Direct low concentration CO₂ electroreduction to multicarbon products via rate-determining step tuning

Abstract

Direct converting low concentration CO₂ in industrial exhaust gases to high-value multi-carbon products via renewable-energy-powered electrochemical catalysis provides a sustainable strategy for CO₂ utilization with minimized CO₂ separation and purification capital and energy cost. Nonetheless, the electrocatalytic conversion of dilute CO₂ into value-added chemicals (C₂₊ products, e.g., ethylene) is frequently impeded by low CO₂ conversion rate and weak carbon intermediates’ surface adsorption strength. Here, we fabricate a range of Cu catalysts comprising fine-tuned Cu(111)/Cu₂O(111) interface boundary density crystal structures aimed at optimizing rate-determining step and decreasing the thermodynamic barriers of intermediates’ adsorption. Utilizing interface boundary engineering, we attain a Faradaic efficiency of (51.9 ± 2.8) % and a partial current density of (34.5 ± 6.4) mA·cm⁻² for C₂₊ products at a dilute CO₂ feed condition (5% CO₂ v/v), comparing to the state-of-art low concentration CO₂ electrolysis. In contrast to the prevailing belief that the CO₂ activation step (${CO}_2+e^{-}+*\to {}^{*}CO_2^{-}$) governs the reaction rate, we discover that, under dilute CO₂ feed conditions, the rate-determining step shifts to the generation of *COOH (${}^{*}CO_2^{-}+H_{2}O\to {}^{*}COOH+{OH}^{-}\left(aq\right)$) at the Cu⁰/Cu¹⁺ interface boundary, resulting in a better C₂₊ production performance.

The development of catalysts that operate under low concentration CO₂ resembling industrial waste gases holds promise for CO₂ reduction. Here, the authors report a vacuum calcination approach for regulating the Cu⁰/Cu¹⁺ density on Cu-based catalysts that can electro-catalyze low-concentration CO₂.

Introduction

Electrochemical reduction of carbon dioxide (CO₂RR) offers a promising avenue for advancing carbon neutrality in an environmentally sustainable manner^1–5. Presently, only metallic Cu-based catalysts exhibit the capability to catalyze CO₂ into a diverse range of hydrocarbons, encompassing C₂₊ products such as C₂H₄ and C₂H₅OH which are in high demand in basic production^6–12. In pursuit of high selectivity for C₂₊ products, previous studies on CO₂ reduction reaction (CO₂RR) have predominantly been conducted using pure CO₂^13–18. It usually comes from large-scale CO₂ capture and sequestration facilities. However, the meticulous enrichment and purification of feed gases still incur substantial capital and energy costs even with latest technologies. For instance, the estimated cost of capturing CO₂ from biomass-based combustion power plant ranges from $150 to $400 per ton of CO₂, with additional purification costs reported to be $70 to $275 per ton of CO₂ depending on the emission source^19–28. The practical CO₂ purity obtainable from military facilities and industrial manufacturing exhausts, such as those from coal power plants and the steel/petrochemical industry, is relatively low, typically ranging from 5 to 33% (CO₂ v/v)^29–38. These exhaust gases are far from being fully utilized resulting in a considerable waste of available carbon source. Therefore, employing electrocatalytic CO₂ reduction with industrially diluted CO₂ as the feed gas, not only avoids the high costs associated with feedstock purification, but also facilitates the sustainable utilization of natural resources and the attainment of the United Nations Sustainable Development Goals (SDGs) 12 & 13, which underscores its considerable significance^39–44.

Recent developments in the electroreduction of diluted CO₂ have centered on two main aspects: (1) controlling the availability of CO₂ (its concentration) to achieve enhanced catalytic performance or modulate the reaction pathway^40,45–50; and (2) pioneering the development of efficient catalysts for the electroreduction of dilute CO₂ while examining their high catalytic activity from a thermodynamic standpoint^51–54. Further studies have utilized simulated flue gas as the feed source for electrocatalytic CO₂ reduction^{35,48–50,55–66}. These innovative catalysts are progressively evolving towards functionalized microporous metal-organic frameworks (MOFs) or active molecular enzyme analogues, which are designed for targeted CO₂ capture or activation^{55,56,58,60–62,67–69}. However, under low CO₂ concentrations, the electroreduction products predominantly consist of C₁ products (e.g., CO, CH₄), which involve relatively simple reduction processes^{23,37,50,55,57,67,70,71}. Most studies have overlooked the relatively low CO₂ feed conditions (≤15% CO₂ v/v) in efforts to mitigate the intense competition with hydrogen evolution^72–76. Some studies have even used CO, an intermediate product of CO₂RR, as the feed source in low-concentration systems to enhance the selectivity for C₂₊ products⁷⁷. Consequently, the characterization of CO₂ reduction reaction (CO₂RR) under such extreme CO₂ feed condition intervals close to realistic situations remains poorly understood. Moreover, the rate of CO₂ reduction reaction (CO₂RR) is closely linked to CO₂ mass transport, which is highly influenced by the purity of the feedstock. However, the impact of dilute CO₂ feedstock on the kinetics of CO₂RR remains insufficiently studied.

Achieving superior selectivity for C₂₊ products under dilute feedstock conditions (≤15% CO₂ v/v) necessitates the construction of catalytically active sites that can promote CO₂ conversion rate. Defect engineering is a prevalent strategy on Cu-based catalysts to manipulate reaction intermediates owing to its unique atomic arrangement^78–82. Among these defects, the adjacent Cu⁰/Cu¹⁺ interface boundary has been recognized as highly active catalytic sites for promoting C₂₊ electroproduction in conventional CO₂RR^83–93 (i.e., CO₂RR in pure CO₂ feed). The synergistic combination of Cu⁰ and Cu¹⁺ active species enhances the electron transfer of CO₂ and promotes *CO dimerization^{79,86,94–103}. Furthermore, several experiments have demonstrated that Cu¹⁺ species enhance the absorption strength of crucial carbon intermediates on the catalytic surface, thereby stabilizing the Cu¹⁺ species¹⁰⁴. Building on these research endeavors, we foresee that the Cu⁰/Cu¹⁺ interface holds promise for achieving outstanding selectivity towards C₂₊ products under dilute-purity feed conditions. Nonetheless, the precise construction of quantifiable Cu⁰/Cu¹⁺ interface boundaries and the exploration of their distinctive kinetic properties in low-purity CO₂ feeds pose significant challenges.

