CO2 electroreduction (CO2RR) into value-added chemical feedstocks and fuels, driven by local-generated renewable energy, is a highly promising strategy for realizing the carbon-neutral cycle together with earning potential economic returns1,2,3,4. Among all CO2RR products, CH4 is of considerable interest based on its well-established infrastructure toward storage, distribution, and utilization5,6. Up to date, mainly Cu-based catalysts are able to generate appreciable CH4 via stabilizing and subsequently hydrogenating the *CO species during CO2RR7. However, owing to the involvement of complicated 8-electron transfer steps and structural degradations (e.g., fragmentation, dissolution, agglomeration), most Cu-based catalysts (e.g., oxide-derived Cu) still suffer from unsatisfactory Faradaic efficiency for CH4 and poor stability8,9.

Perovskite oxides (typically ABO3), featuring distinct merits (e.g., diverse chemical compositions, flexible crystal and electronic structures, and governable physicochemical properties), have provided an attractive platform for accessing high-performance catalysts toward numerous electrochemical reactions10,11,12,13. Upon most occasions, the nature of B-site cations or B–O bonding determines the electrocatalytic properties of perovskite oxides in essence14,15,16. Based on the above characteristics of Cu-based catalysts and perovskite oxides, if the B sites could be occupied entirely or partly by Cu element, the corresponding Cu-based perovskite oxides would be active toward CO2 electromethanation17,18,19,20,21,22. Typical examples involve Cu-based Ruddlesden–Popper perovskite oxides (e.g., La2CuO4−δ)17,18,19,20,21,22. Nonetheless, these catalysts with B sites wholly occupied by Cu exhibit low activity and selectivity for CH4, owing to the distance of their adjacent Cu sites not far enough to inhibit the competitive C–C coupling17,18,19,20,21,22. Furthermore, like traditional Cu-based oxides (e.g., CuO and Cu2O)23,24,25, since the electrode-supplied electrons attack or break the Cu–O bond to reduce the Cu sites, these catalysts also undergo uncontrollable reconstructions (e.g., metallic Cu exsolution) during CO2RR20,21. Such reconstructions could make the active sites unmaintainable, causing lowered catalytic performance or even deactivation8,26,27.

Substitution of another cation (B’) for B to form doped perovskite oxides (e.g., AB1−xB’xO3) has been intensively proved as a tried-and-true strategy to optimize the catalytic performance of perovskite oxides10,11,14,15,16. Accordingly, for Cu-based perovskite oxides, partly occupying their initial Cu sites by the doping cations (B’) could also availably modulate or enhance their catalytic properties toward CO2-to-CH4 conversion. In general, if the B’ and Cu cations are almost equal in molar content, while they are sufficiently different in size and/or charge, Cu-based double perovskite oxides (A2CuB’O6) with B-site rock-salt ordering will be produced28,29,30. The formation of a double perovskite structure is very likely to introduce important benefits to the physicochemical properties, affecting activity, selectivity, and stability in CH4 production28,29,30,31,32,33,34. Specifically, in the rock-salt-type arrangement, the B-site cations alternate in all three crystallographic dimensions, markedly widening the distance between adjacent Cu cations, theoretically almost doubling relative to the undoped ones28,29,30. This increased distance could suppress *CO dimerization and promote activity and/or selectivity for CH4 production31,32. Moreover, the B-site rock-salt ordering could bring superexchange interaction between Cu and B’ cations (mediated by intermediate O anions) and give rise to the redistribution of charge densities of the B-site cations via electron transfer33,34. During CO2RR, this superexchange interaction may transfer the electrode-supplied electrons accumulated around the Cu sites to B’ sites and stabilize the Cu sites, thereby boosting the catalytic stability. However, to our knowledge, such Cu-based double perovskite oxides have not been reported in CO2RR, so the vital roles of their unique physicochemical properties in catalytic performance are yet to be fully uncovered.

Here we present Cu-based double perovskite oxides (A2CuB’O6) with B-site rock-salt ordering and superexchange interaction to facilitate efficient and stable CO2-to-CH4 conversion. As the proof of concept, we employed W6+ cations as the B’ sites, mainly because of their low-lying unoccupied 5d states that strongly hybridized with O 2p states, and synthesized a double perovskite oxide of Sr2CuWO6 as the model catalyst for CO2RR. As expected, for the Sr2CuWO6, its corner-linked octahedra of CuO6 and WO6 were rock-salt ordered. This unique structure made the nearest Cu cations very far apart from each other with a minimum distance of 5.4 Å and introduced superexchange interaction that was mainly manifested by O-anion-mediated electron transfer from Cu to W cations. When evaluated as a catalyst toward CO2RR, relative to its physical-mixture counterpart and the reported Cu-based perovskite oxides, the Sr2CuWO6 delivered remarkable enhancements in activity and selectivity for CH4, together with boosted stability. Our experiments and theoretical calculations suggested that such performance improvements were mainly attributed to the following aspects: the sufficiently long Cu–Cu distances promoting *CO hydrogenation but inhibiting C–C coupling; the superexchange interaction transferring the electrons (around Cu sites) to W sites during CO2RR and thus stabilizing the Cu sites (e.g., Cu+).


