Synthesis and characterization of Cu and bimetallic CuGa catalystsThe Cu and CuGa catalysts were synthesized via a one-step reduction procedure (see Method). By varying the feed ratio of Cu2+ and Ga3+, a series of CuGa catalyst-precursors were obtained. For instance, the notation of “Cu5Ga1” indicates a Cu2+ to Ga3+ molar ratio of 5:1 was used for synthesizing the catalyst precursor. We also determined the precise Cu/Ga ratio by inductively coupled plasma mass spectrometry (ICP-MS). The ICP-determined Cu/Ga ratios are typically lower than the feed ratio, which is likely resulted from the insufficient doping of Ga during the rapid nucleation and growth of Cu nanoparticles (Supplementary Fig. 1). Note, we chose the sample of Cu5Ga1 as the model catalyst for the following discussions owning to its high activity, vide infra.From the results of scanning electron microscopy (SEM), the introduction of Ga slightly increased the surface roughness of Cu nanoparticles (Supplementary Fig. 2), which is further confirmed by the measurement of electrochemical active surface area (Supplementary Fig. 3). Transmission electron microscope (TEM) images suggest that (Supplementary Fig. 4) both Cu and Cu5Ga1 exhibit similar particle size distributions ranging from 20 to 50 nm. Additionally, lattice spacing of 0.18 and 0.21 nm were identified for Cu (200) and Cu (111) in high-resolution TEM (HR-TEM) image of as-prepared Cu sample (Supplementary Fig. 5). Likewise, for the as-prepared Cu5Ga1, obvious crystal planes of Cu2O (110) and (200) with lattice spacing of 0.30 and 0.21 nm, respectively, can be observed on their corresponding HR-TEM images (Fig. 2a). Furthermore, electron energy loss spectroscopy (EELS) mapping (with multiple linear least squares fitting) of the as-prepared Cu5Ga1 confirmed the presence of Cu and Ga and their even distributions across the sample particles (Fig. 2a bottom). We also examined the morphology of these catalysts after CO2R assessments, and observed varying degrees of surface fragmentation in all catalysts. Notably, more pronounced degree of fragmentation and reorganization occurred in the CuGa catalysts, marked by the presence of finely spread small particles (Supplementary Fig. 6). Additionally, post CO2R, Cu predominately exhibits Cu (111) facet with high crystallinity, as evidenced by HR-TEM images (Supplementary Fig. 7). As for Cu5Ga1, a more diverse array of facets, including Cu (200), along with numerous disordered boundaries and defects in addition to the Cu (111) facets, is observable (Fig. 2b). We hypothesize that the introduction of rich boundaries and defects is a result of the Cu-Ga alloying and dealloying process during CO2 reduction. In general, these rich boundaries and defects are typically recognized as highly active sites for CO2R, especially for C2+ production54,55,56. Furthermore, EELS mapping of post-Cu5Ga1 confirmed that the presence of Ga on the catalyst surface (Supplementary Fig. 8).Fig. 2: Synthesis and characterization of Cu and bimetallic CuGa.HR-TEM images and EELS-mapping of the (a) as-prepared and (b) post-CO2R Cu5Ga1 catalyst. The PXRD patterns (c) and Cu LMM Auger XPS spectra (d) of as-prepared Cu and Cu5Ga1 catalysts. e High-resolution Ga 2p XPS of as-prepared Cu5Ga1. f Ex-situ XANES at the Cu and Ga K-edge of as-prepared Cu and Cu5Ga1 catalysts. g FT-EXAF spectra at Cu K-edge of Cu and Cu5Ga1 catalyst. h FT-EXAF at the Ga K-edge of the as-prepared and post-catalysis Cu5Ga1. Cu foil, Cu2O, CuO, and Ga2O3 are shown as reference. Post-catalysis XAS of Cu5Ga1was performed immediately after electrocatalytic COR experiment. Relevant source data are provided in the Source Data file.The powders X-ray diffraction (PXRD) pattern of the as-prepared Cu (Fig. 2c) exhibits three distinctive peaks at 43.4, 50.4, and 74.2°, attributing to Cu (111), (200), and (220), respectively. In contrast, new peaks at 36.5, 42.3, and 61.3° were observed in the as-prepared Cu5Ga1 sample. These patterns further confirm the presence of significant amount of oxidized Cu on the surface of Cu5Ga1, primarily in the form of Cu2O (PDF#77-0199). Notably, no apparent characteristic XRD peaks were not identified for Ga, neither metallic Ga nor oxidized forms of Ga. We attribute