Copper-stabilized bismuth subcarbonate electrocatalysts for durable large-scale formate production at kilowatt power

Abstract

Electrochemical CO₂ reduction (CO₂RR) to formate offers a sustainable route for decarbonization, yet achieving both high catalytic selectivity and stability remains challenging. Herein, we achieve a high formate Faradaic efficiency of 92.5% and full-cell energy efficiency of 53.8% at 30 A in a 100 cm² alkaline membrane electrode assembly (MEA) electrolyzer based on the designed bismuth subcarbonate stabilized with copper (Cu@BOC) catalyst. The electrolyzer demonstrates durable electrolysis at 30 A for 130 h, producing a formate yield of 98.5 mol in 24.5 L of electrolyte (4.02 M). In situ electrochemical spectroscopy measurements reveal the critical role of Cu dopants in stabilizing the catalyst structure from self-reduction, enhancing *OCHO adsorption, and suppressing the competitive hydrogen evolution. A scale-up coupled electrolysis system achieves a molar-scale formate production rate of 10.5 mol h⁻¹ at 2.1 kW and 100 A with a 5 × 100 cm² MEA electrolyzer stack. This rationally designed coupling electrosynthesis system with efficient formate production performance should significantly accelerate practical industrial applications of CO₂RR to formate.

Electrochemical CO₂ reduction to formate with both high catalytic selectivity and stability remains challenging. Here, the authors report high-performance CO₂ electroreduction for molar-scale formate electrosynthesis at kilowatt-scale power by stabilizing bismuth subcarbonate with copper.

Introduction

Electrochemical CO₂ reduction (CO₂RR) into value-added chemical products is one of the most promising carbon-neutral pathways^1–5. Among various CO₂RR products, formate has significant economic value and is widely explored as a chemical feedstock for numerous industrial processes, including its use as a potential fuel in formic acid fuel cells^6–10. However, the production rate of CO₂RR to formate electrolysis remains constrained to the millimole-per-hour range, and it is difficult to scale up the limited electrode area (typically <50 cm²) and low watt-scale power level (<10 W) of laboratory-scale systems for commercialization^11–13. This is because expanding the reaction electrode area inevitably introduces huge challenges related to electrocatalyst selectivity and structural stability^14–19. Therefore, maintaining high selectivity and stability for formate production is crucial for efficient operation of large-scale CO₂RR to bridge the gap between the laboratory research and industrial commercialization²⁰. Catalyst stability is the key to the scale-up process.

Metal oxides as catalysts for CO₂RR to formate exhibit significant activities and have been extensively studied^21–24. Recent studies have demonstrated that CO₂ molecules adsorbed onto the oxygen sites of metal oxides participate in surface reconstruction to form metal carbonates, which plays a key role in subsequent formate generation^23,25–27. During this reconstruction process, however, metal oxides tend to spontaneously reduce to the metallic state with increasing voltage and reaction time, especially under high current densities, thus enhancing the competitive hydrogen evolution reaction to compromise the CO₂RR selectivity and stability^28–31. Although various strategies, including the development of composite catalytic materials for stabilizing the metal oxide state, have been explored to address these challenges, their impact on the CO₂RR to formate performance remains limited^32–34.

On the other hand, the anodic oxygen evolution reaction (OER), commonly paired with CO₂RR to formate, suffers from high overpotentials and low-value products, limiting overall energy efficiency. Many researchers have recently focused on alternative anodic reactions, such as the hydrogen oxidation reaction (HOR)³⁵, glycerol oxidation reaction (GOR)^36,37, 5-hydroxymethylfurfural oxidation reaction (HMFOR)³⁸, and methanol oxidation reaction (MOR)^39,40. Among these, MOR stands out due to its low oxidation potential (0.103 V vs SHE)^41,42, technological maturity, and high Faradaic efficiency toward formate, making it well-suited for scale-up and industrial applications. Therefore, coupling CO₂ reduction (CO₂RR) with MOR to co-produce formate at both the cathode and anode not only lowers the full cell voltage and improves the energy efficiency, but also enhances the overall formate production rate. However, they are still limited to small-scale operation due to catalyst reconstruction, necessitating further research and development into catalytic selectivity and stability, process scalability, and integrated performance.

In this study, we selected bismuth subcarbonate (BOC) as the catalyst for CO₂ reduction to formate. The [Bi₂O₂]²⁺ oxide layers in the BOC provide abundant Bi-O coordination to optimize the adsorption of key *OCHO intermediates⁴³, while the intrinsic ${\mathrm{CO}}_3^{2-}$ layers promote the formation of surface carbonate species²³. Herein, we rationally designed a robust Cu-stabilized BOC (Cu@BOC) electrocatalyst for durable large-scale formate production via CO₂RR (Fig. 1a). In situ electrochemical spectroscopy measurements revealed that the Cu@BOC suppressed the structure damage at high current and enhanced *OCHO adsorption, thereby improving the CO₂RR to formate selectivity and catalytic stability. We assembled this robust catalyst into a 100 cm² membrane electrode assembly (MEA), achieving a maximum formate Faradaic efficiency of 92.5% and a full-cell energy efficiency of 53.8% at 30 A. The device further underwent a stability test at 30 A for 130 h, producing a formate yield of 98.5 mol in 24.5 L of electrolyte (4.02 M). To further increase the formate productivity and validate the commercial potential, scale-up CO₂ electrolysis coupled with MOR was conducted in a 5 × 100 cm² MEA electrolyzer stack, achieving a molar-scale formate production rate of 10.5 mol h⁻¹ at 2.1 kW and 100 A. These results significantly accelerate practical industrial applications of CO₂RR to formate.

