Achieving enhanced and robust propane selective oxidative dehydrogenation via electrically-driven continuous chemical looping Authors: Shihui Zhang, Chang-An Zhou, Jinhuang Cai, Hongjiao Li, Tao Liu, Qiang Hu, Chao Wang, Lei Song, Jiawei Xie, Lirong Zheng, Bin Liang, Hairong Yue, Kui Ma Published: 2026 ABSTRACT Oxidative propane dehydrogenation (OPDH) has emerged as a promising approach for direct propylene production. However, it still confronts great challenge of overoxidation of propane to COx, damaging the products selectivity. Herein, we report an electrically-driven continuous chemical looping process to efficiently produce propylene via an electrochemical CO2-OPDH system within solid oxide electrolysis cells (SOECs). At an optimal current density of 10 mA cm−2, the system with a Co3O4-modified La0.6Sr0.4Co0.2Fe0.8O3-δ (LSCF) anode achieves 92.6% olefin selectivity and 14.2% single-pass propane conversion. This current-driven configuration boosts the propylene yield (5.11 mmol gtotal−1 h−1) by a factor of 18.5 over the open-circuit baseline. The electrochemical CO2-ODHP system also exhibits good stability during a 120-hour durability test. In-situ characterization and theoretical calculations elucidate the electrically driven online replenishment mechanism of Co3O4 lattice oxygen by O2- derived from cathode CO2 reduction. This dynamic cycle maintains moderate anode surface oxygen activity, resulting in highly efficient and selective OPDH without significant over-oxidation or cracking of propane. Introduction Propylene (C3H6), a fundamental feedstock for synthesizing acrolein, polypyrene, acetone, and various industrial chemicals1,2, is predominantly produced through steam cracking of naphtha-based heavy hydrocarbons or fluid catalytic cracking of oil fractions. The increasing exploitation of shale gas resources has positioned direct propane dehydrogenation (PDH) as a promising alternative for propylene production3. Nevertheless, the PDH process still encounters substantial challenges, including high endothermicity, limited thermodynamic equilibrium conversion, undesirable C-C bond cracking, and rapid catalyst deactivation due to coke formation4,5, despite the corresponding industrial units having been established. Besides, the commercial PDH projects utilizing Cr-based or Pt-based catalysts face limitations due to the environmental toxicity of Cr(Ⅵ) compounds and the high cost and poor stability of Pt catalysts6,7. It is an attractive alternative to produce propylene through oxidative propane dehydrogenation (OPDH). OPDH offers exothermic reaction characteristics that circumvent thermodynamic equilibrium limitations while simultaneously mitigating catalyst coking and deactivation risks8. OPDH can be facilitated through conventional oxygen-assisted dehydrogenation of propane (O2-OPDH) or by employing soft oxidants such as CO29,10, N2O11, Sulfur-containing compounds, and halogens/halides12. Although significant advancements of O2-OPDH have been achieved13,14, the challenge of olefins overoxidation to carbon oxides (COx) in the presence of O2 obstacles the attainment of high propylene selectivity and yield. Besides, the mixture of propane and oxygen also poses serious safety hazards. Chemical looping oxidative dehydrogenation (CL-ODH) has emerged as an alternative strategy to address these challenges by employing solid oxygen carrier mediums to achieve spatial or temporal separation of dehydrogenation and oxidation processes15,16. Nevertheless, the conventional CL-ODH processes require frequent regeneration of the reaction bed (Fig. 1a), increasing the operational complexity of the reaction system and posing significant engineering challenges.Fig. 1Comparison between the traditional chemical looping and the electrically-driven chemical looping system.a Schematic of the conventional CL-ODH process. The oxygen carriers are regenerated by the oxidant through batch operation. b Schematic of electrically-driven continuous chemical looping OPDH based on a SOEC system. The oxygen species