All reagents were used as received without further purification.

Electrochemical tests were performed with a CHI 760e electrochemical Station (Shanghai Chenhua Instruments Company). Gaseous products were detected by gas chromatography (Shimadzu, GC-2014c) with a flame ionization detector (FID) and a thermal conductivity detector (TCD). Liquid products were detected using a nuclear magnetic resonance spectrometer (NMR, Ascend 400, Bruker, Germany). The micromorphology, crystalline structure, and element mapping were obtained by a field emission scanning electron microscope (FE-SEM, FEI JEOL-7800F) and a high-resolution transmission electron microscope (HR-TEM, JEM-2100F). The metal amount in the as-synthesized catalyst was detected using inductively coupled plasma-optical emission spectrometry (ICP-OES, OPTIMA2100DV). N₂ adsorption/desorption curves were achieved by a specific surface and porosity analyzer (Micromeritics ASAP 2460) and calculated using the Brunauer–Emmett–Teller (BET) equation. X-ray diffraction (XRD) patterns were recorded with an X-ray powder diffractometer (Rigaku MiniFlex 600) with Cu Kα radiation (k = 1.5406 Å). Raman spectra were acquired with a laser Raman spectrometer (LabRAM HR Evolution, HORBIA FRANCE SAS) with a 633-nm laser excitation. X-ray photoelectron spectra (XPS) were recorded on an X-ray photoelectron spectrometer (ThermoVG Scientific ESCALAB 250) with Al Kα X-ray as the source.

All the five samples in this study were prepared by electrospinning technology. The preparation steps are as follows: 7 ml of N, N-dimethylformamide, 0.5 g polyacrylonitrile (PAN), and 0.75 g ZIF-8 nanoparticles were put into a beaker and stirred until they were evenly mixed into a white viscous solution. Then, 0.2183 g of Co(NO₃)₂·6H₂O (0.00075 mol) and 0.061 g of Cu(NO₃)₂·3H₂O (0.00025 mol) were added, and stirring was continued for at least 20 h or until the mixture was fully mixed to obtain a purple viscous spinning precursor solution. This precursor solution was injected into the syringe and electrospun to polymer fibers. After spinning, the polymer fibers were put into vacuum drying oven at 60°C for at least 12 h, and the dried polymer fiber was pre-oxidized in a muffle furnace. Then those pre-oxidized fibers were carbonized in nitrogen atmosphere. The initial temperature was set at 25°C, raised to 900°C at the rate of 5°C/min, and maintained for another 2 h. The as-synthesized catalyst was named as Co₃Cu/CFs.

Another four catalysts with different metal ratios can be obtained by changing the molar ratio of metal precursors Co (NO₃)₂·6H₂O and Cu (NO₃)₂·3H₂O, including 1/0, 1/1, 1/3, and 0/1. The as-prepared samples were named as Co/CFs, CoCu/CFs, CoCu₃/CFs, and Cu/CFs.

All the five catalysts were powdered and drop-coated onto a carbon paper (SGL Carbon Corporate) to get a useful working electrode. CO₂ reduction activity was tested in a typical H-type electrochemical cell separated by an anion exchange membrane between anodic and cathodic chambers, with a Pt foil as the counter electrode and an Ag/AgCl as the reference electrode; 0.5 M KHCO₃ solution was employed as the electrolyte and bubbled with high purity CO₂ or N₂ (99.995%). The original potentials measured in this manuscript were converted to the reversible hydrogen electrode (RHE) via the Nernst equation: E ( RHE ) = E ( Ag / AgCl ) + 0.199 + 0.059 × pH . (1)

Products from CO₂ reduction were analyzed at various cathodic potentials with a fixed time of 15 min, and the gaseous components were directly injected into gas chromatography. The liquid-phase products were detected via ¹H NMR spectra. The Faraday efficiencies of the products were calculated via the following equations. Q is the total charge transferred through the working electrode at different potentials; m is the number of electrons transferred, which is 2 for HCOOH, CO, and H₂, and 8 for CH₄; n is the mole numbers of products; and F is the Faradaic constant (96,485 C mol⁻¹). FE = Q product Q total = m × n × F Q total (2)