We demonstrate efficient production of C₂₊ compounds under low CO₂ feed conditions (5% v/v) using Cu-based catalysts with abundant Cu⁰/Cu¹⁺ interface boundaries. We utilize a straightforward and stable technique to regulate the density of the Cu⁰/Cu¹⁺ interface between Cu(111) and Cu₂O(111) facets through adjustments in vacuum calcination duration. We discovered that catalytic sites at the Cu⁰/Cu¹⁺ interface boundary display higher catalytic responsiveness with dilute-purity CO₂ feeds than with high-purity CO₂ feeds. Specifically, the selectivity of Cu-based catalysts for C₂₊ compounds correlates positively with the density of Cu⁰/Cu¹⁺ interface boundaries in low CO₂ feed conditions, but this correlation diminishes in high CO₂ feed conditions. Consequently, we attained a faradaic efficiency of (51.9 ± 2.8) % for C₂₊ products at a low CO₂ concentration (5% v/v) using the catalyst featuring the most abundant Cu⁰/Cu¹⁺ interface. Due to the unique catalytic sites at the Cu⁰/Cu¹⁺ interface, the CO₂RR in the alkaline electrolyte microenvironment with dilute feed purity exhibits a distinct reaction mechanism compared to the widely accepted mechanism under high-purity feeds: the rate-determining step shifts backwards to the *COOH formation (${}^{*}CO_2^{-}+H_{2}O\to {}^{*}COOH+{OH}^{-}\left(aq\right)$) instead of maintaining its prior CO₂ activation (${CO}_2+e^{-}+*\to {}^{*}CO_2^{-}$). Furthermore, we investigate the thermodynamic impact of catalytic sites at the Cu⁰/Cu¹⁺ interface on dilute CO₂ electrolysis. Our research offers fresh kinetic insights into the catalytic mechanism of CO₂RR under dilute feed conditions and underscores the significance of directly utilizing dilute industrial CO₂ emissions, which is often overlooked in current studies on heterogeneous electrocatalysis.

Results

Preparation and characterization of catalysts

Our objective was to create Cu-based catalysts with a unique fragmentation structure for CO₂RR experiments. We selected commercial σ-Cu enriched in the Cu(111) and Cu₂O(111) facet as the catalyst precursor. High-level vacuum calcination can delicately modify the surface crystalline structure of the precursor while simultaneously stabilizing its chemical composition. Using this approach, we produced highly fragmented copper with abundant Cu⁰/Cu¹⁺ interface boundaries, termed high-boundary-density copper (HB Cu), through vacuum calcination of the precursor at 250 °C for 30 min (see Methods). To examine the relationship between the Cu⁰/Cu¹⁺ interface structure and C₂₊ selectivity, we crafted control samples with nearly identical chemical compositions and adjustable Cu⁰/Cu¹⁺ interface density. Medium- and low-boundary-density copper samples (MB Cu and LB Cu, respectively. Figure 1a were prepared as controls by vacuum calcination of commercial σ-Cu at 250 °C for 90 and 150 min, respectively. Furthermore, to eliminate the influence of Cu valence, we created a pure Cu catalyst by annealing commercial σ-Cu in an H₂ atmosphere at 250 °C for 2 h. Scanning electron microscopy (SEM) images confirm the dispersed nanoparticles morphology of all Cu-based catalysts (Fig. 1b–d and Supplementary Figs. 1, 3, 5, 7, 9). Analysis of transmission electron microscopy (TEM) images reveals that the particle sizes of these catalysts are quite similar, ranging from approximately 70 to 130 nm, with a slight increase observed with longer annealing times (Fig. 1e–g and Supplementary Figs. 2, 4, 6, 8, 10). X-ray diffraction (XRD) analysis revealed that the chemical compositions of the Cu-based catalysts included Cu(111) and Cu₂O(111), while pure Cu exclusively exhibited the Cu(111) facet.

Fig. 1. Characterization of the copper electrocatalysts.

a Schematic illustration for the preparation of the HB Cu, MB Cu and LB Cu catalysts. TEM of the HB Cu (b), MB Cu (c), and LB Cu (d) catalysts before CO₂RR, respectively. Scale bars, 100 nm. Insets show the particle size distribution of the corresponding catalysts. Source data are provided as a Source Data file. HRTEM images showing crystallographic information of HB Cu (e), MB Cu (f), and LB Cu (g) catalysts before CO₂RR, respectively. Scale bars, 5 nm. Locally magnified images showing the magnified grain boundaries. Scale bars, 2 nm. HRTEM images showing crystallographic information of HB Cu (h), MB Cu (i), and LB Cu (j) catalysts after 30 min of CO₂RR, respectively. Scale bars, 10 nm. Locally magnified images showing the magnified grain boundaries. Scale bars, 2 nm. The yellow and orange dashed lines circle the facet fragments of the Cu(111) and Cu₂O(111), respectively, and the white dashed line highlights the interface boundary between the Cu(111) and Cu₂O(111) facets.

As depicted in Supplementary Fig.11, the X-ray diffraction (XRD) analysis revealed that all samples primarily consisted of Cu(111) with a minor presence of Cu₂O(111), whereas pure Cu exhibited solely the Cu(111) facet. The diffraction peaks for Cu(111) and Cu₂O(111) narrow and sharpen with longer calcination times, suggesting a gradual shift in the crystalline phase from concentrated fine fragments to dispersed bulk structures¹⁰⁵. The average sizes of the crystalline phase fragments in the three samples, determined using the Scherrer equation, increase with longer calcination times (Supplementary Table 1). This trend corresponds to the changes observed in XRD diffraction peaks. X-ray photoelectron spectroscopy (XPS) and Cu LMM Auger spectra (Supplementary Figs. 12, 13) provide further confirmation of the distinctions and occupancy between Cu¹⁺ and Cu⁰. The comparable valence ratios observed in all three samples dismiss Cu¹⁺: Cu⁰ content as a significant factor influencing CO₂RR performance⁷⁹ (Supplementary Table 2).