this to both the low quantity of Ga doping, as well as the atomic dispersions of Ga into the Cu2O lattices. Similar phenomenon was observed in other related systems20,35. X-ray photoelectron spectroscopy (XPS) was employed to reveal the electronic state of Cu of both Cu and Cu5Ga1 (Supplementary Fig. 9). As expected, we observed XPS peaks at 952.5 and 932.5 eV corresponding to Cu 2p1/2 and Cu 2p3/2, respectively. The shake-up peaks at 945 and 965 eV correspond to the surface native oxide layer of Cu. We analyzed the Cu LMM Auger spectra to evaluate the Cu valence state of the bulk samples of Cu and Cu5Ga1. As shown in Fig. 2d, no apparent Cu0 peak was observed in either Cu or Cu5Ga1. Besides, the oxidation state of Cu in Cu5Ga1 is predominantly +1, lower than that of the as-prepared Cu catalyst (+2). We tentatively attribute this difference to the higher oxophilicity of Ga, preventing Cu in CuGa from being oxidized to CuO under ambient conditions, consistent with the results observed by Raffaella et al.51. High-resolution Ga 2p and Ga 3d spectra were also collected (Fig. 2e, and Supplementary Fig. 10). Separately, the peak position of Ga 3d overlaps with that of K 3p, a common impurity expected in CO2R systems. Hence, we collected and analyzed Ga 2p spectra for more reliable interpretations of the catalyst composition. As shown in Fig. 2e, the two peaks at 1145 and 1119 eV observed for the as-prepared Cu5Ga1 can be assigned to Ga 2p1/2 and Ga 2p3/2, respectively.To gather more insights to the valence state and coordination environment of Cu in these two samples, both X-ray adsorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) were monitored. As shown in Fig. 2f, the Cu K-edge features of the as-prepared Cu and Cu5Ga1 are closed to the standard samples of CuO and Cu2O, respectively. Besides, the corresponding Fourier transform of EXAFS spectra (FT-EXAF) indicates a Cu-O bond length close to CuO (~1.49 Å) reference for the as-prepared Cu, and a bond length closer to Cu2O reference (~1.39 Å) for Cu5Ga1 (Fig. 2g), aligning with the above XPS analysis. Additionally, the Ga K-edge signal for Cu5Ga1 catalyst before and post COR revealed that Ga mostly exists in the form of Ga2O3. However, following COR, a new peak at ~2.1 Å appears in FT-EXAF spectra (Fig. 2h), indicating the formation of the bimetallic Cu-Ga composition during electrocatalysis, which we believe will quickly convert to Ga2O3 upon cessation of the reaction. Note that, we chose to conduct the operando experiment under COR due to its generally high data quality, i.e., no carbonate formation. However, as CO2 to CO is not the limiting step, we believe the data obtained is relevant for our analysis.Determining the intrinsic activity of CuGa catalyst towards CO2RThe as-prepared Cu and Cu5Ga1 catalysts were assessed for CO2R under identical conditions. We first conducted a series of CO2R measurements under different temperatures to explore the intrinsic activities of these two catalysts toward CO2R (H-cell, 0.1 M KHCO3, and −1.0VRHE). Note that temperature can affect CO2 solubility in aqueous electrolyte57, we thus carefully determined the temperature window to ensure that there were no apparent CO2 mass transfer limitations, at least at our test potentials (−1.0 VRHE). As shown in Fig. 3a, b, Cu5Ga1 exhibits a similar product distribution as Cu, including ethylene (C2H4), ethanol (C2H5OH), methane (CH4), formic acid (HCOOH), etc. However, Cu5Ga1 demonstrates enhanced selectivity and partial current density towards C2+ products and suppressed HER compared to the reference Cu catalyst at all tested temperatures. Furthermore, we plotted the trends of the selectivity, measured in terms of Faraday efficiency (FE) of each product as a function of the electrolyte temperature (Supplementary Fig. 11). We found that the FEs of both H2 and C2H4 decrease as the temperature decreases, while the FEs of CH4 and C2H5OH exhibit the opposite trend. These different trends may result from direct and/or indirect effects of temperature, such as changes in CO coverage, local pH, and surface structure, etc.58. As ethylene is the major C2+ product, we first plotted the logarithmic dependence of ECSA normalized partial current density of C2H4 with reciprocal of temperature to estimate the corresponding activation energy (Ea) based on the Arrhenius equation (2). The same analysis was conducted for H2 production as it is the major competing reaction.$${\mathrm{ln}}\left(i\right)=-\frac{{{\rm{Ea}}}}{{{\rm{R}}}}\left(\frac{1}{T}\right)+{\mathrm{ln}}\left(A\right)$$