Fig. 1. Catalyst design principles for CO₂RR to formate.

a Schematic of CO₂RR on BOC with and without heteroatom doping. b The formation energies of lattice oxygen vacancies in BOC doped with various transition metals. c The partial p-band density of states of Bi atoms for BOC and Cu@BOC structures. d The Gibbs free energy diagrams for CO₂RR on BOC and Cu@BOC. Source data are provided as a Source Data file.

Results

Catalyst design principles for CO₂RR activity

It is widely acknowledged that p-d orbital hybridization between transition metals and p-block metals can regulate the binding strength of key intermediates in electrochemical reactions, thereby optimizing reaction pathways (i.e., catalytic selectivity) and reducing activation energy barriers to enhance catalytic activity^44–47. To improve both the selectivity and activity as well as stability of BOC catalysts, we tune their electronic structures by doping with transition metals (Supplementary Data 1). Considering the smaller covalent radii of transition metal atoms compared to that of Bi atom, doping BOC with transition metals may introduce lattice distortions in the BOC^16,48, modifying the redox environment of the lattice to influence the formation of lattice oxygen vacancies and readjust the electronic structure at the active sites. Additionally, the electronegativity difference between transition metals and Bi atoms would induce charge redistribution^49,50; it is also expected to modulate the surface electronic density of BOC, stabilizing lattice oxygen and maintaining high-valence states of Bi to ultimately prolong the operational lifespan of the catalyst.

We first calculated the oxygen vacancy formation energies of BOC doped with transition metals of different atomic radii, and found that Cu doping leads to a higher oxygen desorption barrier, demonstrating the ability to maintain the lattice oxygen (Fig. 1b and Supplementary Fig. 1). Additionally, Cu-doping helps shift the first peak of the partial p-band density of states (DOS) under Fermi level to lower energy levels (Fig. 1c), which is beneficial for the weaker adsorption and better desorption of intermediates. Furthermore, we calculated the adsorption free energies of key intermediates involved in the CO₂RR process on the catalyst surfaces of Cu@BOC and BOC, respectively (Fig. 1d and Supplementary Figs. 2 and 3). The exothermicity results show that the first step involved CO₂ activation and *OCHO formation on Bi sites in the BOC is favorable. However, the subsequent desorption is prevented due to strong interaction between adsorbed *HCOOH and Bi sites. Due to the modulation of partial p-band DOS by doped Cu, the energy barrier shifts to be energetically favorable for Cu@BOC, indicating that the introduced Cu optimize the adsorption of reaction intermediates.

Catalyst synthesis and structural characterization

Cu@BOC electrocatalyst was synthesized by a one-pot hydrothermal method, followed by in situ electroreduction. Transmission electron microscopy (TEM) and energy-dispersive X-ray spectroscopy (EDX) elemental mapping images show ultra-thin nanosheets structure with Cu and Bi species evenly distributed, inheriting the morphology of the precursor (Supplementary Figs. 4–6). The X-ray powder diffraction (XRD) pattern exhibits that the catalyst maintains the original BOC structure, indicating that Cu doping does not affect its crystal structure (Supplementary Fig. 7). High-resolution TEM (HRTEM) imaging displays a lattice fringe with a spacing of 0.27 nm, corresponding to the (110) plane of BOC (Fig. 2a). The high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) images provide atomic-level imaging of the catalyst structure, confirming the presence of Cu atoms in Cu@BOC (Fig. 2b, c).

Fig. 2. Structural characterizations of Cu@BOC.

a HRTEM image, b AC HAADF-STEM image of Cu@BOC. c The intensity profiles along the region 1 and region 2 in (b). d Cu K-edge EXAFS spectra of Cu@BOC, along with the spectra of CuO and Cu foil as references. e Bi L₃-edge normalized XANES spectra with an inset highlighting the magnified XANES features near the absorption edge. f Bi L₃-edge EXAFS spectra of Cu@BOC, along with the spectra of Bi₂O₃ and Bi powder as references. Source data are provided as a Source Data file.

We further performed X-ray absorption spectroscopy (XAS)⁵¹ analysis to investigate the electronic structure and coordination environments of Cu and Bi atoms in Cu@BOC. The Cu K-edge X-ray absorption near edge structure (XANES) spectra display that the valence state of Cu atoms is slightly higher than +2 (Supplementary Fig. 8). The absence of a shoulder peak at 8988.5 eV, characteristic of the tetragonal symmetry of CuO, confirms that no CuO are present (inset of Supplementary Fig. 8). The Cu K-edge extended X-ray absorption fine structure (EXAFS) spectrum of Cu@BOC shows only one dominant peak at 1.5 Å attributed to Cu-O bonds, while Cu-Cu bonds at 2.2 Å (metallic Cu) and Cu-O-Cu bonds at 2.5 Å (CuO) are not observed (Fig. 2d). The Wavelet transform (WT) analysis (Supplementary Fig. 9) further corroborates the EXAFS results. The EXAFS fitting results show that the coordination number (CN) of Cu-O in the Cu@BOC is 5, higher than the CN of CuO (4) (Supplementary Fig. 10 and Supplementary Table 1). These findings confirm the presence of atomically dispersed Cu coordinated with O atoms in Cu@BOC. Further, the Bi L₃-edge XANES spectra show that the edge position of Cu@BOC closely aligns with that of Bi₂O₃, indicating the oxidation characteristics of Bi atom (Fig. 2e). A negative shift of Cu@BOC edge compared to the edges of BOC and Bi₂O₃ suggests a slight reduction in the oxidation state of Bi (inset of Fig. 2e). The Bi L₃-edge EXAFS spectrum displays two peaks at 1.6 and 3.5 Å, corresponding to Bi-O and Cu-O-Bi bonds, respectively (Fig. 2f), the latter of which is also observed in the Cu K-edge EXAFS spectrum (Fig. 2d). EXAFS fitting results reveal that the CN of Bi-O in Cu@BOC (2.7) is lower than that in BOC (3.1), due to the disruption of BOC structural symmetry by Cu doping (Supplementary Figs. 11, 12 and Supplementary Table 1). Compared to the Bi-O-Bi bonds in BOC, Cu-O-Bi bonds in Cu@BOC is shortened (Supplementary Table 1), indicating enhanced covalency of Cu-O-Bi bonds that prevent the Cu@BOC from structure damage.