The specific preparation process of the material is described in Figure 1. First, Co(NO₃)₂·6H₂O, Cu(NO₃)₂·3H₂O, ZIF-8 nanoparticles as well as PAN were dissolved in DMF to prepare a precursor solution, and then an electrospinning technology was used under constant conditions to get PAN nanofibers with different Co/Cu mole ratios. Then, the Co–Cu/PAN nanofibers were heated to 900°C under a N₂ atmosphere for carbonization. Notably, there are no extra surfactants or reductants involved in the whole procedure. The polymer linkers were pyrolyzed and carbonized to generate the main body of carbon nanofibers, and ZIF-8 nanoparticles collapsed to form the abundant mesopores and macropores through these nanofibers (Yang et al., 2020a). Co²⁺ and Cu²⁺ ions were reduced by organic linkers, and the bimetallic nanoparticles were engendered with a smaller particle size due to the confinement of the polymer and ZIF-8 nanoparticles. The compositions of these bimetallic nanoparticles were tuned by the feeding ratio of Co(NO₃)₂·6H₂O and Cu(NO₃)₂·3H₂O precursors, eventually generating Co₃Cu/CFs, Co/CFs, CoCu/CFs, CoCu₃/CFs, and Cu/CFs.

As shown in Figure 2, the surface morphology and nanostructure of the as-synthesized Co₃Cu/CF catalyst were recorded by FE-SEM and HR-TEM, respectively. The diameter of the carbon nanofibers in Co₃Cu/CFs ranges from 500 to 600 nm. The length is in the scale of hundreds of micrometers, and the interlaced nanofibers furtherly form a network structure (Figures 2A,B). Moreover, abundant hollow pores, in the size range of dozens of nanometers, could be easily seen throughout Co₃Cu/CFs (Figures 2C,D). N₂ sorption isotherms (Figure 3A) further demonstrate that Co₃Cu/CFs have type Ⅳ sorption isotherm, belonging to the mesoporous structure. Co–Cu nanoparticles, in an ∼20 nm diameter range, are evenly immobilized within the hollow pores of carbon nanofibers (Figures 2D,E).

In addition, Figure 2F shows the clear HR-TEM image of an independent Co–Cu bimetallic nanoparticle, and the interplanar spacing of crystalline lattices marked with red lines is measured as 2.04 Å and 2.46 Å, corresponding to the Co (111) and Cu (111) planes (Liu et al., 2014; Kim et al., 2017), respectively. Furthermore, the high-angle annular dark field STEM (HAADF-STEM) and elemental mapping images indicate that Co₃Cu/CFs contain Co and Cu elements. The good match between the emerging positions of Co and Cu elements directly proves the formation of Co–Cu bimetallic nanoparticles. According to the SEM, HR-TEM (Supplementary Figure S1,S2), and N₂ sorption isotherms (Supplementary Figure S3, Supplementary Table S1), Co/CFs, CoCu/CFs, CoCu₃/CFs, and Cu/CFs display similar interlaced nanofibers, abundant hollow pores, and uniform nanoparticles. Hence, the regulation of metal ratio would not significantly change the surface morphology and nanostructure. The actual Co/Cu ratios in the as-synthesized catalyst were detected by ICP-OES. CoCu/CFs, Co₃Cu/CFs, and CoCu₃/CFs own Co/Cu ratios of 1/0.95, 3/0.96, and 0.98/3, respectively, which are close to the original ratios in precursor solutions.