Crystal structure and long Cu–Cu distances

The Sr2CuWO6 catalyst was synthesized through a facile and scalable solid-state reaction (combined high-energy ball milling) process. Note that, according to the tolerance factor rule10, another alkaline-earth metal cation, i.e., Ba2+, can also be selected as the A-site cation to form a double perovskite of Ba2CuWO6. Since our work mainly focused on uncovering the key roles of superexchange-stabilized long-distance Cu sites in enhancing CO2RR property, either Sr2CuWO6 or Ba2CuWO6 can serve as the model catalyst in our work. For the proof of concept, here we designed and synthesized one of these two, i.e., Sr2CuWO6. The as-prepared sample had uniform particle size with an average value of around 300 nm together with a specific surface area of about 3 m2 g−1 (Supplementary Fig. 1). According to the inductively coupled plasma mass spectroscopy analysis, the chemical constituent of the Sr2CuWO6 sample was compatible with its nominal compositions (Supplementary Table 1). Figure 1a shows the X-ray diffraction (XRD) pattern and corresponding Rietveld refinement analysis (Supplementary Table 2) of the Sr2CuWO6 sample. The Sr2CuWO6 was characterized by a pure tetragonal B-site rock-salt-ordered double perovskite phase that was indexed to a space group of I4/m with lattice parameters of a = 5.436 Å and c = 8.400 Å35. Here we also showed the crystal structure of Sr2CuWO6 in Fig. 1a. The structure consisted of alternating corner-sharing WO6 and Jahn–Teller distorted CuO6 octahedra (with short Cu–Oab bonds in the ab-plane and long Cu–Oc bonds along the c-axis), with Sr cations situated at the void positions between these octahedra. As a result, the probably nearest Cu cations were far apart from each other, with two different distances of 5.4 and 5.7 Å that were induced by the Jahn–Teller distortion of CuO6 octahedra35 (Fig. 1b and Supplementary Fig. 2). These distances between adjacent Cu species were far enough to inhibit the C–C coupling and facilitate CO2-to-CH4 conversion, as to be discussed below.

Fig. 1: Crystal structure and composition of Sr2CuWO6.
figure 1

a Rietveld refinement plot of XRD data and schematic illustrations of crystal structure for Sr2CuWO6. Sr, Cu, W, and O are represented by green, blue, gray, and red dots, respectively. The blue and gray octahedra represent CuO6 and WO6 motifs, respectively. b Schematic illustrations of distances between the probably nearest Cu cations. c HRTEM image of Sr2CuWO6 (scale bar: 10 nm). d Enlarged HRTEM image of Sr2CuWO6 taken from the region marked in (c) (scale bar: 2 nm). e SAED pattern of Sr2CuWO6 (scale bar: 5 1/nm). f EDX mappings of Sr2CuWO6 (scale bar: 100 nm).

We validated the crystal structure of Sr2CuWO6 using high-resolution transmission electron microscopy (HRTEM) and a selected-area electron-diffraction (SAED) pattern along the [1\(\bar{1}\)0] zone axis. In Fig. 1c–e, the tetragonal phase was observed, presenting clear crystal fringes with interplanar spacings of about 0.284 and 0.420 nm, corresponding to its (112) and (002) diffraction planes, respectively. Raman spectra further suggested the phase structure of Sr2CuWO6 crystallized with tetragonal I4/m symmetry (Supplementary Fig. 3)36. The energy dispersive X-ray (EDX) mappings in Fig. 1f suggested the existence and homogeneous distribution of Sr, Cu, W, and O elements in the sample. Wide-scan X-ray photoelectron spectra (XPS) (Supplementary Fig. 4) also indicated that the sample was composed of the Sr, Cu, W, and O elements, without any detected signal of other elements except the reference C element.

Superexchange interaction

We conducted XPS and synchrotron-based X-ray absorption spectra (XAS) to explore electronic structure information and superexchange interaction of Sr2CuWO6 catalyst. A physical mixture of CuO/WO3 was prepared as a control sample (Supplementary Fig. 5), carrying the same molar ratio of Cu and W elements as the Sr2CuWO6. Figure 2a, b shows Cu 2p and W 4f spectra of the Sr2CuWO6. The peaks at 934.1 and 35.1 eV could be assigned to Cu2+ 2p3/2 and W6+ 4f7/2, respectively, illustrating the approximate valence states of Cu (+2) and W (+6) in Sr2CuWO6. Relative to the CuO/WO3, the Cu2+ 2p3/2 peak of Sr2CuWO6 shifted 0.33 eV to higher binding energy, whereas their W6+ 4f7/2 peak underwent a negative shift of 0.26 eV. Such XPS peak shifts preliminarily suggest that there is electron redistribution (from Cu2+ to W6+) in the Sr2CuWO6.