CO₂RR electrolysis performance

Electrochemical performance of CO₂RR over Cu-doped BOC catalysts were evaluated in a flow cell using 1.0 M KOH as electrolyte (Supplementary Fig. 13 and Methods). To avoid the impact caused by insufficient reduction, we pretreated the catalyst at a current density of 1 A cm⁻² for 1 h. Gas and liquid products were separately analyzed using gas chromatography (GC) and nuclear magnetic resonance (NMR), respectively. We also synthesized catalysts with varying Cu doping levels (designated as Cu_L@BOC (low) and Cu_H@BOC (high)), and confirmed the Cu doping contents via inductively coupled plasma optical emission spectrometry (ICP-OES) (Supplementary Table 2 and Supplementary Fig. 14). As shown in Fig. 3a, b, formate was the only liquid product obtained during CO₂RR with all samples investigated in this study (Supplementary Fig. 15). Cu@BOC maintained a high Faradaic efficiency (FE) above 90% within 0.1–0.7 A cm⁻², reaching a maximum formate FE of 93.7% at 0.2 A cm⁻². When the current density further increased to 0.8–1.0 A cm⁻², formate FEs maintained above 84% (Fig. 3a). In contrast, BOC retained formate FE above 90% only within a lower current density range of 0.1–0.5 cm⁻². When the current density increased from 0.6 to 1.0 A cm⁻², formate FEs decreased from 72.6% to 45% (Fig. 3b). This discrepancy resulted in a significant difference in partial current density ( j_formate) for formate production, with the Cu@BOC achieving a maximum j_formate of 0.85 A cm⁻², 1.9 times higher than that of BOC, 1.2 and 1.5 times higher than Cu_L@BOC and Cu_H@BOC, respectively (Fig. 3c and Supplementary Fig. 16). As for the inferior performance of Cu_H@BOC with more doped Cu atoms, we attribute it to excessive Cu doping leads to the sacrifice of Bi active sites.

Fig. 3. Electrochemical CO₂RR performance.

The product distribution of CO₂RR at varying current densities on (a) Cu@BOC and (b) BOC. c Partial current densities of formate against potentials (versus reversible hydrogen electrode, RHE) over BOC and Cu@BOC. d Long-term stability test with Cu@BOC at 0.3 A cm⁻² during 200 h CO₂RR operation. Positions indicated by black arrows represent those of replacement of catholyte and anolyte. e Comparison of this work with state-of-the-art catalysts for CO₂RR to formate, including FE_formate, j_formate, durability and formate production rate. Error bars represent the standard deviation of three independent experiments. Applied potential were 85% iR corrected. The solution resistance (R): Cu@BOC, 1.02  ±  0.09 Ω; BOC, 1.23  ±  0.06 Ω. Details can be found in the Method section. Source data are provided as a Source Data file.

To investigate the stability of the catalyst, we firstly performed a long-term CO₂RR electrolysis at 0.2 A cm⁻², where the maximum formate FE was achieved (Supplementary Fig. 17). Clearly, there are no obvious decays observed in both formate FE (>90%) and potential for Cu@BOC, while the BOC shows a rapid decline toward formate FE, sharply descending below 60% after 2 h electrolysis. The post-reaction analysis reveal that the morphology and structure of Cu@BOC are well preserved after a continuous stability test (Supplementary Figs. 18, 19). In contrast, the appearance of metallic Bi formed on the BOC after the testing (Supplementary Fig. 20). Moreover, the long-term structural stability of Cu@BOC was demonstrated over 200 h of continuous electrolysis under 0.3 A cm⁻², with an average formate FE of ~90% (Fig. 3d). These results highlight the superior stability and structural integrity of Cu@BOC during long-term electrolysis. Notably, the developed Cu@BOC exhibits superior CO₂RR performance in FE_formate, j_formate, durability and formate production rate, outperforming the other previously reported Bi-based catalysts (Fig. 3e and Supplementary Table 3)^{16,21,30,43,52–54}.

Origin and mechanism of catalyst activity and stability

To investigate the origin and mechanism of the enhanced activity, in situ Raman spectra were collected through a custom-built electrolysis cell in CO₂-saturated 0.5 M KHCO₃ electrolyte (Supplementary Fig. 21). The potential-dependent in situ Raman analysis confirmed the exceptional structural integrity of Cu@BOC across an extended potential range from −0.4 to −1.4 V, exhibiting no detectable peak variations (Supplementary Fig. 22). Subsequent time-dependent in situ Raman analysis was conducted at a constant voltage of −0.8 V (Fig. 4a and Supplementary Fig. 23). Obviously, the interlayer Bi-O lattice vibration (67 cm⁻¹) maintained constant intensity, while the in-plane Bi-O stretching mode (161 cm⁻¹) displayed gradual attenuation―indicating minor bond reorganization within [Bi₂O₂]²⁺ layers without compromising the core Bi₂O₂CO₃ framework^23,55,56. Strikingly, BOC underwent rapid degradation under identical conditions. The Bi-O vibration (67 cm⁻¹) vanished within only 5 min, accompanied by progressive weakening of 161 cm⁻¹ band and the emergence of metallic Bi signatures at 70 and 98 cm⁻¹ characteristic of the E_g and A_1g stretching modes of Bi-Bi bonds, respectively^57,58. Moreover, time-dependent cyclic voltammetry measurements at −0.8 V demonstrated that Cu incorporation in Cu@BOC suppressed the excessive reduction peak observed in pristine BOC (Supplementary Fig. 24), further confirming its enhanced structural stability in agreement with the in situ Raman findings²⁴.