(2)
where i is the partial current density of the specific product, Ea is the corresponding activation energy, R is the ideal gas constant, T is the reaction temperature, and A is the pre-exponential factor58. Consequently, as depicted in Fig. 3c, Cu5Ga1 exhibits a substantially lower Ea of 28.5 kJ/mol at −1 VRHE towards C2H4 production compared to that of Cu (33.1 kJ/mol at −1 VRHE) under identical conditions. Moreover, Cu5Ga1 also demonstrates much higher Ea for H2 production (39.7 kJ/mol at −1 VRHE) compare to that of Cu (27.3 kJ/mol at −1 VRHE, Fig. 3d). Taken together, one can anticipate that Cu5Ga1 can inhibit the competing HER during CO2R and facilitate the production of C2+ products, particularly C2H4.Fig. 3: Activation energy assessment and CO2R using CO2 streams with different partial pressures.CO2R current densities and FEs for all measurable products at different temperatures at applied potential of –1.0 VRHE in 0.1 M KHCO3 for Cu5Ga1 (a) and Cu (b). a, b No iR correction. The corresponding Arrhenius plots for C2H4 (c) and H2 (d) productions during CO2R. The ECSA for Cu and Cu5Ga1 in H-cell is 83.4 and 91.3 cm2, respectively. CO2R product distribution using different CO2 partial pressure at current density of 0.5 A cm–2 for Cu (e) and Cu5Ga1 (f), measured in flow cell using 1 M KOH as the electrolyte. gThe applied potential (left) and the ratio of FEC2+ to FEH2 (right) under different CO2 partial pressure for CO2R on Cu and Cu5Ga1 under different current density. 100% iR correction was applied to calculate the applied potential. The solution resistance (R): Cu-100, 3.23 ± 0.9 Ω; Cu-500, 2.61 ± 0.6 Ω; Cu5Ga1-500, 2.21 ± 0.8 Ω; Cu5Ga1-1000, 2.06 ± 0.4 Ω. Details about the recorded process can be found in the Method section. h The FEC2+ for CO2R on Cu and Cu5Ga1 catalysts at different current densities with different CO2 partial pressure. The error bars in a–d were standard deviations which determined from at least 3 independent experiments. Relevant source data for this figure are provided in the Source Data file.To further assess the performance of Cu5Ga1 under practical relevant conditions, we conducted CO2R in a flow cell configuration (Supplementary Fig. 12) and employed diluted CO2 streams. As shown in Fig. 3e, f, we present the product distributions of Cu5Ga1 and Cu at the same current density of 0.5 A cm−2, under different CO2 partial pressure using N2 as the inert balance gas. At 100% CO2 partial pressure, Cu5Ga1 (FE = ~68%) exhibits slightly lower C2+ selectivity compared to Cu (FE = ~80%), which can be attribute the smaller overpotential applied. However, upon reducing the CO2 partial pressure from 100% to 75%, the FEC2+ of Cu decreased rapidly (to ~40%), coupled with a significant increase in FECH4 (from ~5% to ~30%). This result indicates that flooding likely occurred at the GDE surface. Notably, the applied potential was recorded to be approximately −0.87 VRHE (Fig. 3g left) for Cu at 75% CO2 partial pressure, aligning with the mass transport region we proposed earlier (Fig. 1b). When we further reduce the CO2 partial pressure to 50% and 25%, H2 became the predominate product for Cu, as anticipated (Fig. 3g right). In contrast, Cu5Ga1 catalyst exhibits a distinct behavior. Specifically, as the CO2 partial pressure decreases, the FEC2+ increases instead of decreasing, reaching the highest FEC2+ (~83%) and FEC2H4(~53%) at 50% CO2, while the FEH2 only increases modestly from 4% to 8%. This intriguing trend further supports our hypothesis on the potential-dependent regions for electrode flooding, as the applied potential for Cu5Ga1 is only −0.7 VRHE at 50% CO2. When we further reduce the CO2 partial pressure to 25%, the applied potential increased significantly to compensate for the CO2 concentration overpotential. Consequently, we observed a significant decrease in FEC2+ (~60%), and increase in selectivity towards to CH4 and H2, indicative of flooding