Materials synthesis and characterization
The CuPd-IC catalyst was first synthesized by chemically reducing copper and palladium salts in a sodium borohydride solution, followed by annealing in argon gas to accelerate atomic displacement and form the atomically ordered structure (Methods in the Supplementary Information)10. For comparison, Cu3Pd, disordered CuPd (d-CuPd), CuPd3 alloys, and pure Cu and Pd nanoparticles, were also prepared in a similar approach using solutions containing metal species with corresponding ratios and concentrations. The X-ray diffraction (XRD) pattern of CuPd-IC exhibited a body-centred cubic phase with a series of distinctive diffraction peaks at 2θ of 29.88°, 42.77°, 53.05°, 62.09°, 70.42°, and 78.34° (Fig. 2a), corresponding to the (100), (110), (111), (200), (210) and (211) crystal planes of the ordered CuPd-IC structure (ICSD 181913). In contrast, the XRD spectra of Cu3Pd, CuPd3, and disordered CuPd (d-CuPd) alloys presented broadened peaks between the (111) diffractions of pure Cu and Pd (Fig. 2b), confirming the formation of Cu-Pd alloys11. The composition of each Cu-Pd alloy was conducted by scanning electron microscope energy-dispersive X-ray spectroscopy. The Cu/Pd molar ratios were determined as 1:1, 1:1, 3:1, and 1:3, for CuPd-IC, d-CuPd, Cu3Pd, and CuPd3, respectively (Supplementary Table 1).
Transmission electron microscopy (TEM) images confirmed that the CuPd-IC sample had an average particle size of ~ 50 nm (Supplementary Fig. 1). In contrast, Cu, Pd, d-CuPd, Cu3Pd, and CuPd3 nanoparticles had similar particle sizes of ~ 10 nm. TEM mapping analyses identified the co-existence and uniform distributions of Cu and Pd elements in those particles (Supplementary Fig. 2–3). The atomic structure of the CuPd-IC sample was further investigated by aberration-corrected high-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM). The entire lattice presented highly ordered rectangular arrays with alternating bright and dark columns of atoms (Fig. 2c). Due to the difference of atomic numbers of Cu (Z = 29) and Pd (Z = 46), the Pd atom columns appeared much brighter than those of Cu atoms. The two perpendicular orientations of unit cells were highlighted and labelled (Fig. 2d). The (100) direction contained both Cu and Pd atoms in an alternating order, while the (110) direction only had the same kind of atoms within each row. The intensity profile measured from the STEM images presented a periodic oscillation pattern (Fig. 2e), and the average peak-to-peak distances were 0.211 and 0.298 nm, corresponding to the Pd − Pd distances at (110) and (100) directions, respectively.
The chemical states of Cu and Pd species of the CuPd-IC were studied using X-ray photoelectron spectroscopy (XPS). Compared to pure Cu nanoparticles, the Cu 2p3/2 and 2p1/2 peaks of CuPd-IC upshifted 0.31 and 0.15 eV toward higher binding energy, respectively (Supplementary Fig. 4a), owing to the larger electron negativity of Pd than that of copper12. Correspondingly, the Pd 3d5/2 and 3d3/2 peaks of CuPd-IC downshifted 0.28 and 0.25 eV, compared to pure Pd nanoparticles (Supplementary Fig. 4b). Besides, the surface valance band spectra of those Cu-Pd alloys showed that with the increase of the Pd fraction, the d-band center position of these samples exhibited an upshift toward the Fermi-level (Supplementary Fig. 5), which was beneficial for enhancing the adsorption of intermediate species13, 14.
To precisely resolve the local structural information for Pd and Cu atoms, the samples were further characterized by X-ray absorption fine structure (XAFS) spectroscopy. For different Cu-Pd alloys, the Cu and Pd K-edge X-ray absorption near-edge structure (XANES) spectra were displayed, compared to those of the standard Cu and Pd foils. The Cu K-edge spectrum of the Cu-Pd alloys presented an absorption edge at 8979.0 eV (Fig. 3a, b), indicating a dominent metallic Cu feature. The Pd K-edge spectra of the Cu-Pd alloys were also examined and presented an absorption edge at 24,370.0 eV, indicating the feature of metallic Pd (Supplementary Fig. 6). Structural information about local atomic coordination was derived by fitting the extended X-ray adsorption fine structure (EXAFS) spectra, and the fitted structural parameters were summarized in Supplementary Table 2. The main peaks in both the Cu K-edge and Pd K-edge spectra of CuPd-IC fit well with the characteristics of the Cu − Pd bond, suggesting that the Cu − Pd bonds dominate the coordination environments of Cu and Pd atoms in CuPd-IC (Supplementary Fig. 7).