Characterization of crystalline structure

We accurately determined the crystalline structure of HB Cu and the control groups using High-resolution Transmission Electron Microscopy (HRTEM) imaging. We then employed a random sampling method for normalized statistical analysis of the HRTEM images. Cu⁰ is mainly present in Cu(111) planes (outlined in yellow), exhibiting a lattice spacing of 2.1 Å, and in trace amounts of Cu(200) (outlined in yellow) with a lattice spacing of 1.8 Å. The interplanar spacing of 2.5 Å corresponds to Cu₂O(111) planes (outlined in orange) interplanar spacing. HB Cu comprises a considerable amount of fragmented Cu(111) and Cu₂O(111) closely spaced, leading to abundant Cu⁰/Cu¹⁺ interface boundaries (outlined in white in Fig. 1e–g). These boundaries are anticipated to enhance the absorption of vital carbonate intermediates. MB Cu exhibits enlarged and dispersed facet fragments, resulting in a moderately abundant presence of Cu⁰/Cu¹⁺ interface boundaries. Conversely, LB Cu primarily comprises uniform Cu(111) facet fragments, with the interfaces being nearly undetectable (Supplementary Figs. 14–16 present additional HRTEM images of HB Cu, MB Cu and LB Cu). The dimensional characteristics of the grain fragments in the HRTEM images closely correspond to the results of the Scherrer equation. To accurately grasp the structural characteristics of the three catalysts, we quantified the density of Cu⁰/Cu¹⁺ interface boundaries in a large number of random HRTEM images, normalizing their length to the sample’s areas⁷⁷. Supplementary Fig. 17 illustrates that HB Cu, possessing abundant fragmented crystalline structure, has relatively high density of Cu⁰/Cu¹⁺ interface boundaries (22.67 μm⁻¹), while MB Cu (14.92 μm⁻¹) and LB Cu (0.94 μm⁻¹) exhibit lower interface boundaries density. This characterization evidence indicates that the interface boundary density follows the sequence HB Cu>MB Cu>LB Cu. To further verify the interface boundary density of these Cu-based catalysts, CO stripping voltammetry was conducted for HB Cu, MB Cu and LB Cu (Supplementary Fig. 18). The results showed that the anodic peaks associated with CO electrooxidation of HB Cu (~1.030 V vs. RHE) exhibit a significant negative shift relative to MB Cu (~1.055 V vs. RHE) and LB Cu (~1.065 V vs. RHE). This suggests that grain boundaries serve as active sites for CO stripping and that increasing the Cu⁰/Cu¹⁺ interface boundary density lowers the overpotential required for CO oxidation. The order of exposed active site density among the three catalysts is HB Cu>MB Cu>LB Cu, consistent with the random sampling observations from HRTEM. Supplementary Figs. 19–24 offer insights into the morphology and physical composition of all Cu catalysts after 30 min of CO₂RR. The surface structural features exhibit minimal changes, with the gradient trends in crystalline interface boundary density between Cu(111) and Cu₂O(111) remaining largely unchanged.

Electrocatalytic performance

We subsequently assessed the CO₂RR performance of all Cu-based catalysts across various CO₂ purities by adjusting the volume ratio of CO₂ to N₂ in the feed gas stream within the flow cell (Supplementary Fig. 25). Figure 2a, d illustrate the faradaic efficiencies of C₂₊ products generated by HB Cu across different CO₂ purity feedstocks, with the corresponding partial current densities shown in Supplementary Fig. 26. When decreasing the incoming gas purity from pure to 2.5%, we observed a general decrease in the current densities of the C₂₊ products, particularly noticeable for feed gas purities below 30%, although not precisely mirrored in their faradaic efficiency. In the less negative potential range (−0.51 V ~ −0.71 V vs. RHE), the improvement in C₂₊ faradaic efficiencies shows an exponential rise at CO₂ concentrations of 2.5%, 5%, and 10%. This upward trend continues down to −0.91 V vs. RHE and extends into more negative potential ranges for CO₂ concentrations above 30%. The performance across these potential intervals suggests that there is ample *CO₂ availability on the catalyst surface for CO₂ electroreduction. Increasing the applied potential and decreasing feed purity result in decreased CO₂RR efficiency due to insufficient *CO₂ supply for electroreduction. This leads to mass transfer limitations and faint hydrogen evolution, consistent with prior research findings. At 5% CO₂, a peak FE_C2+ of (51.9 ± 2.8) % is attained at −0.71 V vs. RHE (Fig. 2h). The corresponding partial current density (normalized to the geometric area of the GDE, J_C2+) was (34.5 ± 6.4) mA·cm⁻², ranking relatively high among recent studies investigating electrolysis under dilute CO₂ feeds (Supplementary Tables 9-10). Samples with fewer Cu⁰/Cu¹⁺ interface boundaries (MB Cu and LB Cu) also exhibit catalytic performances towards C₂₊ products with volcano-shaped curves, influenced by mass transfer limitations (Supplementary Figs. 27, 28). At 5% CO₂, MB Cu and LB Cu only reach maximum FE_C2+ values of (32.6 ± 4.2) % and (25.2 ± 3.5) %, respectively (Fig. 2h). The C₂₊ selectivity of the samples exhibits a positive correlation with Cu⁰/Cu¹⁺ interface density (Fig. 2c, f). Furthermore, in achieving 0.1 A·cm⁻² in electrolysis, HB Cu demonstrates a lower onset potential compared to MB Cu and LB Cu, indicating its higher intrinsic activity (Fig. 2i). In comparison to the plot in 5% v/v CO₂ (Fig. 2b), the C₂₊ selectivity of samples remains comparable in pure CO₂ (Fig. 2e), indicating that the dominance of Cu⁰/Cu¹⁺ interface is not evident under conditions abundant in mass transfer. To exclude the possibility of N₂ electroreduction reaction(N₂RR) on the Cu⁰/Cu¹⁺ interface, we also examined the main products of N₂RR on HB Cu: NH₃ and N₂H₄ (Supplementary Figs. 29, 30). In addition, several control experiments were conducted: (1) commercial σ-Cu nanoparticles without calcination, (2) pure Cu nanoparticles after H₂ reduction and (3) Cu-based catalysts with varying Cu/Cu₂O ratios obtained through air calcination (the synthesis is shown in Methods). All comparison catalysts were tested under a 5% v/v CO₂ feed and exhibited different C₂₊ selectivities, particularly the pure Cu, which lacked a detectable Cu⁰/Cu¹⁺ interface. Moreover, the selectivity for C₂₊ products in Cu-based catalysts with varying Cu/Cu₂O ratios was strongly and positively correlated with the Cu⁰/Cu¹⁺ interface boundary density (Fig. 2f and Supplementary Figs. 31–38). These experimental observations indicate that the Cu⁰/Cu¹⁺ interface boundary significantly reduces mass transfer limitations, thereby enhancing C₂₊ selectivity under dilute CO₂ feed conditions.

Fig. 2. CO₂RR electroreduction performance.

a Faradaic efficiencies (FE) for C₂₊ products produced by HB Cu at 2.5%, 5%, 7.5%, 10% and 15% CO₂ feed concentration over a range of potentials. Curves connecting data points serve as visual guides. b Faradaic efficiencies (FE) for C₁ and C₂₊ products as well as current densities (j) for CO₂ electroreduction produced by HB Cu, MB Cu, LB Cu and pure Cu at 5% CO₂ feed concentration over a range of potentials. c The FE_C2+ of HB Cu, MB Cu and LB Cu during CO₂ electroreduction process at 5% CO₂ feed concentration. d Faradaic efficiencies (FE) for C₂₊ products produced by HB Cu at 30%, 60% and pure CO₂ feed concentration over a range of potentials. Curves connecting data points serve as visual guides. e Faradaic efficiencies (FE) for C₁ and C₂₊ products as well as current densities (j) for CO₂ electroreduction produced by HB Cu, MB Cu, LB Cu and pure Cu at pure CO₂ feed concentration over a range of potentials. f The Cu⁰/Cu¹⁺ interface boundary length per area, measured from the HRTEM images of each sample, plotted against the C₂₊ products selectivity. g Comparison of FE_C2+/FE_C1 ratios on HB Cu, MB Cu and LB Cu across various CO₂ concentrations. h Comparison of Faradaic efficiencies (FE) of different products produced by the three catalysts under potentials at which the maximum FE for C₂₊ products is reached. i Partial current densities for CO₂ electroreduction under pure N₂ atmosphere (dashed line) and 5% CO₂ feed concentration (solid line) conditions on HB Cu (red line), MB Cu (blue line), LB Cu (gray line). j Horizontal comparison of several factors for HB Cu, MB Cu and LB Cu. There is no iR correction for voltages. Error bars represent the standard deviation based on at least three separate measurements. Source data are provided as a Source Data file.