are provided by the oxidant, driven by an electric field, passed through the solid electrolyte, and used to regenerate the oxygen defect formed in the anode catalysts during the OPDH process. c The influence of oxygen species characteristics on propane oxidative dehydrogenation reaction. The blue spheres, red spheres, and dotted circles represent the metal atoms, the reactive oxygen species, and the oxygen vacancies, respectively. Comparison between the traditional chemical looping and the electrically-driven chemical looping system. a Schematic of the conventional CL-ODH process. The oxygen carriers are regenerated by the oxidant through batch operation. b Schematic of electrically-driven continuous chemical looping OPDH based on a SOEC system. The oxygen species are provided by the oxidant, driven by an electric field, passed through the solid electrolyte, and used to regenerate the oxygen defect formed in the anode catalysts during the OPDH process. c The influence of oxygen species characteristics on propane oxidative dehydrogenation reaction. The blue spheres, red spheres, and dotted circles represent the metal atoms, the reactive oxygen species, and the oxygen vacancies, respectively. The OPDH utilizing CO2 as a soft oxidant presents a viable solution to suppress overoxidation and achieve the CO2 resource utilization17. However, due to the inherent thermodynamic stability (ΔGf = 394 kJ mol−1), most CO2-ODHP systems still struggle to surpass the equilibrium limit of PDH in terms of propylene yield. Although some catalytic systems have been reported to achieve approximately 50% propane conversion with 95% propylene selectivity at 550 °C18–20, they generally face the bottleneck of frequent catalyst regeneration. Furthermore, the demanding requirement for conventional CO2-ODHP catalysts to simultaneously activate C-H bonds, adsorb/activate CO2, and suppress side reactions poses a major design challenge. It should also be emphasized that the weak capability of CO2 in regenerating oxygen vacancies and removing coke makes it difficult to apply in thermal CL-ODH systems. Recently, the electrochemical conversion of alkanes based on solid oxide electrolysis cells (SOECs) has attracted considerable attention21–23. This approach leverages an electrochemical device to decouple the alkane oxidation process into two half-reactions: oxidant electrolysis at the cathode and alkane electrooxidation at the anode. Through electrochemical control over the migration behavior of oxygen species, this strategy offers a feasible framework for balancing coke formation and overoxidation through precise current regulation22–24. Furthermore, surface modifications have proven effective in tuning oxygen species and adsorption properties, thereby enhancing oxidative dehydrogenation activity25. This electrochemical approach presents a promising alternative to conventional processes by bypassing challenges such as dry reforming and separations in CO2-ODHP, simplifying catalyst design, and overcoming the complexities of CL-ODH operations. However, the low C-C bond energy of propane predisposes it to cracking (yielding CH4, C2H4) alongside dehydrogenation, accounting for the poor propylene selectivity in already scarce electrochemical CO2-ODHP literature26–28. Consequently, a deeper understanding of the interfacial reaction mechanisms and rational electrode material design are crucial for achieving selective electrochemical propane-to-propylene conversion. Herein, we propose an electrochemically controllable strategy based on continuous chemical looping for oxidative propane dehydrogenation coupled with CO2 electrolysis. This approach achieves high selectivity and stability in the OPDH process through modification of the LSCF anode and current-driven dynamic lattice oxygen regeneration (Fig. 1b). The Co3O4-modified LSCF anode (Co3O4-LSCF-0.26) exhibits a high olefin selectivity of 92.6% and a propane conversion of 14.2% under an optimal current density of 10 mA cm−2 at 600 °C. Comprehensive characterization reveals