song, J. G. Chen, Adv. Funct. Mater. 2020, 30, [eLocator: 2000407];

b) Y. Cheng, S. Zhao, H. Li, S. He, J.‐P. Veder, B. Johannessen, J. Xiao, S. Lu, J. Pan, M. F. Chisholm, S.‐Z. Yang, C. Liu, J. G. Chen, S. P. Jiang, Appl. Catal. B 2019, 243, 294.

a) X. Liu, J. Xiao, H. Peng, X. Hong, K. Chan, J. K. Nørskov, Nat. Commun. 2017, 8, [eLocator: 15438];

b) X. Zhang, J.‐X. Liu, B. Zijlstra, I. A. W. Filot, Z. Zhou, S. Sun, E. J. M. Hensen, Nano Energy 2018, 43, 200;

c) J. Yang, X. Liu, H. Yuan, J. Sun, L. Li, K. E. J. Goh, Z. G. Yu, J. Xue, J. Wang, Y.‐W. Zhang, Appl. Catal., A 2022, 642, [eLocator: 118708].

G.‐C. Wang, L. Jiang, Y. Morikawa, J. Nakamura, Z.‐S. Cai, Y.‐M. Pan, X.‐Z. Zhao, Surf. Sci. 2004, 570, 205.

N. Osouleddini, S. F. Rastegar, J. Electron Spectrosc. Relat. Phenom. 2019, 232, 105.

K. Zhao, X. Nie, H. Wang, S. Chen, X. Quan, H. Yu, W. Choi, G. Zhang, B. Kim, J. G. Chen, Nat. Commun. 2020, 11, 2455.

A. A. Peterson, J. K. Nørskov, J. Phys. Chem. Lett. 2012, 3, 251.

a) J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865;

b) P. E. Blöchl, Phys. Rev. B 1994, 50, [eLocator: 17953].

a) G. Kresse, J. Non‐Cryst. Solids 1995, 192–193, 222;

b) G. Kresse, J. Hafner, Phys. Rev. B 1993, 48, [eLocator: 13115].

A. A. Peterson, F. Abild‐Pedersen, F. Studt, J. Rossmeisl, J. K. Nørskov, Energy Environ. Sci. 2010, 3, 1311.

K. Mathew, R. Sundararaman, K. Letchworth‐Weaver, T. A. Arias, R. G. Hennig, J. Chem. Phys. 2014, 140, [eLocator: 084106].