Fig. 4. Mechanism analysis.

a Time-dependent in situ Raman spectra for Cu@BOC and BOC at −0.8 V. b In situ ATR-FTIR spectra measured at a series of potentials for Cu@BOC. c Contour map of in situ ATR-FTIR spectra of Cu@BOC. d Integrated FTIR *OCHO peak areas as a function of potentials for Cu@BOC and BOC. Potentials applied were 85% iR corrected. The R for in situ Raman measurements were: BOC, 1.78  ±  0.30 Ω; Cu@BOC, 1.82  ±  0.33 Ω. The R for in situ ATR-FTIR measurements were: BOC, 2.35  ±  0.27 Ω; Cu@BOC, 2.46  ±  0.42 Ω. e The Gibbs free energy diagrams for CO₂RR on Cu@BOC, BOC and Bi (001). Source data are provided as a Source Data file.

To further elucidate the origin of the catalyst stability, we performed ex situ XPS and in situ XAS analyses. The Bi 4 f XPS spectra show the absence of Bi(0) peaks in Cu@BOC, in contrast to BOC (Supplementary Fig. 25), which is consistent with the in situ Raman results. As the formation of Bi(0) is typically associated with the depletion of lattice oxygen, O 1 s XPS spectra were also analyzed. The O 1 s XPS spectrum of Cu@BOC was deconvoluted into three peaks at 531.8, 530.4 and 529.5 eV, corresponding to hydroxyl-O, vacancy-O and lattice-O, respectively (Supplementary Fig. 26). Notably, the lattice-O/vacancy-O ratio in Cu@BOC was higher than that in BOC, demonstrating enhanced retention of lattice oxygen in Cu@BOC. To explain the enhanced retention, we further employed in situ XAS tests to probe the electronic structure under reaction conditions. The Cu oxidation state was observed to gradually decrease during the reaction and stabilized close to +1 (Supplementary Fig. 27a). In situ Cu K-edge EXAFS spectra confirmed the stable presence of Cu-O bonds, highlighting the role of doped Cu in anchoring lattice oxygen in Cu@BOC (Supplementary Figs. 27b, 28 and Supplementary Table 1). Collectively, the results from in situ XAS, in situ Raman, and ex situ XPS analysis demonstrated that Cu doping effectively stabilized the highly active Bi oxidation state via locking the lattice oxygen during CO₂RR.

To investigate the mechanism underlying the observed high formate selectivity, we conducted in situ attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR) (Fig. 4b and Supplementary Fig. 29). The peaks at 1394 cm⁻¹ attributable to intermediate *OCHO for CO₂RR to formate were observed over Cu@BOC across the whole range of applied potentials⁵⁹. No signals were detected at 1900–2100 cm⁻¹, indicating the absence of *CO⁶⁰ (Fig. 4c). Differently, BOC exhibited a distinct infrared vibration peak assigned to competitive *COOH at 1440 cm⁻¹ (Supplementary Fig. 29), indicating that BOC is poorly efficient for formate production. To elucidate the relationship between *OCHO coverage and CO₂RR to formate performance, we compared the FTIR *OCHO peak areas as a function of potentials for Cu@BOC and BOC (Fig. 4d and Supplementary Fig. 30). The *OCHO peak areas sharply increased from −0.8 V to −1.0 V, reflecting the facilitated *OCHO formation with increasing electron and proton availability^61,62, and then gradually decreased from −1.1 V to −1.4 V due to accelerated product desorption and competing hydrogen evolution⁶³. Notably, Cu@BOC exhibited a striking 2.7-fold higher adsorption coverage than BOC at −1.2 V―providing direct evidence of enhanced intermediate stabilization. These results reveal that Cu@BOC improved *OCHO coverage for formate production relative to BOC, consistent with its superior CO₂RR performance (Fig. 3c).

To explore the detailed reaction mechanism, we employed density functional theory (DFT) calculations to determine the reaction energy barriers for CO₂RR to formate on the catalyst surface (Supplementary Data 1). The calculation results (Supplementary Fig. 31) show that CO₂ reduction on Cu sites of Cu@BOC is impeded by a high reaction barrier of 3.3 eV (Supplementary Fig. 32), ruling out doped Cu atom as active site and suggesting it acts as a structural modifier. The Gibbs free energy difference (ΔG) between *OCHO and *COOH on Bi sites in the catalyst surfaces indicates that Cu@BOC exhibits a larger energy difference, which is favoring the adsorption of *OCHO over *COOH (Fig. 4e, Supplementary Figs. 33, 34), in line with the in situ ATR-FTIR results. The favorable adsorption of *OCHO also leads to the overall lowest uphill barrier on Cu@BOC among investigated BOC, Cu@BOC and Bi metal models, demonstrating its optimal theoretical activity for CO₂RR to formate (Fig. 4e and Supplementary Figs. 35–37). Meanwhile, Cu@BOC demonstrates the highest energy barrier for hydrogen evolution among all the investigated models (Supplementary Fig. 38). These theoretical findings are in good agreements with the experimental results.