Fig. 2: Superexchange interaction in Sr2CuWO6.
figure 2

a Cu 2p XPS spectra of Sr2CuWO6 and CuO/WO3. b W 4f XPS spectra of Sr2CuWO6 and CuO/WO3. c Cu K-edge XANES spectra of Sr2CuWO6. d W L3-edge XANES spectra of Sr2CuWO6. e Enlargement of Cu K-edge XANES spectra. f Enlargement of W L3-edge XANES spectra. g Top view of charge distribution between CuO6 and WO6 octahedra in Sr2CuWO6. Cu, W, and O are represented by blue, gray, and red dots, respectively. The light-blue regions (surrounding the Cu, O, and W sites) depict the electron transfer channels. h Schematic illustration of Cu–O–W superexchange interaction (electron transfer from Cu to W cations mediated by O anions) in Sr2CuWO6. The blue and gray octahedra represent CuO6 and WO6, respectively. i Schematic illustration of electronic DOS contributions (from O 2p and Cu 3d states) and Cu–O bond covalency for Sr2CuWO6 and CuO.

Figure 2c, d shows the normalized Cu K-edge and W L3-edge X-ray absorption near-edge structure spectra (XANES) for Sr2CuWO6. The absorption edges (i.e., Cu K-edge and W L3-edge) of Sr2CuWO6 were nearly identical to those of CuO and WO3 references, respectively, confirming the valence states of Cu and W species in Sr2CuWO6 close to +2 and +6. In the enlarged spectrum of the Cu K-edge (Fig. 2e), a positive-energy shift and higher white-line peak intensity were observed for the Sr2CuWO6, as compared to the CuO reference, indicative of the existence of a higher valence state of Cu species in Sr2CuWO6. On the contrary, the spectrum of the W L3-edge for Sr2CuWO6 exhibited a slight shift towards lower energy and a weaker white-line peak intensity relative to the WO3 reference (Fig. 2f), indicating a minor reduction of W valence state in Sr2CuWO6. These XAS results demonstrate electron interaction between CuO6 and WO6 octahedra or electron transfer in the direction from Cu to W species in the Sr2CuWO6. Combined with the above crystal structure characterization, one can believe that this electron transfer between rock-salt-ordered CuO6 and WO6 octahedra must be mediated by the intermediate oxygen anions, thus being defined as a superexchange interaction.

We further performed Bader charge analysis to investigate charge density redistribution. In Fig. 2g, the light-blue regions, surrounding the Cu, O, and W sites, clearly depicted the Cu–O–W charge transfer channels. The Bader charges of Cu and W sites in Sr2CuWO6 were calculated to be 1.27 and 2.95 |e|, respectively, which were different from 1.08 |e| for Cu sites in CuO and 3.08 |e| for W sites in WO3 (Supplementary Table 3). These phenomena also indicate the charge redistribution from Cu to W sites (mediated by O sites) in Sr2CuWO6, consistent with the XPS and XAS results. Thus, the B-site rock-salt-ordered double perovskite lattice was endowed with significant superexchange interaction (Cu–O–W) between alternate CuO6 and WO6 octahedra, mainly characterized by the O-anion-mediated electron transfer from Cu to W cations (Cu2+ + W6+ → Cu>2+ + W<6+), as schematically illustrated in Fig. 2h. As a result, we infer that the superexchange interaction could suppress the accumulation of electrode-supplied electrons around Cu sites via fast electron transport channels (light-blue regions in Fig. 2g), thereby protecting the Cu sites during CO2RR. Besides, in light of the increased valence state (or electronegativity) of Cu sites reducing the electronegativity difference between Cu and O sites, the superexchange interaction could strengthen Cu–O bond covalency and thus maintain the Cu–O lattice integrity during CO2RR. We proved the strengthened Cu–O bond covalency by the computed density of states (DOS) and band centers of Cu 3d and O 2p (Fig. 2i and Supplementary Fig. 6), using CuO as a reference.

Activity and selectivity for CH4

We carried out density functional theory (DFT) calculations to predict CO2RR properties of Sr2CuWO6 catalyst (Fig. 3a, Supplementary Figs. 7 and 8). The DFT calculations were implemented on CuO2/WO2-terminated Sr2CuWO6(001) surface (Supplementary Fig. 7) since such a surface was usually observed and stable37,38. We took the full reaction pathways for CH4 and C2H4 formation starting from *CO as analysis objects and calculated their corresponding energy profiles at the Cu sites39,40. On the Sr2CuWO6(001) surface, the energy difference between *CO and *CHO was about 0.64 eV, much lower than the energy barrier (1.08 eV) for C2H4 production (i.e., 2*CO to the TS) (Fig. 3a and Supplementary Fig. 8). As a result, CH4 formation was more favorable on the Sr2CuWO6(001) surface based on the presumption that the energy of TS for the *CO hydrogenation was not significantly different from the energy of *CO step. This could be ascribed to the fact that the long Cu–Cu distances (at least 5.4 Å) on Sr2CuWO6(001) surface were able to intensify the single-atomic feature of Cu, thereby inhibiting the C–C coupling but facilitating the CH4 production. To this end, associated with its actual physicochemical properties, we can predict that the Sr2CuWO6 catalyst with B-site rock-salt-ordered structure will offer remarkable activity and selectivity toward CO2-to-CH4 conversion.