Figure 2g further demonstrates the influence of Cu⁰/Cu¹⁺ interface boundary density on product distribution, denoted as FE_C2+/FE_C1. In CO₂ purities below 30%, FE_C2+/FE_C1 increases with decreasing CO₂ purity, suggesting that dilute-purity CO₂ has a greater impact on product distribution compared to the Cu⁰/Cu¹⁺ interface boundary. Notably, we hypothesize that catalyst surface roughness may also impact CO₂ electroreduction properties. Therefore, we determined the electrochemically active surface area of the three samples by conducting cyclic voltammetry (CV) scans at various scan rates (Methods and Supplementary Fig. 39). The three samples, each with varying interface boundary densities, exhibited only minor differences in electrochemically active area. Interestingly, we found no correlation between the electrochemically active area and C₂₊ selectivity performance under the 5% CO₂ v/v feed condition. This further confirms the strong correlation between Cu⁰/Cu¹⁺ interface boundary density and C₂₊ selectivity.

Insights into CO₂RR under dilute purity feeds

Given that CO₂RR performance under dilute purity feeds is often hindered by mass transport limitations, which significantly impede electrocatalytic reaction rates, it is imperative to thoroughly evaluate electrocatalytic kinetics under such conditions. In this study, we analyze the reaction kinetics of the HB Cu catalyst by examining plots of log(j) versus potential (V vs. RHE). Figure 3a–c present the kinetics data across the entire range of CO₂ purity feeds for the primary electrocatalytic products—hydrogen, carbon monoxide, and ethylene. The calculated transfer coefficient (α = 2.303RT/F dlog(j)/dE) provides insight into the number of electrons transferred in the rate-determining and preceding steps, while the Tafel slope (dE/dlog(j)) identifies the rate-determining step (RDS) in the reaction process^106–108 (the raw data is presented in the Supplementary Figs. 40–43 and Supplementary Tables 4–7).

Fig. 3. Insights into CO₂RR mechanism from the reaction kinetics perspective and corresponding Operando Raman spectroscopy distinction.

Tafel plots of the HB Cu’s partial current density for H₂ production (a), CO evolution (b), and C₂H₄ production (c) under varying CO₂ purities in 1 M KOH at 25 °C. There is no iR correction for voltages. Operando Raman spectroscopy of HB Cu under pure CO₂ condition (d) and 5% CO₂ feedstock condition (e) over a range of potentials. f Peak *CO and *CO₂⁻ intensity during CO₂RR on HB Cu under 5% CO₂ feedstock condition (orange lines) and pure CO₂ condition (blue lines). g Proposed rate-determining reaction mechanisms for CO₂RR on rich Cu⁰/Cu¹⁺ interface surface(left) and poor Cu⁰/Cu¹⁺ interface surface(right). Source data are provided as a Source Data file.

Initially, we examine the kinetic characteristics of hydrogen production, which are particularly sensitive to disruptions under dilute CO₂ feeds. As depicted in Fig. 3a, the Tafel curves flatten as the CO₂ feed purity decreases, with the slopes decreasing from 251 mV dec⁻¹ to 93 mV dec⁻¹. The transfer coefficients calculated from these curves are less than 0.5, suggesting that the formation of surface-bound H* species is the rate determining step in H₂ evolution¹⁰⁹, regardless of the CO₂ feed purity (Detailed derivation is illustrated in the Supporting information). Notably, MB Cu and LB Cu,

which have sparser Cu⁰/Cu¹⁺ interface boundary densities, exhibit similar kinetic characteristics in H₂ evolution. Specifically, the transfer coefficients are less than 0.5 regardless of the CO₂ feed conditions, with H* formation dominating H₂ evolution (Supplementary Fig. 40 and Supplementary Table 7). This observation suggests that catalytic sites at the Cu⁰/Cu¹⁺ interface boundaries contribute to regulating H* formation, and these interface boundaries facilitate the rate of H₂ evolution as CO₂ feed concentration decreases.

In our experimental setup, the kinetics of CO production demonstrates distinctive behavior in contrast to H₂ production (Fig. 3b). This is analyzed by examining a mechanistic sequence that entails the rate-determining formation of surface-bound *CO₂⁻ species, which serves as the initial intermediate in CO production. As the CO₂ feed purity reduced from pure to 2.5%, the Tafel slopes of HB Cu dramatically decrease from 112 mV dec⁻¹ to 39 mV dec⁻¹, corresponding to an increase in the transfer coefficient from 0.5 to 1. These kinetics data reveal an overall increase in the rate of CO production as the CO₂ feed concentration decreases. Specifically, the Tafel slopes all fall below 57 mV dec⁻¹ at CO₂ feed concentrations below 7.5%, indicating a shift from the initial rate-determining CO₂ activation step to the rate-limiting *CO₂⁻ hydrogenation. Upon reducing the CO₂ feed purity from pure to 5%, both MB Cu and LB Cu samples showed Tafel slope changes from 133 mV dec⁻¹ to 107 mV dec⁻¹ and 192 mV dec⁻¹ to 110 mV dec⁻¹, respectively (Supplementary Fig. 40). These values all correspond to a transfer coefficient close to 0.5, suggesting that the initial CO₂ activation step constantly remains RDS during CO formation on Cu catalysts with sparser Cu⁰/Cu¹⁺ interface. These observations indicate that the Cu⁰/Cu¹⁺ interface can notably facilitate CO₂ activation, leading to a tendency to revert the rate-determining step, particularly evident in the HB Cu catalyst with a high proportion of Cu⁰/Cu¹⁺ interface. This, in turn, improves the conversion rate of essential carbon intermediates (notably *CO₂⁻ and *COOH), thereby establishing a robust basis for subsequent C-C coupling reactions.