Scale-up electrolysis for formate production

To systematically validate the scalability of formate electrosynthesis via CO₂RR, we first evaluated Cu@BOC in a 4 cm² MEA (benchmark configuration, Supplementary Figs. 39, 40). This foundational analysis confirmed operational feasibility before transitioning to an industrially relevant 100 cm² MEA electrolyzer (Supplementary Fig. 41). Fig. 5a showed the electrolysis performance for CO₂RR to formate conversion across a total current range from 10 to 100 A. The formate FEs remained above 90% across 10–40 A, reached 92.5% at 30 A, and maintained robust value of 80.6% even at 100 A. Notably, we achieved a high full-cell energy efficiency of 53.8% and formate production rate of 0.5 mol h⁻¹ at a low cell voltage of 2.4 V under a total current of 30 A (Fig. 5b), outperforming the majority of current electrolysis systems (Supplementary Table 4). Considering the high overpotential associated with anodic OER and the limited economic value of O₂ as a product, we substituted the anodic reaction with MOR, enabling formate generation at both the anode and cathode⁶⁴. To decouple MOR contributions, we first evaluated formate FEs in a 4 cm² MEA with an Ar-purged cathode (Supplementary Fig. 42). Subsequently, full CO₂RR | | MOR operation was assessed in an industrially relevant 100 cm² MEA electrolyzer. In comparison to CO₂RR | | OER system, the coupled electrolysis achieved a reduction in cell voltage and an increase in formate production rate by up to 1.5 times (Supplementary Fig. 43). Furthermore, a stability test was conducted at a total current of 30 A in CO₂RR | | MOR system. Over 130 h of continuous electrolysis, the coupled system produced a formate yield of 98.5 mol in 24.5 L of electrolyte (4.02 M) (Fig. 5c and Supplementary Fig. 44). Additionally, scale-up CO₂RR electrolysis coupled with MOR was performed using a 5 × 100 cm² MEA electrolyzer stack (Fig. 5d). The coupled electrolysis system achieved a total current of 100 A at a kilowatt-scale power of 2.1 kW, delivering a maximum formate production rate of 10.5 mol h⁻¹ and a low energy consumption of 4.3 kWh kg⁻¹ (Fig. 5e, f). The lower voltage at 100 A may be ascribed to localized heating under high current operation, which accelerates ion transport and reduces ohmic resistance⁶⁵. Overall, this system not only demonstrates large-area operation of CO₂RR to formate within a 5 × 100 cm² MEA electrolyzer stack, but also achieves an order-of-magnitude increase in formate production rate compared to previous systems (Supplementary Table 5). Promisingly, the high formate production rate at kilowatt scale highlights the industrial potential of the coupled electrolysis system.

Fig. 5. Scale-up CO₂RR electrolysis.

a The product distribution of CO₂RR and cell voltage as a function of total current in a 100 cm² MEA electrolyzer. b Energy efficiency and formate formation rate of Cu@BOC under different total currents. c Stability test at a total current of 30 A during 130 h in a 100 cm² MEA electrolyzer. Positions indicated by black arrows represent those of replacement of anolyte. d Photograph of the 5 × 100 cm² MEA electrolyzer stack. e Formate formation rate and cell voltage as a function of total current in a 5 × 100 cm² MEA electrolyzer. f Power input and energy consumption in coupled electrolyzer under different total currents. The Cell voltage represents full-cell voltage measured without iR correction. Source data are provided as a Source Data file.

Discussion

In summary, we have rationally designed and developed a coupled electrolysis system based on the stabilized Cu@BOC catalyst for high-performance CO₂RR to achieve molar-scale formate production at kilowatt power. In situ spectroscopy measurements demonstrated that the Cu@BOC inhibited the self-reduction of catalyst to metallic Bi, enhanced the adsorption of the key *OCHO intermediate, and suppressed the hydrogen evolution to enhance the catalytic activity and stability. We assembled the Cu@BOC catalyst in an alkaline MEA with 100 cm², achieving a high formate FE of 92.5% and full-cell energy efficiency of 53.8% at 30 A. A scale-up CO₂RR coupled with MOR in a 5 × 100 cm² MEA electrolyzer stack achieved a maximum formate production rate of 10.5 mol h⁻¹ at 2.1 kW. Our system offers a viable strategy for formate electrosynthesis, paving the way to develop high-performance CO₂ reduction technologies for large-scale commercialization.

Methods

Chemicals

Copper(II) nitrate trihydrate (CuNO₃·3H₂O, AR, 99.0%, SCR), bismuth nitrate (Bi(NO₃)₃, 99.99%, Macklin), polyethylene glycol (PEG-200, 99.5%, Aladdin), aqueous ammonia (GR, 25.0–28.0%, SCR), nickel (II) nitrate hexahydrate (Ni(NO₃)₂·6H₂O, AR, ≥ 98.0%, SCR), iron (Ⅲ) nitrate nonahydrate (Fe(NO₃)₃·9H₂O, ≥99.99%, Aladdin), urea (CO(NH₂)₂, 99 at.% 13 C, Macklin), potassium bicarbonate (KHCO₃, 99.5%, Macklin), potassium hydroxide (KOH, ≥99.999%, Aladdin), ethanol (AR, ≥99.7%, SCR), isopropyl alcohol (≥99.9%, Aladdin), methanol (AR, ≥ 99.5%, SCR), deuterium oxide (D₂O, 99.9 at.% D, Macklin), dimethyl sulfoxide (DMSO, 99%, Macklin), Nafion (5 wt.%, D520; DuPont) were all purchased from commercial sources and used without further purification unless otherwise stated. Sustainion XA-9 solutions (5 wt.%, Dioxide Materials) were prepared by dissolving the powder in ethanol. Anion exchange membrane (FAB-PK-130, 130 μm thickness, Fumasep), Sustainion anion-exchange membrane (Sustainion X37-50 Grade FA, 50 μm thickness, Dioxide Materials), polytetrafluoroethylene (PTFE, 60 wt.%, DuPont) and carbon black (Vulcan XC-72R, Fuel Cell Store), Sigracet 28BC gas diffusion layer (GDL, 235 μm thickness, Fuel Cell Store), Ni foam (0.5 mm thickness, 110 ppi, Suzhou Sinero technology Co., Ltd) and IrO₂ loaded on Ti mesh (1 mg cm⁻², 0.25 mm thickness, Zhenjie New Materials Co., Ltd) were purchased and used as received unless otherwise noted.