Fig. 3: Activity and selectivity for CH4 over Sr2CuWO6.
figure 3

a DFT-calculated energy diagrams for CH4 and C2H4 formation on Sr2CuWO6(001) surface starting with *CO (TS: transition state). b FEs for various gas products over Sr2CuWO6 at different applied current densities. c \({{{\mbox{FE}}}}_{{{\mbox{C}}}{{{\mbox{H}}}}_{4}}\) of Sr2CuWO6 and CuO/WO3 at different applied current densities. d \({{{\mbox{FE}}}}_{{{\mbox{C}}}{{{\mbox{H}}}}_{4}}/{{{\mbox{FE}}}}_{{{{\mbox{C}}}}_{2}{{{\mbox{H}}}}_{4}}\) or \({j}_{{{\mbox{C}}}{{{\mbox{H}}}}_{4}}/{j}_{{{{\mbox{C}}}}_{2}{{{\mbox{H}}}}_{4}}\) of Sr2CuWO6 and CuO/WO3 at different applied current densities. e \({{{\mbox{FE}}}}_{{{\mbox{C}}}{{{\mbox{H}}}}_{4}}\) and \({j}_{{{\mbox{C}}}{{{\mbox{H}}}}_{4}}\) of Sr2CuWO6, in comparison with those of Cu-based perovskite oxides reported in the literature (Supplementary Table 4). The error bars represent the mean ± standard deviation (SD, n = 3 replicates).

We preliminarily checked the probability of CO2RR occurring over the Sr2CuWO6 catalyst by linear sweep voltammogram (LSV) curves recorded in a CO2- and Ar-flowed liquid-electrolyte (1 M KOH) flow cell, respectively (Supplementary Figs. 9 and 10). Relative to Ar-flowed electrolyte, there were higher current densities as well as a less negative onset potential in CO2-flowed electrolyte, suggesting that the Sr2CuWO6 catalyst is indeed active toward CO2RR. We then systematically evaluated CO2RR properties of the Sr2CuWO6 catalyst at various applied current densities in CO2-flowed liquid-electrolyte (1 M KOH) flow cell (Fig. 3b, Supplementary Figs. 11 and 12). As a note, the 1 M KOH was adopted as the electrolyte in the flow cell because it was able to improve charge transfer, inhibit HER, and thus give rise to marked improvements in CO2RR activity and selectivity, relative to the bicarbonate/carbonate electrolytes13. In the applied current density range (from 100 to 600 mA cm−2), the main product was CH4, with high Faradaic efficiencies (FEs) more than 42.3% (Fig. 3b). At a current density of 400 mA cm−2, the CH4 product displayed a maximum FE of 73.1%, corresponding to a high partial current density of 292.4 mA cm−2 exceeding the industrial-level requirements (>200 mA cm−2) (Fig. 3b and Supplementary Fig. 13). Meanwhile, the \({{{{{\rm{FE}}}}}}_{{{{{{\rm{C}}}}}}_{2}{{{{{\rm{H}}}}}}_{4}}\) and FEliquid C2+ ranged from 2.2% to 7.1% (Fig. 3b and Supplementary Fig. 12), indicating an efficient suppression of C–C coupling. These results reveal that upon serving as a catalyst toward CO2RR, the Sr2CuWO6 is prone to generate CH4 rather than C2H4, in line with the above DFT calculations (Fig. 3a).

We also benchmarked the CO2RR properties of the Sr2CuWO6 against the CuO/WO3. The detailed CO2RR properties of the CuO/WO3 were shown in Supplementary Fig. 14. The Sr2CuWO6 significantly promoted CO2-to-CH4 conversion, whereas its physical-mixture counterpart enhanced C–C coupling (similar to oxide-derived Cu catalysts23,24). To be specific, in the applied current density range, relative to the CuO/WO3, the Sr2CuWO6 exhibited 3.7- to 14.1-fold higher \({{{{{\rm{FE}}}}}}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) or \({j}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) (Fig. 3c and Supplementary Fig. 13), together with much lower \({{{{{\rm{FE}}}}}}_{{{{{{\rm{C}}}}}}_{2}{{{{{\rm{H}}}}}}_{4}}\) or \({j}_{{{{{{\rm{C}}}}}}_{2}{{{{{\rm{H}}}}}}_{4}}\) (Supplementary Fig. 15). And the values (10.8–26.2) of \({{{{{\rm{FE}}}}}}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}/{{{{{\rm{FE}}}}}}_{{{{{{\rm{C}}}}}}_{2}{{{{{\rm{H}}}}}}_{4}}\) or \({j}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}/{j}_{{{{{{\rm{C}}}}}}_{2}{{{{{\rm{H}}}}}}_{4}}\) for the Sr2CuWO6 were almost 13.2–59.8 times higher than those (0.44–1.98) for the CuO/WO3 (Fig. 3d). Combined with the above physicochemical property characterizations (Figs. 1 and 2) and DFT calculations (Fig. 3a), we could attribute these results to the sufficient-long Cu–Cu distances of Sr2CuWO6 that regulated the adsorption/activation of key intermediates, thus inhibiting C–C dimerization and promoting *CO hydrogenation. We compared the \({{{{{\rm{FE}}}}}}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) and \({j}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) of Sr2CuWO6 catalyst with those of the reported Cu-based perovskite oxides (Fig. 3e and Supplementary Table 4). The Sr2CuWO6 performed much better than all these perovskites reported in the literature. For instance, the \({j}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) of Sr2CuWO6 was about 2.5–1562.5 times higher than that of the reported perovskite-based catalysts. To our knowledge, the Sr2CuWO6 was the most effective Cu-based-perovskite catalyst for CO2-to-CH4 conversion. Moreover, Supplementary Fig. 16 highlights that the activity and selectivity for CH4 of Sr2CuWO6 are comparable to or higher than those of most reported representative Cu-based catalysts in flow cells (Supplementary Table 5).