As shown in the XRD patterns (Figure 3B), the sharp diffraction peaks at 44.9°, 52.5°, and 76.2° can be seen in Co/CF samples, which are attributed to the Co (111), (200), and (220) planes (JCPDS 89-7093), respectively. Those peaks at 43.3°, 50.5°, and 74.2° of the Cu/CF samples are in accordance with the Cu (111), (200), and (220) planes (JCPDS 89-7093), respectively. The XRD patterns of Co₃Cu/CFs, CoCu/CFs, and CoCu₃/CFs possess both Co and Cu diffraction peaks, further verifying the formation of Co–Cu bimetallic nanoparticles. The Raman spectra (Figure 3C) of the five samples contain characteristic peaks around 1,350 cm⁻¹ and 1,578 cm⁻¹, related to the d band of defective carbon and the G band of graphite carbon (Kumar et al., 2013), respectively. Co₃Cu/CFs have the largest intensity ratio of D and G bands (I_D/I_G, 1.01), indicating more defects and more potential active sites. The survey XPS spectra of Co₃Cu/CFs (Figure 3D, Supplementary Figure S4) confirm the presence of Co, Cu, C, N, and O elements. The high-resolution Co 2p XPS spectrum of Co₃Cu/CFs (Figure 3E) exhibits four peaks, including Co 2p _3/2 (780.3 eV), Co 2p _1/2 (795.6 eV), and two Co satellite peaks. Two strong peaks in the Cu 2p XPS spectrum (Figure 2F) that appear at 932.3 and 952.0 eV can be indexed to Cu 2p _3/2 and Cu 2p _1/2 , respectively. Notably, oxide peaks can be detected in Co and Cu 2p XPS spectra of Co₃Cu/CFs because of the easy oxidation of metallic Co and Cu nanoparticles (Xie et al., 2021; Kim et al., 2017). Similar C, N, Co, or Cu species could also be observed in the XPS spectra of another four samples (Supplementary Figure S5–S8).

The CO₂ electroreduction performances of Co–Cu bimetallic catalysts were investigated and compared with the performance of pure Co or Cu catalysts in a typical H-type electrochemical cell. All the catalysts were pre-activated using cyclic voltammetry until stable profiles were obtained. Figure 4A and Supplementary Figure S9 present the linear sweep voltammetry (LSV) curves of five samples in N₂-saturated or CO₂-saturated 0.5 M KHCO₃. The cathodic current densities of all five catalysts were measured to be approximately doubled in CO₂ than those in the N₂-saturated electrolyte, indicating the potential catalytic activity in CO₂ reduction (Zhang et al., 2014). In addition, LSV tests also prove that Co₃Cu/CFs show the highest current density among the five catalysts (Figure 4A).

In order to further evaluate the catalytic activities and quantify the product distribution from CO₂ electroreduction, potential dependent CO₂ electrolysis using five catalysts was conducted in a 0.5 M KHCO₃ electrolyte from –0.5 V_RHE to –1.2 V_RHE applied cathode potential. Gas products were directly injected into gas chromatography for on-line analysis, and liquid products in the catholyte were detected using ¹H NMR after each electrolysis. The product distribution and faradaic efficiencies (FEs) are summarized in Figure 4 and Supplementary Figure S10. For comparison of the electrocatalytic activities, Co/CFs and Cu/CFs with pure Co and Cu nanoparticles were also synthesized and assessed in CO₂ electrolysis. As shown in Figure 4B, Co/CFs generated CO as the only product from the CO₂ reduction and H₂ from the hydrogen evolution reaction (HER), with a 57% maximum FE for CO at –0.8 V_RHE applied potential. As for Cu/CFs (Figure 4D), it produced H₂, CO, and CH₄ during electrolysis, with a 49% maximum FE for CO at –0.9 V_RHE. CH₄ is one of the urgent products with the transfer of eight electrons (Sharifian et al., 2021); however, highest FEs of CH₄ only reach 7% at –0.8 V_RHE. Both Co/CFs and Cu/CFs show certain activities in CO₂ electroreduction, but neither of them could effectively suppress the HER process, and the FEs of H₂ range from 43 to 74%. As displayed in Figure 4C and Supplementary Figure S10, doping Co with Cu causes a significant increment in CO₂ catalysis. Compared to Co/CFs and Cu/CFs, the CO₂ reduction procedure on Co₃Cu/CFs, CoCu/CFs, and CoCu₃/CFs became much more dominant than the HER procedure. In particular, the maximum FE for C1 production (HCOOH and CO) increases to 97% at –0.8 V_RHE using the Co₃Cu/CFs catalyst, and the HER is totally suppressed to only a 3% FE of H₂. CoCu/CFs and CoCu₃/CFs have a similar tendency of FEs for CO₂ electrolysis as that of Co₃Cu/CFs, but they acquire lower total FEs of CO and HCOOH throughout the applied potential.