Preparation of Cu@BiO_x precursor

1.9 g Cu(NO₃)₂·3H₂O and 3.2 g Bi(NO₃)₃ were dissolved in 40 mL deionized water with 178 μL PEG-200 to form a homogeneous suspension. 5 mL aqueous ammonia diluted with 5 mL deionized water was slowly dropped into the above suspension and stirred continuously for 30 min. Then, the blue suspension was transferred to a Teflon-lined autoclave. The hydrothermal program was set to 180 °C for 3 h, followed by cooling to room temperature (25 °C). The Cu@BiO_x precursor was collected via centrifugation several times and washed with deionized water and ethanol, subsequently dried in vacuum at 80 °C overnight. By varying the mass of Cu(NO₃)₂·3H₂O in preparation, samples BiO_x (0 g), Cu_L@BiO_x (1.0 g), Cu_H@BiO_x (3.9 g) were obtained, respectively.

Preparation of GDL

100 mg of Vulcan XC-72R carbon black was dispersed in 30 mL of isopropyl alcohol and 10 mL of deionized water, together with 50 μL of Sustainion XA-9 solution and 200 μL of 60 wt.% PTFE suspension. The mixture was mechanically stirred for 1 h to obtain a homogeneous ink, which was then sprayed onto one side of carbon paper. Subsequently, 200 µL cm⁻² of 1 wt.% PTFE emulsion was sprayed onto the coated surface using an air-pump-driven spray gun. Then, the carbon paper was dried at room temperature before heat treatment. Annealing treatment was performed in a muffle furnace in three steps: first at 150 °C for 30 min, then ramped to 260 °C and held for another 30 min, followed by a final heating to 350 °C for 20 min. The resulting carbon black loading (approximately 1.0 mg cm⁻²) was determined by weighing the carbon paper before and after spraying.

Preparation of Cu@BOC gas diffusion electrode (Cu@BOC GDE)

The Cu@BOC GDE was prepared via an in situ electroreduction method. Typically, 300 mg of Cu@BiO_x precursor and 60 mg of Vulcan XC-72R carbon black, 26.5 mL of ethanol, 500 µL of 5 wt.% Nafion D520 solution, and 3 mL of 1 wt.% PTFE dispersion were mixed and ultrasonicated for 3 h to form a homogeneous catalyst ink. The resulting ink was then loaded on a modified GDL (2 × 1.5 cm², details provided below) to fabricate the working electrode, and the catalyst loading of 3 mg cm⁻² was determined gravimetrically. Prior to use, the electrode was activated via chronoamperometry at 1 A cm⁻² for 1 h in CO₂-saturated 0.5 M KHCO₃ solution. The obtained electrode was used directly for subsequent CO₂RR test. The BOC, Cu_L@BOC, and Cu_H@BOC electrodes were prepared using the same procedure, except that different amounts of Cu salt were added during precursor synthesis.

Preparation of NiFeLDH/NF anode

The NiFe layered double hydroxide (NiFeLDH) anode supported on Ni foam (NiFeLDH/NF) was prepared via a hydrothermal method. Specifically, a piece of Ni foam (10 × 10 cm²) was ultrasonically cleaned with diluted 2.0 M HCl solution, ethanol and deionized water for 15 min each, and then dried for subsequent use. Ni(NO₃)₂·6H₂O (977 mg), Fe(NO₃)₃·9H₂O (339 mg) and CO(NH₂)₂ (1681.7 mg) were dissolved in 210 mL of deionized water under continuous stirring for 20 min. Then, the resulting solution, along with the pretreated Ni foam, was transferred into a 250 mL Teflon-lined stainless-steel autoclave and subjected to hydrothermal treatment at 120 °C for 12 h. After naturally cooling to room temperature, the obtained NiFeLDH/NF electrode was thoroughly rinsed with deionized water and ethanol (three times each), and then dried under vacuum at 60 °C overnight.

Material characterization

The morphologies of as-prepared samples were characterized by transmission electron microscopy (TEM) (HT-7700) and high-resolution transmission electron microscopy (HRTEM) (Tecnai G2 F20 S-TWIN). Atomic force microscopy (AFM) was obtained with MultiMode VEECO. The atomic-scale morphology of as-prepared samples was obtained by aberration-corrected scanning transmission electron microscopy (HAADF-STEM) (Titan ChemiSTEM). X-ray powder diffractometer (XRD) tests were performed on a X-pert Powder diffractometer with Cu-Kα radiation. X-ray photoelectron spectroscopy (XPS) data were collected on an ESCALAB with MARK Ⅱ Energy Collector. Ex situ X-ray absorption spectroscopy (XAS) data of as-prepared samples were collected at the 20U beamline of the Shanghai Synchrotron Radiation Facility (SSRF).