Cu sites stabilized by superexchange interaction

We performed a series of ex-situ and in-situ characterizations to investigate the structural evolution of Sr2CuWO6 and stabilization of Cu sites during CO2RR (Fig. 4). The reduction tolerance of Sr2CuWO6 was probed under a high-temperature reducing atmosphere. At 300 °C (in H2/Ar for 1 h), the Cu2+ in Sr2CuWO6 was reduced to Cu+, instead of metallic Cu (Fig. 4a and Supplementary Fig. 17), with the generation of oxygen vacancies (Supplementary Fig. 18 and Supplementary Table 6). Whereas the CuO/WO3 was gradually reduced to Cu2O/WO3 (at 250 °C) and Cu/WO3 (at 300 °C) (Supplementary Fig. 19). According to the Rietveld refinement analysis (Supplementary Fig. 20 and Supplementary Table 7), the Sr2CuWO6 underwent a phase transition from I4/m to Fm-3m during thermochemical reduction but still belonged to the category of B-site rock-salt-ordered double perovskites (Fig. 4a, b)41. This phase transition could be ascribed to lattice expansion of CuO6 octahedra induced by reduction of smaller-size Cu2+ (0.87 Å) to larger-size Cu+ (0.91 Å). Notably, in the newly generated structure, the Cu–Cu distance (about 5.8 Å) was still very long (Fig. 4c and Supplementary Fig. 21), and the superexchange interaction (Cu–O–W) could still exist or even be strengthened due to the easier electron transfer from Cu+ to W6+ sites relative to that from Cu2+ to W6+ sites (Supplementary Fig. 22). These results indicate that the superexchange interaction can inhibit deep reduction of the Cu sites and thus avoid structural collapse of the Sr2CuWO6. These may partly imply the high structural stability of Sr2CuWO6 during CO2RR.

Fig. 4: Stability of the Cu sites.
figure 4

a XRD pattern of Sr2CuWO6 after thermochemical reduction. b Schematic illustration of phase transition of Sr2CuWO6 after reduction. Sr, Cu, W, and O are represented by green, blue, gray, and red dots, respectively. The red-dotted circle, blue, and gray octahedra represent oxygen vacancy, CuO6, and WO6, respectively. c Schematic illustration of the distances between the nearest Cu cations in Fm-3m phase of Sr2CuWO6. d XRD patterns of Sr2CuWO6 after CO2RR at different current densities (GDE: gas diffusion layer). e Cu LMM XPS spectra of Sr2CuWO6 after CO2RR at different current densities. f W XPS spectra of Sr2CuWO6 after different current densities. g CO2RR stability test of Sr2CuWO6 and CuO/WO3 in a flow cell at 400 mA cm−2. In-situ Raman spectra of h Sr2CuWO6 and i CuO/WO3 as a function of CO2RR time at −1.2 V vs. RHE (RHE: reversible hydrogen electrode).

The possible structural changes of Sr2CuWO6 after CO2RR were analyzed using ex-situ XRD and XPS (Fig. 4d–f). Similar to the thermochemical reduction, after CO2RR (e.g., at 200 mA cm-2), part of the I4/m phase of Sr2CuWO6 was converted into Fm-3m phase (Fig. 4d), without detectable impurity, and the Cu2+ and part W6+ species on the surface were reduced to Cu+ and W5+, respectively (Fig. 4e, f, Supplementary Fig. 23, and Supplementary Table 8). This suggests that the Cu+ species (in the Fm-3m phase) might be active sites for CO2 methanation. By contrast, as fully evidenced by previous studies, the CuO (in CuO/WO3) can completely be reduced to metallic Cu under similar CO2RR conditions26,27. As a result, the Cu sites of oxidation states in double perovskite structure are well stabilized by the superexchange interaction during CO2RR. On this basis, we evaluated CO2RR stability of the Sr2CuWO6 in comparison with the CuO/WO3 through chronopotentiometric polarization in the CO2-flowed liquid-electrolyte flow cell (Fig. 4g). During 20,000 s of electrolysis (at 400 mA cm−2), for the Sr2CuWO6, the applied potential was stable at 1.23 ± 0.15 V vs. RHE, and the \({{{{{\rm{FE}}}}}}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) was maintained at 64% ± 6%. Whereas the CuO/WO3 showed severe deterioration in potentials and obvious fluctuations in \({{{{{\rm{FE}}}}}}_{{{{{\rm{C}}}}}{{{{{\rm{H}}}}}}_{4}}\) during 8000 s of electrolysis. These results demonstrate excellent CO2RR stability of the Sr2CuWO6. As a note, our gas diffusion layer suffered flooding issues when the CO2RR stability test of Sr2CuWO6 catalyst ran for more than 20,000 s. Since the flooding issues can cause an essential failure of the CO2RR system, we terminated the electrolysis on the Sr2CuWO6 catalyst at about 20,000 s; but the steady CO2RR testing time for the Sr2CuWO6 catalyst itself was supposed to be for more than 20,000 s by considering the well-stabilized Cu sites. Usually, the flooding issues can be mitigated by washing away the carbonate precipitation. This has been widely used to reactivate the electrodes42,43.