at the GDE reaction layer.To further validate the hypothesis on the potential-dependent electrode flooding behavior, we varied the applied potential for both Cu and Cu5Ga1 based on our preliminary categorization of different kinetic regions (Fig. 1b). As shown in Fig. 3g, we chose current densities of 0.1 and 0.5 A cm−2 for Cu, and current densities of 0.5 and 1 A cm−2 for Cu5Ga1 for this assessment, respectively. At the current density of 0.1 A cm–2, applied potentials for Cu at different CO2 partial pressure lie in the kinetic and mixed-control region, indicating flooding may not occur significantly. These applied potentials are similar to those applied to Cu5Ga1 at 0.5 A cm–2, as anticipated, we observed similar trends in the ratio of CO2R/HER (Fig. 3g right) and the changes of FEC2+ (Fig. 3h and Supplementary Fig. 13). To drive a higher current density, as depicted in Fig. 3g, the applied potentials for both Cu5Ga1(1 A cm−2) and Cu (0.5 A cm–2) quickly enter the mixed-control region (negative than −0.75 VRHE) and then the severely flooded region as CO2 partial pressure decreases. As expected, we observed a rapid decrease in both the ratio of FECO2R/FEH2 and the overall FEC2+ (Fig. 3g, h). These correlations indicate that the GDE reaction interface stability regarding flooding is most likely related to applied potential. Taken together, our results suggested that the introduction of Ga can significantly reduce the activation overpotential for the production of C2+ products, e.g., C2H4, and inhibit HER process. Benefiting from the reduced activation overpotential, the overall applied cathodic potential is sufficiently low in the case of Cu5Ga1 for achieving high current density for C2+ production while preventing sever electrode flooding.Exploring the origin of improved activity of the CuGa catalystIn addition to the morphological changes (Fig. 2d), we also observed that the presence of the Ga exhibits a noticeable effect on preventing Cu oxidation under ambient conditions, as evidenced by XRD, Cu 2p XPS, and Cu LMM Auger results (Supplementary Figs. 14 and 15). We believe this effect is likely due to the high oxophilicity and low electronegativity of Ga. Furthermore, we observed a substantial loss of Ga from the CuGa catalysts after CO2R, as evidenced by EELS fitting, Ga 2p XPS, and ICP-MS (Supplementary Figs. 8, 16, and 17). We propose that the Ga loss can be attributed to two factors. Firstly, it may occur during the alloying–dealloying process. Secondly, the hydrophobicity of the GDE electrode decreases upon reaction cessation, leading to the etching of the Ga2O3 layer by KOH. Nevertheless, we believe the residual minor Ga on the catalyst surface could still play a significant role during catalysis. We will delve into a detailed discussion on this aspect thereinafter.We also investigated the dynamic changes in Cu and CuGa catalysts under working conditions by in-situ XANES. Due to the insufficient signal-to-noise ratio caused by interfacial instability in alkaline CO2R reaction (i.e., carbonate formation) under practical relevant current densities, we opted to perform COR instead of CO2R, since the CO2 to CO step is most likely not the limiting step on both Cu and CuGa in our experimental conditions59. As shown in Fig. 4a, the Cu K-edge normalized absorption spectra of Cu5Ga1 and Cu catalysts at the OCV state reveal that in the as prepared Cu5Ga1 sample, Cu exists predominately in the state of Cu2O, whereas in the as synthesized Cu catalyst, it mostly exists in the state of CuO (Supplementary Fig. 18). Upon the application of a negative potential to initiate the COR, the Cu matrix of both catalysts undergoes a swift conversion from oxidized states to the metallic state during electrocatalysis. While previous reports suggest that the presence of Cu+1 can promote C2H4 selectivity and stability60,61, there is insufficient evidence to support the notion that Ga can help maintain the +1-oxidation state in our system. As depicted in Fig. 4b, FT-EXAF conversion reveals that at applied negative potentials (–0.48, and –0.63 VRHE), the length of Cu-Cu bond in Cu5Ga1 (~2.25 Å) slightly exceeds that of the standard Cu foil (~2.18 Å). As the potential becomes more negative, the Cu-Cu bond