that the current-driven migration of O2- continuously replenishes the consumed lattice oxygen in the Co3O4 anode. This dynamic equilibrium maintains the population of active oxygen species on the anode surface, ensuring high propane conversion and reaction stability. Density functional theory (DFT) calculations further elucidate that the appropriate oxygen activity of the Co3O4-LSCF electrode is critical for maintaining the balance between propane dehydrogenation conversion and propylene selectivity. This study provides an alternative approach for developing a highly selective and stable propane-to-propylene conversion process. Results CO2-OPDH system and characterizations of Co3O4-LSCF anode The electrically-driven continuous chemical looping CO2-OPDH system was illustrated in Fig. 1b. The electrolytic reaction system employs yttria-stabilized zirconia (YSZ) as the solid electrolyte, with Ni-YSZ serving as the cathode for CO2 reduction and Co3O4-modified LSCF as the anode for propane dehydrogenation. The oxygen anions (O2-) generated during cathodic CO2 electrolysis are continuously transported to the anode under current drive, supplying active oxygen species for the oxidative dehydrogenation of propane to produce propylene. However, the migration behavior of oxygen species and their oxidation activity significantly influence the propane oxidation process at the anode. For instance, sluggish oxygen transfer can lead to an insufficient supply of surface-active oxygen, which would restrict the conversion of propane. In contrast, excessively high oxygen activity may cause over-oxidation of propane, decreasing propylene selectivity (Fig. 1c). Thus, to achieve an optimal balance between propane conversion and propylene selectivity of the propane oxidative dehydrogenation process, precise regulation of oxygen species migration behavior and surface oxygen activity is essential. Oxide modification is a feasible strategy for manipulating the oxygen species features at the reaction interface to enhance the performance of electrocatalytic oxidation. We prepared LSCF anodes with different cobalt oxide loadings from 0, 0.13, 0.26, to 0.39 mg cm−2 by an impregnation method. The Co modified-LSCF anode was integrated into the YSZ electrolyte (Fig. 2a). Scanning electron microscopy (SEM) images clearly demonstrate the structure of electrode interface between Ni-YSZ, YSZ, and Co-modified LSCF (Fig. 2b and Supplementary Fig. 1, 2). TEM and corresponding element mapping images confirmed that cobalt oxide particles were homogenously loaded on the surface of LSCF (Supplementary Fig. 3, 4). The X-ray diffraction (XRD) patterns revealed that the (110) plane of LSCF exhibits the strongest diffraction peak (Supplementary Fig. 5). No additional peaks are observed after Co modification, indicating that the cobalt species are highly dispersed29. High-resolution TEM image shows a lattice fringe spacing of approximately 0.283 nm (Fig. 2c), corresponding to the (110) plane of LSCF, which is consistent with the XRD observations. Together, these results indicate that the (110) plane is the dominant exposed crystal facet in LSCF. HRTEM image and the corresponding fast Fourier transform (FFT) patterns indicate that the cobalt species exist as Co3O4 nanoparticles30, and the (220) plane is the preferred orientation facet of Co3O4 (Fig. 2c). As the Co loading amount increases, the size of Co3O4 particles in Co3O4-LSCF-0.13, Co3O4-LSCF-0.26, and Co3O4-LSCF-0.39 gradually increases, with average grain diameters of 7.66, 17.3, and 28.2 nm, respectively (Supplementary Fig. 6). The Raman spectrum of the Co3O4 powder reveals the presence of characteristic bands at 469, 515, 607, and 672 cm-1, corresponding to Eg, F2g, F2g and A1g mode31, respectively (Supplementary Fig. 7). In contrast, the Co3O4-modified LSCF sample displayed exclusively a prominent A1g mode attributed to the A1g symmetric Co-O stretching vibration. The notable redshift of the bands for Co species in the Co3O4-modified LSCF arises from the phonon confinement