Moreover, the stability of Cu sites in CuO6 octahedra of Sr2CuWO6 during CO2RR was further verified by in-situ Raman spectroscopy in an operando electrolyzer (Supplementary Fig. 24). As expected, the in-situ Raman spectroscopic analyses were consistent with the above ex-situ characterizations. To be specific, as the applied potential negatively shifted (from −0.8 to −1.2 V vs. RHE), the Raman spectra of Sr2CuWO6 displayed no change in characteristic peaks and no formation of any new peak (Supplementary Fig. 25). In addition, these characteristic peaks were also retained during 1200 s of electrolysis at −1.2 V vs. RHE (Fig. 4h). Note that, within the electrolysis time, only a new peak at about 308.2 cm−1 appeared (Fig. 4h), possibly originating from the electrochemical reduction of W6+ to W5+44. It was also observed that a main characteristic peak at 864.1 cm−1 gradually moved to lower Raman shifts (Fig. 4h), probably corresponding to the reduction-induced lattice expansion and phase transition (from I4/m to Fm-3m) as mentioned above36. However, for the CuO/WO3 during 1200 s of electrolysis (Fig. 4i), the characteristic peaks at 278.1 and 331.2 cm−1 of CuO gradually disappeared, while a characteristic peak at 613.6 cm−1 of Cu2O appeared and then disappeared, and the intensity of Cu–CO peaks at 294.2 and 385.6 cm−1 was gradually improved45. These indicate that the CuO in CuO/WO3 is initially reduced to Cu2O and then to metallic Cu.

In light of the above analyses, we plotted out the structural evolution diagrams to graphically describe the key role of superexchange interaction in stabilizing the Cu sites in Sr2CuWO6 during CO2RR (Fig. 5). As shown in Fig. 5a, once the CO2RR was initiated, the Cu2+ on Sr2CuWO6 surface began to be reduced to Cu+ by the electrode-supplied electrons. At the same time, the Sr2CuWO6 surface was transformed from the I4/m to Fm-3m phase, still being the B-site rock-salt ordered structure. Further reduction was not able to convert the Cu+ to Cu0 but rather led to the transformation of W6+ to W5+ on the surface while maintaining the Fm-3m phase. The most plausible reason was that, during CO2RR, the superexchange interaction effectively transferred the electrode-supplied excessive electrons accumulated around the Cu+ sites to W6+ sites (to form W5+) through the fast electron transport channels (Fig. 2g), thereby protecting the Cu sites from electron attack and preserving the double perovskite phase. By contrast, the CuO/WO3 without superexchange interaction was successively reduced to Cu2O/WO3 and Cu/WO3 (Fig. 5b). Taken together, during CO2RR, although these changes occurred on the Sr2CuWO6 surface, the superexchange interaction prevented structural collapse, stabilized the Cu+ sites, and maintained the long Cu–Cu distances, thereby promoting the efficient and stable CO2-to-CH4 conversion.

Fig. 5: Schematic illustrations of catalyst structure evolution during CO2RR.
figure 5

a Sr2CuWO6. b CuO/WO3. Sr, Cu, W, and O are represented by green, blue, gray, and red dots, respectively. The red dotted circle represents oxygen vacancy.


Employing Sr2CuWO6 as the proof-of-concept catalyst, we have developed Cu-based rock-salt-ordered double perovskite oxides for efficient and stable CO2-to-CH4 conversion and uncovered the key roles of their unique physicochemical properties in boosting activity, selectivity, and stability toward CH4 production. In the rock-salt-ordered structure, the corner-linked CuO6 and WO6 octahedra alternated in all three crystallographic dimensions, leading to sufficiently long Cu–Cu distances (at least 5.4 Å) and marked Cu–O–W superexchange interaction. When explored as a catalyst toward CO2RR, relative to its physical-mixture counterpart, the Sr2CuWO6 featured not only enhancements in terms of activity and selectivity for CH4 but also significantly boosted stability. Moreover, the Sr2CuWO6 was the most effective Cu-based-perovskite catalyst for CO2 methanation and performed comparably to or better than most reported representative Cu-based catalysts. According to the experiments and theoretical calculations, the superb performance could be attributed to the following factors: (i) the long-distance Cu sites facilitating *CO hydrogenation while inhibiting C–C coupling; (ii) the superexchange interaction stabilizing the Cu sites and preventing structural collapse. This work discovered efficient and stable Cu-based double perovskite oxides for CO2RR, providing a new avenue for the rational design of more advanced Cu-based catalysts.