length returns to the standard length. However, obtaining high-quality in-situ signals for the Ga K-edge was not successful due to its low content and significant background fluorescence interference from Cu35,62. Nonetheless, Cu-Ga bond can be clearly observed in the same measurement after COR reduction (Fig. 2h). Taking into account the theoretical potentials for Ga3+/Ga0 (EGa2O3/Ga0 = –0.485 VRHE, EGaOOH/Ga0 = –0.493 VRHE, EGa(OH)3/Ga0 = –0.415 VRHE)63 and the above results, we reasonably speculate that the above potential-dependent change in Cu-Cu bond length is a result of alloying and dealloying processes. We believe that Ga tends to redeposit onto the Cu lattice during catalysis, consistent with similar observations made elsewhere51.Fig. 4: In-situ characterization of Cu and Cu5Ga1 during CO/CO2R.a In-situ XANES at the Cu K-edge of Cu5Ga1 catalysts during COR condition (1 M KOH, flow cell) with different applied potentials. 100% iR (4.3 ± 0.3 Ω) correction was applied to calculate the applied potential. b the corresponding FT-EXAF spectra at Cu K-edge of Cu5Ga1 catalyst. Cu foil is shown as reference. In-situ ATR-SEIRAS spectra of Cu (c) and Cu5Ga1 (d) during COR at different applied potentials. 100% iR correction was applied to calculate the applied potential. The solution resistance (R) for Cu and Cu5Ga1 is18.2 and 10.2 Ω, respectively. e Potential-dependent COad peak area on Cu and Cu5Ga1 in in-situ ATR-SEIRAS spectra. e The partial current density of CO (top), C2+ products(middle), and its sum (bottom) in CO2R reduction (1 M KOH, flow cell) under different applied potential. 100% iR correction was applied to calculate the applied potential. The solution resistance (R) for above analysis can be found in figure caption of supplementary Fig. S22. Details about the recorded process can be found in the Method section. f The free energy surface (FES) for CO2 reduction to CO and CO-CO coupling with Cu (100) surface with 0.0, 1/6, and 1/4 ML of Ga. Right sides are the structures of the *OCCO on the1/6 ML Ga and pristine Cu (100). Atom color code: Cu, dark yellow; Ga, light purple; C, black; O, red; H, white. Relevant source data are provided in the Source Data file.Attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) was employed to monitor adsorption behavior of *CO, a key intermediate for CO2/COR reaction. Previous work has suggested that the CO2 to CO steps is not rate-limiting on Cu for C2+ products59. Therefore, to eliminate the influence from the CO2 to CO steps, we carried out IR measurements directly under the COR conditions to monitor the *CO on Cu5Ga1 and Cu, ensuring a fair comparison. As shown in Supplementary Fig. 19a, b, as the cathodic potential increases from OCV, a distinct band in the range of 2000 -2100 cm–1, attributed to the atop-bound CO (COad) on the catalyst surface, emerged on both Cu5Ga1 and Cu. Note, the observed COad peak are well described by a commonly used model consisting of the “low frequency band” and the “high frequency band” (Fig. 4c, d)64,65. Assuming Beer’s law holds, the integrated COad peak area is directly proportional to the COad coverage66. In this case, one could observe that the COad coverage on Cu5Ga1 is higher than that of Cu (Supplementary Fig. 19c). Noted, other factors, such as surface enhancement efficiency, ECSA of catalysts, etc., will also affect this high “COad peak area”. On the other hand, the wavenumber of COad is more closely associated with the COad binding strength on a given catalyst67. As a result, we observed that the wavenumbers of both COad peaks on Cu5Ga1 consistently remain slightly higher than those on Cu (Supplementary Fig. 19d), indicating that the binding strength of COad on Cu5Ga1 is slightly lower than on the pristine Cu.To further understand the effect of CO adsorptions, we analyzed CO generation during CO2R and found that Cu5Ga1 indeed exhibited substantially higher selectivity and partial current density towards CO, particularly at low potentials compared to Cu (Supplementary Fig. 20). As shown in Fig. 4e, Cu5Ga1 consistently exhibits a higher partial current density (CO and C2+ combined) than that of Cu across a broad overpotential window, even when normalized to its higher ECSA. Given the minimal