Although ethylene evolution shares a common *CO₂⁻ intermediate with CO production in the early stages of electroreduction. To differentiate it from the CO formation pathway and ensure the independence of C₂₊ product evolution route, we define the formation pathway of C₂₊ product as originating from *CO. As illustrated in Fig. 3c, the Tafel slopes of C₂H₄ catalyzed by HB Cu exhibit two distinct regions. Under fully saturated CO₂ conditions, C₂H₄ evolution displays multiple parallel Tafel slopes of approximately 192 mV dec⁻¹ with a transfer coefficient of less than 0.5. This indicates that the chemical coupling of two surface-bound carbon-based intermediates predominantly governs ethylene formation, consistent with earlier findings^45,110–115. When CO₂ purity decreases, we note a series of noticeably flattened Tafel slopes at 85 mV dec⁻¹, along with a corresponding transfer coefficient nearing 0.5. This suggests an increased reaction rate while maintaining the same rate-determining process. Interestingly, as the CO₂ feed purity decreased from pure to 5%, we observed a decrease in the Tafel slopes of both MB Cu and LB Cu, from 146 mV dec⁻¹ to 113 mV dec⁻¹ and from 158 mV dec⁻¹ to 142 mV dec⁻¹, respectively, showing comparable downward trends (Supplementary Fig. 40). The results suggest that the Cu⁰/Cu¹⁺ interface boundary marginally enhances the kinetics of C-C coupling under low CO₂ concentrations. However, the enhancement is not significant.

In order to identify essential reaction intermediates and validate the kinetic evolution mechanism in CO₂ electroreduction under low CO₂ concentration conditions, we performed operando Raman experiments on the three Cu-based catalysts, applying potentials ranging from −0.11 to −1.51 V vs. RHE. Figure 3d, e depict the operando Raman spectra of CO₂RR on HB Cu under both pure and 5% CO₂ feeds. A series of bands ranging from 140 to 290 cm⁻¹ and a peak at 524 cm⁻¹, identified as characteristic peaks of Cu₂O, were consistently observed over HB Cu, indicating the persistent presence of the Cu₂O(111) facet and interface boundary during catalysis (in contrast, the Cu₂O peaks for MB Cu and LB Cu were less persistent. Refer to Supplementary Figs. 44, 45 for operando Raman spectra of MB Cu and LB Cu). In pure CO₂ feed conditions, the peak at approximately 2090 cm⁻¹, attributed to the stretching vibration of *C ≡ O (noted as ν_*C≡O), consistently maintains prominence across the entire cathodic potential range. This suggests the buildup of *CO, aligning with a product distribution featuring increased selectivity for CO under high-concentration CO₂ feed conditions. Nonetheless, the decrease in the ν_*C≡O peak below −0.31 V vs. RHE under a 5% v/v CO₂ feed condition suggests the swift depletion of *CO, aligning with an augmented selectivity towards C₂₊ products (Upper section of Fig. 3f). These findings corroborate the experimental results indicating an improved C₂₊ selectivity under dilute CO₂ feedstock conditions. Moreover, the positive correlation between CO peak intensity and feed gas concentration indicates a strong dependence of *CO surface coverage on CO₂ concentration. This insight can guide subsequent simulations of carbonate intermediate coverage on the catalyst surface. Additionally, the peaks at 1292 cm⁻¹ and 1590 cm⁻¹, attributed to the asymmetric stretching vibrations of *CO₂⁻ (ν_as*CO2-), exhibit greater prominence with 5% CO₂ compared to pure CO₂ conditions at low potentials (Lower section of Fig. 3f), confirming a shift of RDS to *COOH generation under dilute feeds, facilitating the accumulation of *CO₂⁻. Conversely, under pure CO₂ conditions, the formation of *CO₂⁻ is limited by the rate and challenging to preserve, aligning with the kinetic trends inferred from the Tafel slope. The pronounced weakening of the ν_as*CO2- peak with increasing potential can be ascribed to limitations in reactant transfer (Fig. 3g, Detailed derivation of the rate-determining step evolution is provided in the Supporting Information).

We aimed to understand the thermodynamic impact of the Cu⁰/Cu¹⁺ interface boundary under CO₂ feed conditions with varying purities through DFT simulations. According to the Henry’s law: local CO₂ concentration is proportional to local CO₂ partial pressure; local CO₂ partial pressure is proportional to CO₂ surface coverage. It is reasonable to express local CO₂ concentration in terms of the coverage of key carbonate intermediates, such as *CO₂⁻ and *CO^{54,110,116–119}. Based on the unstable adsorption of *CO₂⁻ species observed in simulation and the strong correlation between *CO peak intensity and CO₂ feed concentration in Operando Raman measurements. *CO coverage was ultimately selected to represent CO₂ feed concentration in this experimental system. Furthermore, since the *CO peak position in Operando Raman spectra changes gradually with applied potential, we infer that the active sites for *CO adsorption are in a dynamic adsorption state, and CO poisoning does not occur on the catalytic surfaces. We then investigated the coverage effect by computing the Gibbs free energies of reaction at low coverage (1/30 monolayer, ML) and high coverage (14/30 ML). Here, 1/30 ML and 14/30 ML represent the surface coverage of the crucial intermediate *CO under 5% and pure CO₂ feed conditions, respectively^120–122 (Fig. 4a and Supplementary Figs. 46–49). Thus, stable models of Cu(111), Cu₂O(111) surfaces, and symmetric low-angle tilted Cu⁰/Cu¹⁺ interface boundaries at both high and low coverage were constructed based on HRTEM images. Initially, we examined the adsorption of *CO on the four models. In the Cu⁰/Cu¹⁺ interface model at low coverage, *CO predominantly binds to the boundary sites. We determined that the adsorption free energy of *CO on the Cu⁰/Cu¹⁺ interface is −0.052 eV, which is stronger than that on the Cu(111) surface (0.191 eV) but weaker than that on the Cu₂O(111) surface (−0.469 eV). At high *CO surface coverage, the adsorption free energy of *CO at the Cu⁰/Cu¹⁺ interface is 0.113 eV, which is thermodynamically more favorable than on the pristine Cu(111) surface, where it is 0.191 eV. Consequently, the Cu⁰/Cu¹⁺ interface exhibits stronger *CO binding than pure Cu, promoting the accumulation of the crucial intermediate *CO. To further validate the higher reactivity of the Cu⁰/Cu¹⁺ interface towards *CO, CO stripping experiments were conducted on three catalysts (Supplementary Fig. 18). The results demonstrated that increasing in Cu⁰/Cu¹⁺ interface boundary density on the Cu-based catalysts effectively reduces the overpotential required for *CO oxidation.