Chemicals and materials

All chemicals were used directly without any further purification. SrCO3 (AR, ≥99%) and isopropanol (AR, ≥99.7%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Dimethyl sulfoxide (DMSO, ≥99.9%) was purchased from Shanghai Macklin Biochemical Co., Ltd. CuO (AR, 99%) and WO3 (AR, 99.8%) were purchased from Shanghai Aladdin Biochemical Technology. Nafion 117 solution (5 wt%) and D2O (99.9 atom% D) were purchased from Sigma-Aldrich Biochemical Technology. HNO3 (AR, 65–68%), H2O2 (AR, 30%), HF (AR, ≥40%), and HCl (AR, 36–38%) were purchased from Xilong Scientific. High-purity CO2 gas (99.999%), Ar gas (99.999%), and 10 vol% H2–Ar gas (99.999%) were purchased from Qingdao Dehaiweiye Technology Co., Ltd.


Sr2CuWO6 was synthesized by a high-temperature solid-state reaction/high-energy ball-milling process. In a typical procedure, stoichiometric SrCO3, CuO, and WO3 were well mixed by ball-milling process and then calcined at 900 °C in Air for 12 h. The admixture was ground again and then pressed into pellets under the pressure of 10 MPa for subsequent re-calcination at 1050 °C in Air for 24 h. Finally, the as-prepared powder was ground with the high-energy ball-milling process (900 rpm) to obtain uniform nanoparticles. The high-temperature reducing-atmosphere treatments of Sr2CuWO6 were processed in a sealed tube furnace in 10 vol% H2–Ar mixture with a flow rate of 20 mL min−1 for 1 h.

Theoretical calculations

First-principles calculations were carried out on the basis of periodic density functional theory (DFT) using a generalized gradient approximation within the Perdew–Burke–Ernzerh of exchange correction functional with Vienna ab initio simulation package (VASP)46,47. Geometry optimization was conducted in Sr2CuWO6, CuO, and WO3. The wave functions were constructed from the expansion of plane waves with an energy cutoff of 450 eV. Gamma-centered k-point of 3 × 3 × 1 has been used. The consistence tolerances for the geometry optimization were set as 1.0 × 10−6 eV/atom for total energy and 0.02 eV/Å for force, respectively. In order to avoid the interaction between the two surfaces, a large vacuum gap of 15 Å has been selected in the periodically repeated slabs. Static calculations were conducted with a convergence condition of 1.0 × 10−6 eV for density of state (DOS), Bader charge, and electron localization function analysis. The band center of Cu 3d or O 2p was calculated using the following equation48:

$${{E}}_{{{{{{\rm{t}}}}}}}=\frac{{\int }_{-\infty }^{\infty }{E}\cdot {T}({E}){{{{{\rm{d}}}}}}{E}}{{\int }_{-\infty }^{\infty }{T}({E}){{{{{\rm{d}}}}}}{E}}$$

where T(E) is the density of states (DOS) of orbitals. E corresponds to the occupied state ranges below the fermi energy level (EF) in DOS. Climbing image nudged elastic band (CI-NEB) was used for transition state searching. In free energy calculations, the entropic corrections and zero-point energy (ZPE) have been included. The free energy of species was calculated according to the standard formula:

$$\Delta G=E+\Delta {ZPE}+\Delta H{-} \Delta {TS}$$

where ZPE is the zero-point energy, ΔH is the integrated heat capacity, T is the temperature of the product, and S is the entropy.


X-ray diffraction (XRD) patterns were recorded by Rigaku Miniflex 600 (Hitachi) diffractometer with Cu Kα radiation (1.5418 Å). The Rietveld refinements of obtained data were conducted using FullProf software. Scanning electron microscopy (SEM) images were taken by a Hitachi S4800 microscope. Transmission electron microscopy (TEM) images were taken by a JEOL 2010F microscope (operated at 200 kV). To further confirm the structure and elements distribution, high-resolution TEM (HRTEM) and energy dispersive X-ray (EDX) spectra/mappings were performed on a JEOL ARM 300 F microscope equipped with dual EDX detectors. X-ray photoelectron spectroscopy (XPS) analyses were carried out by the Thermo ESCALAB 250Xi spectrometer with monochromated Al Kα radiation (hv = 1486.6 eV) operating at 150 W. The energies of each element were calibrated by the adventitious C1s (284.8 eV). Raman spectra were performed on a Renishaw Qontor spectrometer equipped with a 532 nm laser beam and a ×63 water-immersion objective lens. X-ray absorption spectroscopy (XAS) of Cu K-edge and W L3-edge were obtained in a Singapore synchrotron light source (SSLS), using an XAFCA Beamline (operated at 700 MeV) with a maximum current of 200 mA. The reference samples, such as CuO, Cu foil, WO3, and W foil were also measured for comparison and energy calibration. All XANES data were measured in transmission mode using an ion chamber detector with a Si 111 monochromator and analyzed by the Athena program49. The nitrogen adsorption and desorption processes were recorded on an Autosorb-iO (Quantachrome) device at the boiling point of liquid nitrogen to calculate the specific surface areas by the Brunauer–Emmett–Teller (BET) method. The inductively coupled plasma mass spectrometer (ICP–MS) (Agilent 730) was applied to test the metal contents of Sr2CuWO6. The samples for ICP–MS analysis were obtained by dissolving 10 mg sample powder with the mixture of 5 mL HNO3, 1 mL H2O2, 1 mL HCl, and 0.5 mL HF in the oven at 180 °C for 8 h. The cooled-down solution was further diluted to a level of 100 ppb by using a 1% HNO3 solution.