production of CH4 and HCOOH, we believe this enhanced intrinsic activity of Cu5Ga1 is most likely resulted from its weaker COad binding strength and more favorable COad activation coupling step.We then conducted density function theoretical (DFT) calculations on CuGa catalyst to further understand the improved intrinsic activity of the bimetallic CuGa catalyst. The corresponding optimized computational model structure of Ga on Cu (100) are provided in Supplementary Data 1. Despite the experimentally determined Ga/Cu ratio in post-Cu5Ga being ~1:20, we found that Ga doping at the topmost layer is energetically favored by 0.43 eV than at the subsurface, which is substantial. Therefore, a higher Ga concentration on the surface of CuGa catalysts should be expected. However, determining the actual coverage of Ga is challenging. Alternatively, we have considered four different Ga coverages on Cu surface: 0.0 ML, 1/6 ML, 1/4 ML, and 1/2 ML. On the pristine Cu (100) surface, we found that CO prefers to be adsorbed on the hollow site, which is more stable by 0.06 eV than the bridge site, and by 0.13 eV than the atop site. The calculated free energy barrier for CO-CO coupling is 0.76 eV, aligning well with the previous calculations68. In a recent work, the Ga-doping on Cu(111) surface was found to promote the CO adsorption52. We found that CO adsorption at the Ga site on Cu (100) is suppressed rather than promoted. The adsorption energy of CO at the Ga site was calculated to be 0.86 eV higher than that of at the Cu atop site on the pristine Cu (100) surface. Additionally, after structural optimization, all the CO molecules initially placed at the Cu-Ga bridge site or the 3Cu-Ga hollow site will diffuse to the Cu atop site. Therefore, the most favorable adsorption site for CO is expected to vary with the Ga coverage. With 1/6 ML Ga, CO can still park at hollow sites that are not adjacent to Ga. However, with 1/4 ML and 1/2 ML Ga, the hollow site becomes coordinated to at least one Ga atom, making the atop site becomes the favorable site for CO adsorption. In all, as the Ga coverage increases, we found a weakening in the adsorption of all the reaction intermediates (Fig. 4f). For instance, the adsorption free energy of CO on the pristine Cu (100) surface is –0.32 eV, and it is weakened to –0.23 eV, –0.04 eV, and 0.12 eV with 1/6 ML, 1/4 ML, and 1/2 ML of Ga, respectively. Besides, Bader charge analysis reveals that each Ga atom doped on Cu (100) surface will donate about 0.22 electrons to the surrounding Cu atoms due to the lower electronegativity of Ga compared to Cu. It has been previously demonstrated that the interaction between CO and Cu is governed by both the π-bonding and σ-repulsion69, with the latter being sensitive to the occupancy of Cu sp-band. Consequently, the electrons donated by Ga occupy more Cu sp-bands, resulting in higher σ-repulsion and, therefore, weaker CO adsorption.Owning to the weakening of CO adsorption, the free energy surfaces suggest that the surface can efficiently reduce CO2 to CO (Fig. 4f, Supplementary Fig. 21). As a result, we found that the effective barriers for CO-CO coupling to form OCCO are substantially lowered upon Ga-doping. Specifically, the reaction barriers are reduced by 0.08 eV, 0.10 eV, and 0.05 eV with Ga coverages of 1/6 ML, 1/4 ML, and 1/2 ML, respectively, suggesting that Ga-doping can promote the formation of C2+ products across a wide range of Ga coverages. We believe the weakened adsorption of CO leads to reduced activation energy required for reaching the transition state for CO-CO coupling or forming other activated CO states, resulting in enhanced activity towards C2+ products as we observed experimentally.Energy efficiency assessment of CO2/COR on Cu5Ga1