effects or lattice deformation induced by the small particle size32. FTIR spectra show that the vibration modes related to Co3O4 on the Co-modified LSCF anodes are at 667 cm-1 (Supplementary Fig. 8). X-ray photoelectron spectroscopy (XPS) was employed to investigate the surface changes of Co3O4-LSCF anodes (Supplementary Fig. 9a). Interestingly, after Co3O4 modification of LSCF, there are a negative shift of 0.2 eV and a positive shift of 0.4 eV in the binding energies of Co 2p (Fig. 2d) and Fe 2p (Fig. 2e), respectively. However, the binding energies of La 3d and Sr 3d stay unchanged (Supplementary Fig. 9b, c), which represents the electron transfer from Fe to Co due to the strong interaction between the Co3O4 nanoparticles and LSCF. The O 1s XPS spectra demonstrate that the peak of lattice oxygen in LSCF is at 528.7 eV21, while the surface chemisorbed oxygen signal is at 531.6 eV (Fig. 2f)33,34. Notably, a new O 1s peak appears at 529.9 eV in the sample modified with Co3O4, and its peak intensity enhances with the increase of Co3O4 loading amounts (Supplementary Fig. 9d). Such results confirm the contribution from the lattice oxygen of Co3O435,36.Fig. 2Characterization of the Co3O4-LSCF anodes.a A simplified schematic of the Ni-YSZ/YSZ/Co3O4-LSCF integrated electrode for electrochemical CO2-assisted OPDH. b, c SEM image (b) and TEM image (c) of Co3O4-LSCF-0.26. d, e Co 2p (d) and Fe 2p XPS spectra (e) of Co3O4-LSCF with different Co3O4 loadings. f O 1s XPS spectra of LSCF, Co3O4-LSCF, and Co3O4. The O 1s peak in the shaded region is attributed to the lattice oxygen of Co3O4. g XANES spectra of the Co K-edge. CoO, Co2O3, and Co3O4 served as references. h The corresponding Wavelet transform of the k3-weighted EXAFS data with LSCF, Co3O4 and Co3O4-LSCF-0.26. i, O2-TPD profiles of Co3O4-LSCF with different Co3O4 loadings. Source data for Fig. 2d–i are provided as a Source Data file. Characterization of the Co3O4-LSCF anodes. a A simplified schematic of the Ni-YSZ/YSZ/Co3O4-LSCF integrated electrode for electrochemical CO2-assisted OPDH. b, c SEM image (b) and TEM image (c) of Co3O4-LSCF-0.26. d, e Co 2p (d) and Fe 2p XPS spectra (e) of Co3O4-LSCF with different Co3O4 loadings. f O 1s XPS spectra of LSCF, Co3O4-LSCF, and Co3O4. The O 1s peak in the shaded region is attributed to the lattice oxygen of Co3O4. g XANES spectra of the Co K-edge. CoO, Co2O3, and Co3O4 served as references. h The corresponding Wavelet transform of the k3-weighted EXAFS data with LSCF, Co3O4 and Co3O4-LSCF-0.26. i, O2-TPD profiles of Co3O4-LSCF with different Co3O4 loadings. Source data for Fig. 2d–i are provided as a Source Data file. X-ray absorption fine structure (XAFS) analysis was performed to reveal the detailed structure information of the Co3O4-modified LSCF anode. The Co K-edge X-ray absorption near-edge structure (XANES) spectra of Co3O4-LSCF-0.26, similar to Co3O4 but distinct from CoO and Co2O3 (Fig. 2g), indicate the mixed valence of Co2+ and Co3+ (Supplementary Fig. 10)37. Extended-edge X-ray absorption fine structure (EXAFS) spectra were further used to analysis the coordination structure of Co on Co3O4-LSCF anodes. We observe a remarkable difference in the local coordination environment between LSCF and Co3O4-LSCF-0.26 (Supplementary Fig. 11). Similar to Co3O4, the Co3O4-LSCF-0.26 with Co scattering features shows a characteristic peak at 1.5 Å that corresponds to the Co-O scattering paths. Besides, we find that the intensity of the Co-Co coordination at 2.4 Å decreases significantly compared to commercial Co3O4, which may relate to the size confinement of Co3O4 nanoparticles38,39. Figure 2h shows the Wavelet transform results, in which a maximum intensity corresponding to Co-Co coordination appeared in the Co3O4-LSCF-0.26 samples. The intensity of Co-Co coordination is clearly weaker than that of Co3O4, further proving the successfully loading of small size Co3O4 nanoparticles on the LSCF anode. O2-temperature-programmed desorption (O2-TPD) were performed to reveal the reactivity of oxygen species