In addition, the Cu K-edge EXAFS spectra and fitting results were compared with those of different Cu-Pd alloys and standard Cu foil (Fig. 3c). The bond lengths of the Cu–Pd and Cu–Cu bonds were determined as ~ 2.58 and ~ 2.51 Å, respectively. For all the Cu-Pd alloys, the Cu coordination numbers (CNs) were calculated as ~ 8, while the CN ratios between Cu–Pd and Cu–Cu for were 7.2:0.5 (CuPd-IC), 4.8:3.1 (d-CuPd), 2.2:5.7 (Cu3Pd), and 6.6:1.9 (CuPd3), respectively (Fig. 3c). The highest ratio for CuPd-IC suggests its highest density of Cu-Pd pairs. Similar conclusion was also drawn from the Pd K-edge. The CuPd-IC showed the largest CN ratio between Pd–Cu and Pd–Pd bonds as 6.6:1.6 (Supplementary Fig. 8 and Supplementary Table 2), confirming the existence of high-density Cu-Pd pairs within the CuPd-IC.
Further investigation using wavelet-transform EXAFS (WT-EXAFS) of Cu K-edge and Pd K-edge also confirmed the ordered Cu-Pd atomic arrangement in the CuPd-IC sample (Fig. 3d). WT-EXAFS analysis is useful for separating backscattering atoms, providing a radial distance and the k-space resolution15. For the Cu–Cu path, the WT maximum was at 7.53 Å−1 for the copper foil, which rose to higher k-values of 8.55 for d-CuPd, and 9.80 Å−1 for CuPd-IC, respectively. This increased k-value and intensity of CuPd-IC suggested its higher Cu-Pd mixing degree compared to d-CuPd. Similarly, compared to d-CuPd (8.39 Å−1) and pure Pd foil (9.86 Å−1), the CuPd-IC presented a WT maximum at a k-value of 8.62 Å−1 (Supplementary Fig. 9), indicating a higher structural order around Pd sites of CuPd-IC than d-CuPd15, 16.
CO electroreduction performance
The electrochemical CORR measurements of different Cu-Pd alloys were then conducted to investigate their catalytic performances. A flow-cell electrolyzer was used to break the low solubility limitation of CO in aqueous electrolytes for a high CO conversion rate (Methods in the Supplementary Information)17, 18. For CuPd-IC (Fig. 4a, Supplementary Table 3), the total current density increased rapidly with increasing applied potential and reached ~ 650 mA·cm− 2 at − 1.03 V vs. RHE, with a partial current density of 425 mA·cm− 2 and a corresponding FE of ~ 70 ± 5% for producing acetate (Supplementary Table 3), substantially superior to the state-of-the-art electrocatalysts for CO-to-acetate reduction5–7. A representative H-NMR spectrum of the electrolyte after electrochemical CORR was displayed (Supplementary Fig. 10). In comparison, for pure Cu nanoparticles (Fig. 4b, Supplementary Table 4), although the total current density was over 800 mA·cm− 2 at − 0.88 V vs. RHE, the major reduction products were C2H4 and ethanol, with corresponding FE values of 43 ± 2% and 27 ± 3%, respectively. The FE of acetate production was below 10% in the entire range of applied voltage.
To further explore the catalytic function of the Cu-Pd pairs, the CO electroreduction performances of CuPd-IC and other Cu-Pd alloys with different compositions and structures were measured and compared (Supplementary Fig. 11–13 and Supplementary Table 5–7). The acetate selectivity showed a volcano type with the Pd content (Fig. 4c), in which the CuPd-IC presented the highest acetate selectivity compared to other Cu-Pd alloys. The excellent CORR performance of the CuPd-IC catalyst with high-selectivity-at-high-current-density was further translated to a high acetate production rate of ~ 1.1 µmol·s− 1·cm− 2 (Fig. 4d). In contrast, pure Pd was inactive for CO electroreduction due to its too strong CO adsorption12, 19, and thus only H2 evolution was observed (Supplementary Fig. 14).
To assess the potential toward commercially-relevant acetate manufacture, we conducted a long-term electrochemical test using a 5 cm2 MEA electrolyzer equipped with CuPd-IC catalyst (Methods in the Supplementary Information). The MEA allowed to reduce the electrolyte ohmic losses and avoid electrolyte flooding through gas diffusion layer20. The full cell voltage was well maintained between − 3.6 ± 0.1 V during a 100-hour continuous operation at a constant current density of 500 mA·cm− 2 (Fig. 4e). The total acetate production was ~ 2 mol, corresponding to an average acetate production FE of ~ 43%.