After pinpointing the enhanced activity of the CuGa catalysts, we proceeded to validate its performance for CO2/COR reaction in more practical relevant reactors. We first analyzed the product distributions for both CO2R and COR on CuGa (Supplementary Figs. 22–27) and compared them to those observed on pristine Cu to investigate the effect of Ga-doping on C2+ products selectivity. As shown in Supplementary Figs. 26, 27, we observed similar products distributions, i.e., ethylene-to-oxygenates ratios, on both Cu5Ga1 and Cu, for both CO2R and COR before flooding occurs. However, as the current density increased, this ratio on Cu5Ga1 only increased slightly, while the ratio on pristine Cu increased dramatically due to the decrease in liquid products. This is likely induced by electrolyte flooding caused by the rapidly increased cathodic potential for the pristine Cu. Separately, favored liquid C2+ products (~50%), especially for acetate (~20%), were observed in COR compared to CO2R (liquid C2+ products ~30%, acetate ~5%) on both CuGa and Cu. We believe this observation is likely attributed to the higher local pH and CO partial pressure during COR (Supplementary Note 1, Supplementary Table 4)70. Taken together, we believe that Ga-doping into Cu predominately facilitates the kinetic rate-limiting step (*CO activation) and consequently enhances the overall C2+ production activity. As this kinetic rate-limiting step precedes the selectivity-determining step for C2+ production, we did not observe an obvious difference in terms of C2+ selectivity and ethylene-to-oxygenate ratios between CuGa and Cu, for both CO2R and COR. Consequently, as shown in Fig. 5a, the optimal CuGa catalyst (Cu5Ga1) achieves high FEC2+ (> 80%) and low FEH2 (<~5%) (Fig. 5b) across a wide current density range of 0.8−1.3 A cm−2 in CO2R. The excellent performance of Cu5Ga1 is even more pronounced in COR, achieving more than 85% FEC2+ selectivity in the wide current density range of 0.5−1.5 A cm−2. Furthermore, we also found that, for both CO2/COR, CuGa catalysts achieve the highest FEC2+ at much lower overpotential compared to Cu. Notably, the decrease in overpotential is more pronounced in the case of COR. For instance, at small current densities (0.1−0.3 A cm−2) where flooding may not occur, or at least not severely, the reduction in overpotentials already exceeds 100 mV. (Fig. 5c, Supplementary Fig. 28).Fig. 5: Energy efficiency analysis of Cu and Cu5Ga1 during CO/CO2R.The FEC2+ (a), FEH2 (b) of Cu and Cu5Ga1 in CO2 (solid) and COR (hollow). c The FE C2+ distribution of two catalysts at different applied potentials in CO2/COR. d Partial current densities of C2+ products of Cu and Cu5Ga1 in CO2 /COR. e The normalized partial current density of C2+ as function of applied potentials for four catalysts in CO2R. f The EE1/2 for two catalysts at different current density under CO2R. a–e The CO2/COR tests were performed in flow cell with 1 M KOH as electrolyte, Ni foam as counter electrode, and the flow rate of electrolyte and CO2/CO feeding is 25 mL/min. g Stability test at −0.3 A cm−2 current density in flow cell for Cu and Cu5Ga1, the anode electrolyte was regularly renewed. a–g 100% iR correction was applied to calculate the applied potential. The solution resistance (R) for above analysis can be found in figure caption of supplementary Figs. S22 and S23. Details about the recorded process can be found in the Method section. h The current density for two catalysts at different cell voltages; purple area for COR with 2 M KOH as anode electrolyte; pink area for CO2R with 1 M KOH as anode electrolyte. h The CO2/COR tests were performed in MEA with different electrolytes, IrOx/Ti mesh as counter electrode, and the flow rate of CO2/CO is 25 mL/min. The error bars in a-f were standard deviations which determined from at least 3 independent experiments. Relevant source data are provided in the Source Data file.Our attention also extended to the production of H2 and CH4, two representative indictors of interface instability in Cu-based systems71. We found that ~ −0.75 VRHE emerged as a turning point in our measurements for both CO2/COR (Fig. 5d, Supplementary Figs. 29–31). When the applied potential went below ~ −0.75 VRHE, both the FEH2 and FECH4 quickly increased for all catalysts in both CO2/COR, indicating flooding occurred at the reaction interface. Additionally, the ratios of FEC2+ to FEC1+ (C1 includes both CO and CH4) at different potentials also exhibit the same trends (Supplementary Fig. 32). Overall, these observations align well with our hypothesis on potential-dependent electrode-flooding behavior, and the high intrinsic activity of CuGa catalysts ensures substantially enhanced current density can