In situ ATR-FTIR measurements

In situ ATR-FTIR measurements were performed on a Nicolet iS50 FTIR spectrometer using a custom-made three-electrode electrochemical single-cell for electrochemical tests, equipped with MCT detector and a reflectance unit for the electrochemical cell at an incidence angle of 60°. For working electrode, 30 μL catalyst ink (5 mg mL⁻¹) was sprayed on a polished Si crystal with a uniform gold coating. The Ag/AgCl electrode and a 1 cm² Pt sheet were used as the reference electrode and counter electrode, respectively. The electrolyte was 0.5 M KHCO₃, with CO₂ continuously purged. All infrared spectra were acquired within the electrochemical reaction potential range of –0.8 to –1.4 V (vs RHE). For each spectrum, 32 scans were collected with a spectral resolution of 4 cm⁻¹, and the open circuit potential (OCP) was recorded for comparison.

In situ Raman measurements

In situ Raman measurements was performed on a fully automated micro-Raman spectrometer (XploRA PLUS) equipped with a high-precision automated scanning stage system (SCANplus 75 × 50 2 mm) and a CCD detector (Synapse Plus). The laser wavelength was 532 nm. Electrochemical measurements were performed using a custom-made electrochemical single-cell, equipped with a quartz window through which the laser was focused to detect the working electrode. The electrolyte was 0.5 M KHCO₃, with CO₂ continuously purged throughout the measurements. The working electrode was prepared by spraying the catalyst ink uniformly on the surface of a clean 1 × 1 cm² gold foil. The Ag/AgCl electrode and graphite rod were used as the reference electrode and counter electrode, respectively. Potential-dependent spectra were obtained at the OCP and between –0.4 to –1.4 V (vs RHE). Time-dependent spectra were recorded at a potential of –0.8 V, with measurements taken sequentially at OCP, 5, 15, 30, and 60 min.

In situ XAS measurements

Cu K-edge XAS measurements were performed at the 20U beamline of the Shanghai Synchrotron Radiation Facility. In situ XAS measurements were conducted in a custom-made electrochemical cell. The working electrode was prepared by uniformly spraying the catalyst ink onto carbon paper (3 mg cm⁻²), with a platinum wire and Ag/AgCl electrode used as the counter and reference electrodes, respectively. The electrolyte was 0.5 M KHCO₃, while CO₂ was continuously purged into the electrochemical cell at a flow rate of 10 mL cm⁻¹. During the XAS measurements, the absorption edge position (E₀) was calibrated using a Cu foil and the position of E₀ is defined as the point corresponding to the maximum value in the derivative curve of the XANES spectrum. XAS raw data were processed using the Athena module in the Ifeffit software package.

CO₂RR electrolysis

CO₂RR electrolysis experiments were performed at room temperature in a flow-cell separated by anion exchange membrane (FAB-PK-130, 3 × 3 cm²). The Cu@BOC catalyst with 1 × 1 cm² active area was used as the working electrode. The Ag/AgCl electrode and Ni foam (2 × 2 cm²) were employed as the reference electrode and counter electrode, respectively. The electrolyte employed is 1.0 M KOH solution (pH=14.1 ± 0.1). Both the cathode chamber (1 mL) and anode chamber (4 mL) were recirculated with the electrolyte using a peristaltic pump at 15 mL cm⁻¹, while CO₂ was continuously fed into the 1 mL gas chamber at 20 mL cm⁻¹.

An MEA electrolyzer with 100 cm² was assembled using two titanium flow field plates for CO₂ feeding at the cathode and aqueous solution feeding at the anode, respectively. The cathode, anion exchange membrane (Sustainion X37-50 Grade FA, 11 × 11 cm²) and NiFeLDH/NF anode were assembled with a torque wrench so that the catalyst is stressed uniformly. The cathodic flow field was fed with humidified CO₂ gas at a flow rate of 80 mL min⁻¹ through a mass flow controller. The anodic flow field was fed with 1.0 M KOH solution at a flow rate of 0.1 L min⁻¹ recirculated using a peristaltic pump. The electrolysis was carried out in the galvanostatic mode with an Autolab electrochemical station (PGSTAT 302 N equipped with 10 A booster) and a DC Power Supply (Tuodapu 6230, 0–30 V/ 0–500 A, 15 kW), both of which automatically collected and stored test data. The scale-up measurements using a 5 × 100 cm² MEA electrolyzer stack were carried out with the cathodic flow field fed with humidified CO₂ gas at a flow rate of 2 L min⁻¹ through a mass flow controller and anolyte fed at a flow rate of 0.5 L min⁻¹ recirculated using a peristaltic pump. For the coupled electrolysis with CO₂RR and MOR, the anolyte of 1.0 M KOH was replaced with 1.5 M methanol + 1.0 M KOH (pH=13.9 ± 0.1). In the MEA system, heat dissipation was achieved by cooling the anolyte, thereby maintaining the MEA temperature within the range of 40–45 °C. During the stability tests, to better collect the cathodic products, the cathode gas stream is passed through a cold trap to capture evaporated formic acid. The 1.0 M KOH and 1.5 M methanol + 1.0 M KOH solutions were prepared in deionized water at room temperature and used immediately. The FAB-PK-130 and Sustainion X37-50 Grade FA were pretreated by immersion in 1.0 M KOH for 48 h before use.

The Ag/AgCl reference electrode (saturated KCl solution) was calibrated prior to use. The calibration was performed using a platinum electrode as both the working and counter electrodes, with the Ag/AgCl electrode as the reference, in H₂-saturated 0.5 M H₂SO₄ electrolyte (pH=0 ± 0.04). Cyclic voltammetry (CV) was conducted at a scan rate of 1 mV s⁻¹, and the average potential at zero current was taken as the calibrated E_Ag/AgCl. For the flow cell, all the potentials applied were converted to RHE following the equation below:

$$E_{\mathrm{RHE}}=E_{\mathrm{Ag}/\mathrm{AgCl}}+0.197\mathrm{V}+0.0591\times \mathrm{pH}-0.85iR$$ 1

The R (solution resistance) was determined by potentiostatic electrochemical impedance spectroscopy (EIS) using the workstation’s FRA module, with a frequency range from 100 kHz to 0.01 Hz. For MEA electrolyzer, the Cell voltage represents full-cell voltage measured without iR correction.