Density functional theory calculations
Density function theory (DFT) calculations were conducted to obtain an atomic insight into the enhanced acetate activity and selectivity on the CuPd-IC catalyst. As CuPd(110) is the most stable surface of the ordered body-centered-cubic (BCC) CuPd alloy, while Cu(111) and Pd(111) are the most stable surfaces for the face-centered-cubic (FCC) Cu and Pd, respectively, these surfaces were employed as the model systems. In order to examine the coverage effect, the adsorption free energies of various reaction intermediates were also calculated at a low coverage (0.06 monolayer, ML) and a high coverage (0.50 ML), respectively.
We started by checking the adsorption free energies of CO on CuPd(110), Cu(111) and Pd(111). The adsorption of CO on CuPd(110) is stronger than that on pure Cu (–0.06 eV), but weaker than that on pure Pd (–1.01 eV). For CuPd(110), it was found that CO prefers to locate on a hollow site consisting of two Pd atoms and one Cu atom, rather than that of a Pd atom and two Cu atom (i.e., CuPd2 hollow site vs. Cu2Pd hollow site, Fig. 5a). As shown in Supplementary Table 8, the adsorption free energies of CO on CuPd(110) are more negative as compared to that on the Cu(111) surface both at the low coverage (–0.78 eV vs. − 0.06 eV) and at the high coverage (–0.49 eV vs. 0.02 eV), showing that CuPd-IC can bind CO more strongly to enrich CO on its surfaces as compared to pure Cu.
Then, the formation of acetate on CuPd(110) at both low and high surface coverages of *CO was examined. The reaction energetics were calculated based on the standard hydrogen electrode (SHE) model21, 22, in which one-half of the chemical potential of hydrogen was used as the chemical potential of the proton-electron pair. The initial step was the formation of *CO–COH, which is a common intermediate proposed in the literature for the formation of the C2+ products23–26. Afterwards, the acetate formation pathway, proposed by Luc et. al. for Cu(111)7, was employed. The free energy changes for the acetic acid formation on CuPd(110) at 0 V (vs. SHE) at the low coverage (blue curve in Fig. 5b) and the high coverage (red curve in Fig. 5b) are both presented (data in Supplementary Table 9), and the adsorption configurations of the intermediates are summarized (Supplementary Fig. 15–16). The results show that the ΔG for the C–C coupling is uphilled by 1.53 eV at the low coverage, which is significantly decreased to 0.87 eV at the high coverage. For all the following steps, the ΔG values are more negative at the high coverage than those at the low coverage, and thus they shall overcome lower barriers according to the Brønsted–Evans–Polanyi relations27, 28. In particular, for ethenone (*CH2–C = O) that is considered as the key intermediate on the acetate formation pathway, the ΔG for the ethenone formation at the high coverage is − 0.61 eV, significantly lower than that at the low coverage (–0.22 eV). Thus, a high coverage of *CO can further accelerate the acetate formation.
In addition to a stronger CO adsorption, the role of Pd in CuPd-IC in enhancing the acetate formation was also studied by comparing the acetate formation pathways on both CuPd(110) and Cu(111) surfaces (Fig. 5c), where only the results of free energy changes at the low coverage were presented. Starting from the CO–COH*, it was found that the ΔG of the ethenone formation is 0.04 eV on Cu(111) surface, but is − 0.22 eV on CuPd(110) (Fig. 5c, Supplementary Table 10), suggesting that the introduction of Pd can stabilize ethenone, the key intermediate on the acetate formation pathway proposed in the literature7, 29.
In addition, for alloys with Pd/Cu > 1, the aggregation of Pd atoms leads to more severe hydrogen evolution reaction. As shown in Supplementary Table 11, under 298.15 K and 105 kPa H2 pressure, the hydrogen binding free energy was calculated to be 0.16 eV on Pd(111) with *CO coverage as high as 0.69 ML, much lower than that of 0.32 eV for *H on CuPd(110) with *CO coverage reaching 0.50 ML. Thus, we hypothesized that the highly ordered CuPd-IC presents a high density of the Cu–Pd pairs to minimize the Pd–Pd clustering, which allows to enrich the CO adsorption, stabilize ethenone and depress hydrogen evolution, thus leading to an efficient production of acetate.