be achieved with sufficiently low cathodic potentials (Fig. 5e, Supplementary Figs. 33, 34). As a result, showing in Fig. 5f, Cu5Ga1 achieves an encouragingly high cathodic EE of >50% towards C2+ products at a high current density of 1 A cm−2, surpassing that of Cu (~30%) significantly. Even at a further increased current of 1.5 A cm−2, Cu5Ga1 maintained over 40% cathodic EE1/2 for C2+ products, while that of Cu rapidly dropped to less than 10%.Single-pass carbon conversion efficiency (SPCE) serves as another crucial indicator for evaluating the CO2R performance. To mitigate the severe CO2 loss in alkaline electrolyte, we assessed the CO2R performance and SPCE in acidic electrolytes (0.5 M K2SO4, pH = 2). As anticipated, Cu5Ga1 catalyst demonstrated substantially higher C2+ selectivity and lower overpotential at all current densities compared to those observed on Cu, confirming its enhanced performance observed in alkaline CO2R (Supplementary Fig. 35). Furthermore, at the optimal CO2 flow rate, an appreciable SPCE of 40% could be achieved at a current density of 500 mA cm−2 in Cu5Ga1-based CO2R in acidic electrolyte.Stability is equally crucial for CO(2)R. In the CO2R, Cu GDE rapidly flooded due to the higher cathodic potential required for 0.5 A cm−2, while the Cu5Ga1 GDE could last for significantly longer time under the identical operating current density (Supplementary Fig. 36). Considering that salt crystallization in the CO2R also contributes to destabilize of the GDE reaction interface. Hence, we focused on assessing the impact of electrode potential on the stability of COR (Fig. 5g). The results revealed that the Cu5Ga1-GDE can maintain ~90% FEC2+ at an industrial current density (0.3 A cm−2) for more than 120 h, while the Cu-GDE experiences system failure at ~20 h. Cu-based catalysts may undergo degradations due to dissolution, restructuring, and fermentations, which can lead to unsatisfactory stability for long-term COR electrolysis72,73. Therefore, to explore the origin of the instability of pristine Cu electrode, we carried out detailed analysis on the changes in ECSA, product distributions, and OH− adsorption of the two catalysts during and after COR for various time periods (Supplementary Figs. 37–39). As a result, we conclude that the electrolyte flooding is the predominant cause of the instability of pristine Cu during long-term COR electrolysis, despite a certain level of surface restructuring occurring in both catalysts.We proceeded to assemble a MEA reactor to minimize the ohmic overpotential and assess the performance of CuGa catalysts for CO2/COR by using different anode electrolytes. In the CO2R reaction, resembled those obtained in the flow-cell, the Cu5Ga1 catalyst demonstrated a low cell voltage towards CO2R products than that of Cu. Specifically, at a current density of 0.8 A cm−2, the cell voltage observed in Cu5Ga1-MEA was ~134 mV and 247 mV lower compared to those of Cu-MEA, in 0.1 M KHCO3 and 1 M KOH, respectively (Fig. 5h, Supplementary Figs. 40–43). In the COR reaction, both Cu and Cu5Ga1 exhibited enhanced activity compared to CO2R and achieved higher FEC2+ (~90%) across a wide current density range of 0.1–1.5 A cm−2 (Supplementary Fig. 44). More remarkably, benefiting from the lower cell voltage, Cu5Ga1 catalyst can achieve a high current density of 2.5 A cm−2 at a cell voltage of 3 V and still maintaining a ~70% FEC2+ (with a partial current density of C2+ of 1.6 A cm−2). As expected, the highest EE (~40%) of Cu5Ga1 in COR can be achieved at 0.8 A cm−2, and it can maintain ~30% at high current density (2 A cm−2). In contrast, Cu-MEA will quickly shift to the HER due to system instability at current densities exceeding 1.5 A cm−2. As for the EE in CO2R, Cu5Ga1 also maintains high full-cell EE at larger current densities, compared to Cu (Supplementary Fig. 45). Additionally, upon comparing the EE1/2 and full-cell EE achieved in the flow cell and MEA with the latest research results, we found that the Cu5Ga1 catalyst in this study demonstrated advantages in energy efficiency especially at high current densities (Supplementary Fig. 46 and Tables 5, 6).
Energy-efficient CO(2) conversion to multicarbon products at high rates on CuGa bimetallic catalyst