Chronoamperometry (CA) was used to evaluate the total current density during CO₂ electrolysis, and five samples achieved very close total current densities from –0.5 V_RHE to –1.2 V_RHE (Supplementary Figure S11). In addition, the partial current densities (j _C1 ) for C1 products (CO, HCOOH, and CH₄) were normalized by the total current densities and FEs at each cathode potential. Co₃Cu/CFs brought forth a significantly higher partial current density than Co/CFs, Cu/CFs, CoCu/CFs, and CoCu₃/CFs within the potential range, and got a maximum j _C1 of 78.1 mA cm⁻² at –1.2 V_RHE (Figure 4E). Therefore, Co₃Cu/CFs possess remarkable catalytic activity in CO₂ reduction, and they can also successfully suppress the HER procedure at relatively high cathode potentials. Moreover, it is extremely important to estimate the long-term durability of bimetallic catalysts because increasing the applied potentials and heavy current densities might seriously impact the structural stability (Vasileff et al., 2018; Jia et al., 2022). Long-term tests of potentiostatic CO₂ electrolysis were conducted using Co₃Cu/CFs catalysts at –0.8 V_RHE cathode potential where the best FEs of C1 products were obtained. The gaseous products were detected on-line every 6 h, and the corresponding CO FEs and current densities versus time are plotted in Figure 4F. Both CO FEs and partial current densities of the Co₃Cu/CF catalyst exhibited only small declines during the 60 h electrolysis, retaining approximately 90% of the original values and manifesting excellent stability in CO₂ electroreduction. Co₃Cu/CFs were characterized after a long-term electrolysis by TEM (Supplementary Figure S12), and the small Co₃Cu/CFs did not agglomerate together. As described previously (Figures 2D,E), the Co–Cu bimetallic nanoparticles are uniformly dispersed and firmly confined within the hierarchical pores of carbon nanofibers, and separated from easy agglomeration during electrolysis.

The aforementioned experimental results demonstrate the outstanding activities of Co₃Cu/CFs in CO₂ reduction. The mechanism of the high performance was first investigated using Tafel slopes within sufficiently low overpotential ranges. As presented in Figure 5A, 105 mV dec⁻¹ Tafel value is observed on Co₃Cu/CFs, confirming the first electron transfer from CO₂ to CO₂•⁻ as the rate determining step (Chen Jia et al., 2021). Compared to those of Co/CFs (108 mV dec⁻¹), Cu/CFs (127 mV dec⁻¹), CoCu/CFs (111 mV dec⁻¹), and CoCu₃/CFs (137 mV dec⁻¹), the lower Tafel value of Co₃Cu/CFs indicates faster reaction kinetics in CO₂ reduction. As mentioned previously, four electron product CH₄ was only obtained with the Cu/CF catalyst, and the bimetallic Co_xCu_y/CFs could produce CO and HCOOH. Metallic Cu owns relatively strong binding energies of *COOH and *CO intermediates compared to pure metallic Co, and these intermediates could be stabilized and further reduced to hydrocarbons or alcohols (Duan et al., 2018; Mun et al., 2019). In Co_xCu_y/CF samples, the catalytic behavior of metallic Cu was totally altered via fusing it with Co composition within the same nanoparticles. The binding energies of *COOH and *CO intermediates were weakened enough to be released from the catalyst surface, increasing the tendency toward HCOOH and CO production.

The electrochemical active surface area and the conductivity property of these five samples were also measured via double-layer capacitance (C_dl) and electrochemical impedance spectroscopy (EIS). Compared with Co/CFs (31.5 mF cm⁻²), Cu/CFs (17.5 mF cm⁻²), CoCu/CFs (30.0 mF cm⁻²), and CoCu₃/CFs (21.5 mF cm⁻²), Co₃Cu/CFs have a much higher C_dl value of 45.5 mF cm⁻² (Supplementary Figure S13), manifesting the larger ECSA and more active sites for CO₂ reduction (Yang et al., 2020b; Hao et al., 2022). In addition, the EIS curves in Figure 5B prove that Co₃Cu/CFs show a relatively small impedance than those of the other samples, which is beneficial to faster electron transport as well as better conductivity (Wei et al., 2022; Liu et al., 2014). The larger ECSA and good conductivity of Co₃Cu/CFs is consistent with its higher current densities in LSV and electrolysis tests. As illustrated in Figure 5C, Co₃Cu/CFs possess highly graphitized and multi-level porous carbon nanofibers, which can accelerate the electron transmission and expose abundant bimetallic Co–Cu sites for CO₂ reduction, eventually leading to the remarkable partial current densities for C1 products.