Fig. 4. DFT calculations for different surfaces regarding the effect of *CO coverage.

a Adsorption configurations and adsorption energies (ΔG_ads) of *CO at Cu(111), Cu₂O(111), Cu⁰/Cu¹⁺ interface boundary at low *CO coverage and Cu⁰/Cu¹⁺ interface boundary at high *CO coverage. b The energy profiles of CO₂ reduced to *CO and the subsequent C-C coupling at the Cu(111), Cu₂O(111) and Cu⁰/Cu¹⁺ interface boundary. c Reaction free energy difference between CO₂ activation, *CO₂⁻ hydrogenation and *COOH dehydrogenation with consequential C-C coupling on Cu⁰/Cu¹⁺ interface boundary at low (orange) and high *CO coverage (gray). The dark brick-red, grey and red spheres represent Cu, C and O atoms respectively. Raw computational and simulation data are provided as Source Data and Supplementary Data 1 files.

Subsequently, we investigated the CO₂ electroreduction pathway to OC*COH on different surfaces, a recognized route for generating C₂₊ products. Using this pathway, we calculated the Gibbs free energy change (ΔG) for each intermediate formation during CO₂ electrocatalysis on Cu, Cu₂O, and Cu-Cu₂O interface surfaces at 0 V (vs. RHE) and under low *CO coverage conditions (Fig. 4b). The ΔG value for *COOH formation on the Cu-Cu₂O interface (0.20 eV, compared to the 0 eV state for CO₂) was lower than that on the pure Cu surface (1.12 eV). Additionally, the ΔG value for C-C coupling on the Cu-Cu₂O interface is 0.59 eV, significantly lower than those on the Cu and Cu₂O surfaces. These observations demonstrate that C-C coupling predominantly takes place at the Cu⁰/Cu¹⁺ interface boundary, particularly under dilute CO₂ feed conditions. This finding aligns with our experimental results, indicating that the HB Cu catalyst, which has a more abundant Cu⁰/Cu¹⁺ interface, exhibits higher selectivity towards C₂₊ products. To illustrate the impact of *CO coverage on the Cu⁰/Cu¹⁺ interface boundary, we examined the energy changes along the CO₂ reaction pathway using the Cu-Cu₂O interface model. We observed a notable reduction in the activation energy for CO₂ to *CO₂⁻ conversion as the *CO coverage decreased from 14/30 ML to 1/30 ML. This trend mirrors the subsequent hydrogenation of *CO₂⁻ to *COOH (Fig. 4c and Supplementary Table 8), indicating that the Cu⁰/Cu¹⁺ interface boundary can thermodynamically promote the activation and electroreduction of CO₂ under conditions of low *CO coverage. Specifically, at high *CO coverage, the activation of CO₂ to *CO₂⁻ is the dominant factor controlling the overall reaction potential in the CO formation pathway, consistent with the kinetics inferred from Tafel slopes, which suggest that the activation of CO₂ to *CO₂⁻ is the rate-determining step. As *CO coverage decreases, the dimerization of carbon-containing intermediates becomes thermodynamically favorable, aligning with the kinetic behavior indicated by the reduction of the Tafel slope observed during ethylene formation. The thermodynamic promotion of CO₂ activation and reduction in CO₂RR becomes more evident with increasing Cu⁰/Cu¹⁺ interface boundary density. This is consistent with experimental findings demonstrating a positive correlation between the density of Cu⁰/Cu¹⁺ interface boundary and the performance of C₂₊ products under dilute CO₂ feed conditions.

To further investigate the electronic interaction between the Cu⁰/Cu¹⁺ interface and the crucial intermediate *CO absorbed on the surface, we performed differential charge analysis (DCA) on the Cu⁰/Cu¹⁺ interface and pure Cu (Cu(111) surface in this case) with adsorbed *CO. As illustrated in Fig. 5a–d, electrons are transferred from the two Cu surfaces to *CO intermediates. The calculated Bader charge reveals stronger electronic interactions for *CO adsorbed on the Cu⁰/Cu¹⁺ interface (gaining 0.49 electrons) compared to those on Cu(111) surface (gaining 0.47 electrons). This enhanced *CO adsorption facilitates C-C coupling, which aligns with our experimental results. Raw simulation data are provided as Supplementary Data 1 file. Furthermore, to assess the catalytic stability of grain boundary-rich catalysts under low-concentration CO₂ feeds, we conducted stability tests of CO₂RR in a 1 M KOH electrolyte with a 5% v/v CO₂ feed, observing stable CO₂ reduction by HB Cu for 28.5 hours (Fig. 5e, f). During electrolysis, we maintained the test potential at −0.71 V vs. RHE, the optimal potential for C₂₊ products. The partial current density of C₂₊ products remained stable at (35 ± 5) mA cm⁻², with a corresponding FE stable at (49 ± 3) % over the 28.5 h period. Based on our empirical observations, although a low-concentration CO₂ feed is more representative of real production conditions and significantly reduces feed gas purification costs. Optimization of the experimental set-up and the electrocatalytic process is necessary to achieve low energy consumption and high stability for industrial-scale CO₂RR application^123–125.

Fig. 5. DCA models for Cu(111) and Cu⁰/Cu¹⁺ interface and experiment setup & stability measurement of CO₂RR at 5% CO₂ feed concentration.

Plots of differential charge analysis for *CO absorbed on Cu(111) surface (front view: a top view: b) and Cu⁰/Cu¹⁺ interface (front view: c, top view: d). Blue, red and gray balls stand for copper, oxygen and carbon atoms, respectively. The yellow and blue isosurfaces correspond to the electron accumulation and depletion regions, respectively. e Schematic illustration of electrocatalytic equipment with the gas diffusion electrode for CO₂RR. f Stability measurement of CO₂RR for 28.5 h of electrolysis at an optimum potential of −0.71 V vs. RHE at 5% CO₂ feed concentration and current densities for each product. There is no iR correction for voltages. Source data are provided as a Source Data file. Raw simulation data are provided as Supplementary Data 1 file.

Discussion

Previous techno-economic studies have shown that electrolysis of low-concentration CO₂ to produce sustainable chemicals is economically advantageous, as it avoids the high energy demands of feed gas enrichment. This process holds significant potential for environmental remediation, chemical production and other applicative scenarios³⁸. However, poor product selectivity and low catalytic current densities (below the industrial benchmark of 200 mA·cm^–2) are the major challenges for industrial-scale low concentrations CO₂ electrolysis^123–125. To address these challenges: firstly, improving the catalysts’ ability to effectively capture low-concentration CO₂ and increase the local concentration of CO₂ at catalytic active sites is critical to improve catalytic performance^126–130. Secondly, enhancing the catalyst’s capacity to directly convert low-concentration CO₂ would significantly improve product selectivity. However, it is worth noting that although some encouraging results have been achieved in low concentrations CO₂ electrolysis^69,131, the field is still in its early stages with many challenges to be addressed. These include unclear thermodynamics and kinetics mechanisms and the instability of catalytic systems under low-concentration conditions^23,48,63.