Products analysis

Gas products were analyzed using an online gas chromatograph (Fuli 9790 PLUS) equipped with a thermal conductivity detector (TCD) and two flame ionization detectors (FID1 for C₁ gases and FID2 for C₂₊ gases). The actual gas flow rate was measured at the GC outlet using a bubble flowmeter, and this flow rate was used for quantification of gaseous products in each individual experiment. Liquid products collected from anolyte were analyzed by a Bruker AVANCE NEO 500 MHz nuclear magnetic resonance (NMR) spectrometer. 500 μL of electrolyte was mixed with 100 μL D₂O with 10 mM dimethyl sulfoxide as internal standard. The ¹H-NMR spectrum was measured with water suppression using a pre-saturation method.

The Faradaic efficiency for gas product was calculated as follows:

$${\mathrm{FE}}_{\mathrm{product}}=\frac{Q_{\mathrm{product}}}{Q_{\mathrm{total}}}\times 100=\frac{n\times m\times F}{Q_{\mathrm{total}}}\times 100$$ 2

where $Q_{\mathrm{total}}$ is the total consumed charge, $Q_{\mathrm{product}}$ is the charge used for the formation of the product, n is the number of electrons transferred to form product, $m$ is the amount of the product, F is the Faraday constant, which is 96485 C mol⁻¹.

The partial current density of a specific product is calculated as follows:

$$j_{\mathrm{product}}=j_{\mathrm{total}}\times {\mathrm{FE}}_{\mathrm{product}}$$ 3

where $j_{\mathrm{total}}$ is the total current density.

The formate production rate is calculated as follows:

$$\mathrm{Formate}\mathrm{production}\mathrm{rate}=\frac{I_{\mathrm{total}}\times {\mathrm{FE}}_{\mathrm{formate}}\times 3600}{n\times F\times A}=\frac{I_{\mathrm{formate}}\times 3600}{n\times F\times A}$$ 4

where $I_{\mathrm{total}}$ is the total current, $A$ is the electrode area.

The energy efficiency is calculated as follows:

$$\mathrm{EE}=\frac{E_{\mathrm{OER}}^0-E_{{\mathrm{CO}}_2/\mathrm{formate}}^0}{E_{\mathrm{cell}}}\times {\mathrm{FE}}_{\mathrm{formate}}$$ 5

where $E_{\mathrm{OER}}^0$ and $E_{{\mathrm{CO}}_2/\mathrm{formate}}^0$ are the theoretical potentials for oxygen evolution and CO₂RR to formate, respectively. $E_{\mathrm{cell}}$ is the applied cell voltage.

The power (kW) is calculated as follows:

$$P=E_{\mathrm{cell}}\times I_{\mathrm{total}}$$ 6

The energy consumption (kWh kg⁻¹) is calculated as follows:

$$\mathrm{Energy\; consumption}=\frac{E_{\mathrm{cell}}\times n\times F}{{\mathrm{FE}}_{\mathrm{formate}}\times M_{\mathrm{formate}}\times 3600}=\frac{E_{\mathrm{cell}}\times I_{\mathrm{total}}}{R_{\mathrm{formate}}\times M_{\mathrm{formate}}}$$ 7

where $R_{\mathrm{formate}}$ is the formate production rate

DFT calculations

All density functional theory (DFT) calculations were conducted using the Vienna Ab Initio Simulation Package (VASP)⁶⁶. The projector-augmented wave (PAW) method⁶⁷ was applied to describe electron-ion interactions, while electronic exchange and correlation energies were treated with the Perdew-Burke-Ernzerhof (PBE) functional⁶⁸ within the generalized gradient approximation (GGA)⁶⁹. A kinetic energy cutoff of 450 eV was set, with convergence thresholds for energy and force established at 10⁻⁵ eV and 0.05 eV Å⁻¹, respectively. For Brillouin zone integration, a 1  × 1 × 1 Monkhorst-Pack grid with special k-points was used for geometric optimizations, while a 2 × 2 × 1 grid was employed for electronic property calculations. To account for van der Waals interactions, the DFT-D3 method⁷⁰ was utilized. Bader charge analysis was performed using the code developed by the Henkelman group⁷¹. The (2 × 2) three-layer slab of BOC(001) model, Bi(001) and Cu@BOC(001) model were built as the models for DFT calculations. In each surface model, atoms in the bottom layer were fixed, while the remaining atoms were fully relaxed. A vacuum gap larger than 15 Å was introduced to prevent interactions between adjacent periodic structures.

Gibbs free energies were calculated for all the elementary reactions at 298.15 K and 1 atm. The Gibbs free energy was calculated through the following equation:

$$G=E+ZPE-TS$$ 8

where E was the total energy by the DFT optimization, ZPE was zero-point energy, S was entropy, and T = 298.15 K. For all calculations, the optimized structures are provided in Supplementary Data 1.

Author contributions

Y.H., Z.B., and L.D. supervised the project. H.Z. carried out the experiments, analyzed the data and wrote the manuscript. M.W. and Y.L. performed and analyzed the DFT calculations. W.L. and X.S. contributed to the AC HAADF-STEM measurements. N.L. and H.Z. performed XAS measurements. B.Y., Z.L. and L.L. provided the help in manuscript editing. All authors discussed the results and contributed to review and editing of the manuscript.

Peer review

Peer review information

Nature Communications thanks Guillermo Díaz-Sainz and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

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/s41467-025-67274-7.