Preparation of working electrodes

The working electrodes were prepared by coating the catalyst ink onto the hydrophobic carbon paper (i.e. the gas diffusion layer, GDL). To be specific, for the preparation of the catalyst ink, 10 mg sample powder was homogenously dispersed into a mixed solution of isopropanol (1 mL) and Nafion (50 μL) by ultrasonic processing for 1 h. The catalyst ink was then coated on the hydrophobic carbon paper (Toray, YLS-30T, 1.5 × 1.5 cm2) with a loading amount of 0.5 mg cm−2 and dried under the infrared lamp. This method was used to prepare the working electrodes for both electrochemical measurements and the in-situ Raman spectroscopic measurements.

Electrochemical measurement

The CO2 electrochemical reduction measurements were processed in a homemade flow cell with a three-electrode system controlled by a CS310M electrochemical workstation (Wuhan, Corrtest). The Ag/AgCl electrode (filled with saturated KCl solution) and Pt mesh were used as the reference and counter electrodes, respectively. 1 M KOH was used as the electrolyte, filling, and cycling in the flow cell with a pumped rate of 20 mL min−1 controlled by a double channel peristaltic pump (Hebei, Leadfluid, BQ80s). An anion-exchange membrane (Hefei, ChemJoy Polymer Materials Co., Ltd., SYMA-2) was used for separating the anodic and cathodic compartments to avoid crossover pollution. High-purity CO2 gas was continuously supplied into the gas chamber with a flow rate of 35 mL min−1 controlled by a mass flow controller (D07-19B, Sevenstar Electronics Co., Ltd, Beijing) and the flow rate was further verified by a soap bubble flowmeter. The LSV curves were also recorded in the flow-cell configuration flowed with Ar or CO2 gas at a scan rate of 10 mV s−1. All applied potentials were converted into the standard reversible hydrogen electrode (RHE) potentials by the equation of ERHE = EAg/AgCl + 0.197 V + 0.0591 V × pH, with 70% iR compensation. The cell resistance was measured using the function of Rs measurement in the measurement soft of Corrtest CS310M electrochemical workstation (the value at 10,000 HZ from the electrochemical impedance spectroscopy) under open circuit potentials before every independence test.

Quantification of products

The gas products were detected by online gas chromatography (GC2060, Ramiin, Shanghai) equipped with flame ionization (FID) and thermal conductivity (TCD) detectors. A standard gas mixture (containing 1 vol% each of H2, CO, CH4, C2H4, C2H6, and 95 vol% CO2) was used to calibrate the gas products. The Faradaic efficiency (FE) of each gas product under different current densities was gained based on more than three parallel experiments. After the reaction, the catholyte was collected for liquid product analyses by a Nuclear magnetic resonance spectrometer (NMR, Bruker, AVANCE-III 600 Hz). Typically, 2 mL catholyte was mixed with 100 μL 5 mM DMSO (as an internal standard substance). And then 250 μL mixture was mixed with 350 μL D2O for the NMR measurement. The FEs of the products were calculated by the following equation:

$${{\rm {FE}}}=\frac{{Q}_{{{\rm {product}}}}}{{Q}_{{{\rm {total}}}}}=\frac{n{{\cdot }}N{\cdot F}}{{Q}_{{{\rm {total}}}}}\,$$

where Qproduct and Qtotal present the charge consumption of the target product and CO2RR process, respectively, n presents the electron transfer number of the target product, N presents the amount of substance for the product and can be calculated from the product concentration, and F presents the faradaic content (96,485 C mol−1).

In-situ Raman test

The in-situ Raman test was processed in an electrochemical operando cell (C031-2, Tianjin Gaoss Union Technology Co. Ltd.) with a three-electrode system. The in-situ Raman spectra were recorded by the Renishaw Qontor spectrometer using a 532 nm laser beam and a ×63 water-immersion objective lens. The Ag/AgCl (filled with saturated KCl solution) and graphite electrodes were used as reference and counter electrodes, respectively. The carbon paper coated with catalyst ink was used as the cathode, immersed in the CO2-saturated 0.1 M KHCO3 electrolyte. The 0.1 M KHCO3 was filled in the cathodic compartment while flowing in the anodic compartment with a flow rate of 20 mL min−1 to remove bubbles. The water-immersion objective lens was immersed in the cathodic compartment to directly observe the surface of the catalyst. The cathodic and anodic compartments were separated by a Nafion 117 proton exchange membrane. The power of the laser was kept at 1 mW to avoid irradiation damage on the catalyst. The surface exposure time was 20 s, and each signal line was collected twice. All Raman raw data were recorded and processed by Wire soft.