In summary, we have devised a simple vacuum calcination approach to regulate the density of Cu⁰/Cu¹⁺ interface boundaries on Cu-based catalysts. By manipulating the crystalline interface, we successfully crafted the Cu-based HB Cu catalyst, which displayed a high interface boundaries density of 22.67 μm⁻¹. At a low flue gas concentration of 5% CO₂ feed condition, the catalyst HB Cu exhibited a FE_C2+ of (51.9 ± 2.8) % and robust stability during the CO₂RR. Kinetic mechanism analysis based on the Tafel slopes engineering reveals that, only at low concentration CO₂ feed conditions, the catalytic sites at the Cu⁰/Cu¹⁺ interface facilitate a backward migration of the rate-determining step (from ${CO}_2+e^{-}+*\to {}^{*}CO_2^{-}$ to ${}^{*}CO_2^{-}+H_{2}O\to {}^{*}COOH+{OH}^{-}\left(aq\right)$), facilitating the easy formation of carbon-containing intermediates, but intriguingly, this phenomenon does not occur with high purity CO₂ feed conditions. Moreover, based on thermodynamic calculations, the energy barriers for CO₂ activation and subsequent reduction at Cu⁰/Cu¹⁺ interface boundary sites are significantly reduced at low CO₂ purity comparing to high purity feed conditions. This kinetically and thermodynamically synergistic enhancement endowed Cu catalysts with abundant interface boundary sites with strong capability to directly catalyze low CO₂ gas concentrations. HRTEM engineering and CO₂RR performance investigated the correlation between crystalline interface boundary density and C₂₊ selectivity, while ECSA measurements and XRD analysis negated the impact of electrochemically active area and oxidation state content on the electrolysis performance. This study adjusts the density of the Cu⁰/Cu¹⁺ interface boundary using a straightforward method and reveals the distinctive activity of interface boundary sites under nontraditional catalytic conditions. It not only establishing a green and sustainable electrocatalysis strategy for C₂₊ hydrocarbons production with high selectivity at near-minimum concentrations of industrial exhaust CO₂ but also encourages the investigation of catalytic mechanisms in conditions more representative of real-world production scenarios, such as using dilute feedstocks.

Methods

Reagents

Commercial Cu nanoparticles (99.9%, 25 nm), Nafion solution (5 wt% in water), dimethyl sulfoxide (DMSO, 99.99%) and deuterated water (D₂O, 99.9 atom % deuterium) were purchased from Sigma-Aldrich. Cu(CH₃COO)₂·H₂O (purity ≥99.0%), H₄N₂·H₂O (85.0% of volume fraction) and KOH was purchased from Aladdin Ltd. Electrolyte solutions were prepared with deionized water (DIW), which was obtained from a Millipore Auto pure system. All chemicals were of analytical grade and used without further purification. Nitrogen (99.99%), hydrogen (99.99%) and carbon dioxide (99.99%) were purchased from Tianze gas, Inc.

Characterizations

The high-resolution transmission electron microscopy (HRTEM) was observed on a JEOL JEM-2800 at 200 kV. SEM measurements were performed with a JEOL JSM-7800F. Raman spectra were obtained by a confocal Raman spectrometer (Renishaw inVia-Reflex) with a laser wavelength of 633 nm. Electrochemical measurements were performed with the Corrtest instrument (CS350H) electrochemical measurement using a three-electrode system with Ag/AgCl (filled with 3 M KCl) as a reference and Pt plate as a counter electrode. Agilent 7890B gas chromatography (GC) was used to analyzing the gaseous products. 1HNMR spectra were recorded on Bruker DRX 400 Avance MHz spectrometer.

Synthesis of HB Cu, MB Cu and LB Cu

The HB Cu catalyst was synthesized from 99.9% pure Cu nanoparticles with an average size of 25 nm using the vacuum calcination method. In brief, 0.2 g of commercial σ-Cu nanoparticles was placed into a porcelain boat and inserted into a tubular furnace. The sample was evacuated to a pressure below 10 mmHg and heated to 200 °C at a heating rate of 5 °C/min under a vacuum for 30 minutes to produce HB Cu. The MB Cu and LB Cu catalysts were synthesized using a procedure similar to that for HB Cu, with the exception that the vacuum calcination times were extended to 90 min and 150 min, respectively.

Synthesis of the control catalysts

A series of additional controls were obtained by calcination as well as conventional hydrothermal synthesis: (1) Original commercial σ-Cu nanoparticles, without any calcination; (2) Pure Cu nanoparticles after H₂ reduction and (3) Cu-based catalysts with different Cu/Cu₂O ratios obtained after air calcination and typical synthesis. The pure Cu catalyst was synthesized from 99.9% σ-Cu nanoparticles with an average size of 25 nm using hydrogen calcination method. In brief, 0.2 g of commercial σ-Cu nanoparticles was placed into a porcelain boat and inserted into a tubular furnace. The sample was evacuated to a pressure below 10 mmHg, then pure hydrogen was introduced to atmospheric pressure, repeated 3-4 times, and heated to 200 °C at a heating rate of 5 °C/min under hydrogen atmosphere for 30 min to produce pure Cu. The Cu-based catalysts with different Cu/Cu₂O ratios: 1 Cu, 2 Cu and 3 Cu were synthesized using a procedure similar to that for pure Cu, except that the calcination atmosphere was air, and the calcination times were 10 min, 20 min and 30 min, respectively. As for the control catalyst 4 Cu (Cu₂O) obtained by the typical synthesis, in brief, 0.2 g of Cu(CH₃COO)₂·H₂O was dissolved in 40 mL of deionized water (DIW), and then 200 μL of H₄N₂·H₂O was added dropwise successively. The mixture was stirred rigorously for 30 min to prepare Cu₂O. The achieved yellow precipitate was washed with excess DIW and absolute ethanol for at least three times, which was then dried at 60 °C in an oven under vacuum for next use.

Preparation of the working electrodes

To prepare catalyst inks for the working electrodes, 10 mg of the as-prepared catalyst powder and 30 µL of a 5 wt% Nafion solution were dispersed in 1.0 mL of isopropyl alcohol solution and subjected to ultrasonication for 10 min. The resulting catalyst inks were subsequently applied to carbon paper (AvCarb GDS3250, obtained from Fuel Cell Store) with an area of 0.8 × 1.6 cm² (catalyst loading ~1.0 mg/cm²) using an airbrush and dried overnight at 60 °C in a vacuum oven.

CO stripping

CO stripping experiment was performed in a three-electrode system by employing a DONGHUA electrochemical workstation (DH7002A). The electrolyte was first bubbled with Ar₂ for 30 min, followed by 20 min of pure CO, while the potential of catalyst was held constant at ~0 V (vs. RHE), during which CO monolayer adsorption was developed on the catalyst surface. Extra CO was then expelled from the electrolyte by bubbling the solution with Ar₂ for 20 min. Immediately afterwards, CV sweep setup at 20 mV/s was carried out.

CO₂RR measurements

Electrochemical measurement was performed in a flow cell electrolyzer by employing a Corrtest instrument (CS350H). The flow cell electrolyzer comprises of two electrolyte chambers separated by an anion exchange membrane and a gas chamber. GDE coated with the prepared catalyst was used as the cathode (working electrode), while the Ag/AgCl (filled with 3 M KCl) electrode and the nickel foam were used as reference and anode (counter electrode), respectively. For this three-electrode experiment, the cathode-anode spacing was small enough that the solution contact resistance was negligible. Therefore, the potential (vs. Ag/AgCl) were converted to RHE values without iR correction:

$$E\left(vs.RHE\right)=E\left(vs.Ag/AgCl\right)+0.197V+0.0592\times pHV$$

During electrocatalysis, both gas and electrolyte keep flowing. The flow rates of CO₂ (1, 2, 4, 12, 24, 40 sccm) and N₂ gases (39, 38, 36, 28, 16, 0 sccm) are regulated by two mass flow controllers according to the gas ratios (2.5%, 5%, 10%, 30%, 60% 100% v/v), and the overall gas flow rate is controlled by another mass flow controller (40 sccm, Alicat), while the electrolyte flow rate is controlled by peristaltic pumps (1.1 mL min⁻¹, Chuangrui, BT100M/YZ1515X). The gaseous phase composition was analyzed online by gas chromatograph (Agilent 7890B). Meanwhile, the cathode electrolyte was collected and analyzed for the liquid product with a Bruker Avance DRX 400 MHz nuclear magnetic resonance spectrometer (1H NMR). Essentially, the diluted cathode electrolyte (500 µL) after electrolysis was mixed with 100 µL of the internal standard solution (25 ppm (v/v) dimethyl sulfoxide, D₂O) followed by vortex to be uniform. The 1D 1H spectrum is measured with a water suppression method. The FEs of the products is calculated by using the following formula:

$$FE\left(\%\right)=\frac{nxF}{Q}\times 100\%$$

where F is the Faradaic constant, n is the number of electrons transferred, x refers to the mole number of products, and Q is the total charge.

Theoretical methods

In our computational study, we utilized the density functional theory (DFT) as implemented in the Vienna Ab initio Simulation Package (VASP) to carry out all the calculations. The generalized gradient approximation (GGA) with the Revised Perdew-Burke-Ernzerhof (r-PBE) functional was selected to describe the exchange-correlation energy. The ionic cores were represented using the projected augmented wave (PAW) method, and the valence electrons were accounted for within a plane wave basis set, setting the kinetic energy cutoff to 450 eV. To address solvent effects, we employed the implicit solvation model available in VASP, allowing for the explicit calculation of the solvation energy Esol. Additionally, the DFT-D3 method was incorporated to account for van der Waals forces, which are typically underestimated in standard DFT calculations.

The optimization of geometries was conducted with a stringent convergence criterion of 0.05 eV/Å for the forces. For the Brillouin zone integration, we used a 2 x 2 x 1 Monkhorst-Pack grid for all calculations, and the bottom half of the atoms were kept fixed. The changes in free energy (ΔG) for each step of the CO₂ reduction reaction (CO₂RR) were determined using the computational hydrogen electrode (CHE) model. This model assumes that the chemical potential corresponds to the energy of half a molecule of gas-phase H₂ at the standard hydrogen electrode (SHE) potential of 0 V.

To consider the effect of the electrode potential U, ($G\left(U\right)=G\left(0V\right)-neU$) we adjusted the free energy by adding -eU for electron transfer steps, where e is the elementary charge, n is the number of transferred proton-electron pairs, and U is the applied potential. The Gibbs free energy was computed using the following formula:

$$△G=△E+{△E}_{ZPE}-T△S$$

Here, $△E$ represents the change in DFT energy, $△E_{ZPE}$ is the zero-point energy, $△S$ is the entropy at 298.15 K, and T is the absolute temperature (in Kelvin).

For the Cu⁰/Cu¹⁺ interface boundary model, we constructed a surface consisting of Cu(111) (44) and Cu₂O(111) (12) units, forming a (6 × 5) Cu surface structure with specific lattice parameters: a = 16.396, b = 12.396, c = 23.581; α = β = 90°, γ = 120°. The lattice mismatch ratios were 8.55% along the a-axis and 2.58% along the b-axis. The surface coverage of *CO was set to 1/30 ML and 14/30 ML to simulate dilute (5% CO₂ v/v) and pure reactant feed conditions, respectively. The reaction free energy for converting CO₂ to *CO on the catalyst surface under varying *CO coverages was simulated based on the following reaction scheme:

$${CO}_2+e^{-}+*\to {}^{*}CO_2^{-}$$

$${}^{*}CO_2^{-}+H_{2}O\to {}^{*}COOH+{OH}^{-}\left(aq\right)$$

$${}^{*}COOH+H_{2}O+e^{-}\to {}^{*}CO+{OH}^{-}\left(aq\right)$$

This approach allowed us to investigate the reaction energetics under different conditions and provide insights into the catalytic performance.

Operando Raman experiments

Operando Raman measurements were conducted in a commercial spectro-electrochemical cell with a three-electrode configuration. To prepare the customized working electrodes, the obtained catalyst inks (10 mg/mL of each of the three catalysts: HB Cu, MB Cu and LB Cu) was uniformly dropped onto the carbon paper (AvCarb GDS 3250, purchased from Fuel Cell Store) with a 1.6 × 1.6 cm² area (catalyst loading ~1.0 mg/cm²) and dried overnight at 60 °C in a vacuum oven. The obtained carbon paper was then used as working electrodes, with a nickel tape and a saturated Ag/AgCl (filled with 3 M KCl) electrode as the counter electrode and reference electrode, respectively. Operando Raman experiments were conducted on a confocal Raman spectrometer (Renishaw inVia-Reflex) with a laser wavelength of 633 nm. The spectrochemical cell was coupled to a CHI 660E electrochemical analyzer for the electrochemical measurements. Before operando experiments, background spectra were collected without any applied potential. Subsequently, operando spectra were recorded at stepped potentials from −0.11 to −1.51 V vs. RHE in 1 M KOH. All spectra were recorded with a 4 cm⁻¹ resolution and 10 scans.

Author contributions

L.X. and Y.C. conceived the project. L.X. performed the initial preparation experiments and catalytic performance evaluations. Y.C. and W.Z. supervised research. L.X. wrote the manuscript. Y.J., M.S., J.L., J.Z. and W.Z. supervised the whole project and involved in manuscript preparation and revision.

Peer review

Peer review information

Nature Communications thanks Jing Li, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available within the paper and its Supplementary Information file or are available from the corresponding authors upon request. Source Data and Supplementary Data files are provided with this paper. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-54590-7.