in different anodes (Fig. 2i). Two significant oxygen desorption peaks emerge after loading Co3O4 onto the anode surface. The desorption peak that appears at 600-800 °C is attributed to the spill flow of bulk Co3O4 lattice oxygen40, while the peak ~320 °C is ascribed to the desorption of active surface oxygen41. The surface oxygen desorption peak intensity gradually increases as the loading amounts of Co3O4 increases, while the lattice oxygen overflow temperature decreases noticeably. Such results demonstrate that the interaction between the LSCF and the lattice oxygen of Co3O4 weakens with the growth of Co3O4 particle size. H2-temperature-programmed reduction (H2-TPR) profiles (Supplementary Fig. 12) show that the onset reduction temperature of the Co3O4-modified LSCF anodes is decreased obviously and a new reduction peak appears at ~350 °C with high Co3O4 loading, suggesting the enhancement of oxygen mobility. Therefore, we concluded that both the amount of lattice oxygen species and the oxygen mobility are increased due to the decoration of Co3O4 on LSCF. Further, the electrochemical impedance spectra (EIS) of the constructed Co3O4-LSCF/YSZ/Ni-YSZ cell show that the polarization resistance decreases with the loading of Co3O4 (Supplementary Fig. 13). The decreasing electrode polarization resistance reflects the improvement of the electrode reaction kinetics. All in all, because of the enhanced ability to generate active oxygen species, Co3O4-modified LSCF anodes are expected to achieve high electrochemical OPDH activities42,43. Electrochemical performance of continuous chemical looping CO2-OPDH system The electrochemical CO2-OPDH activities of the Co3O4-LSCF anodes were measured on a homemade device presented in Supplementary Fig. 14. Based on our prior findings, the CO2-OPDH reaction initiates above 600 °C44. While elevating reaction temperature better matches the optimal operating condition of YSZ, thereby enhancing propane conversion, it also promotes propane cracking and over-oxidation, which depresses olefin selectivity and ultimately provides no gain in propylene yield (Supplementary Fig. 15). Hence, the reaction temperature of 600 °C was selected for this work. Figures 3a and 3b display the anode product analysis in the electrically-driven continuous chemical looping propane oxidative dehydrogenation while CO2 electrolysis is simultaneously conducted in cathode. We obtained propylene and ethylene are the dominant products, while C2H6, CH4, CO, and CO2 are also present as by-products. We find that the modification of the LSCF anode with Co3O4 significantly promotes the conversion of propane and improves olefin products selectivity. The olefin selectivity is ~ 77.0% for unmodified LSCF anode and the propane conversion is barely 9.4% (at 600 °C, 10 mA cm-2). With 0.13 mg cm-2 Co3O4 loading, we increase the olefin selectivity and the propane conversion to nearly 90.6% and 10.7%, respectively. The olefin selectivity further increases to 92.6% (over 80% for propylene), and the propane conversion rises to 14.2% for 0.26 mg cm-2 Co3O4 modified anode. Nevertheless, as the Co3O4 loading amounts reaches 0.39 mg cm-2, the propylene selectivity decreases to 74.3% while the COx selectivity further increases to 8.4%, suggesting the excessive oxidation of propane. Notably, methane selectivity on Co3O4-LSCF-0.26 electrode drops remarkably to 4.3% (11.9% for unmodified LSCF), and ethylene selectivity decreases from 25.5% (unmodified LSCF) to 12.0%. All these results indicate that the modification of Co3O4 effectively inhibits the cracking of propane.Fig. 3Electrochemical CO2-OPDH performance.a Products selectivity and propane conversion on anodes with different Co3O4 loading at a driven current density of 10 mA cm-2, 600 °C (Cathode: 10% CO2 + 5% N2 + 85% Ar, with a flow rate of 2.5 mL min-1; Anode: 5% C3H8 + 5% N2 + 90% Ar, with a flow rate of 10 mL min-1). b Selectivity and propane conversion on Co3O4-LSCF-0.26 anode at different driven currents. The activity data are derived from a single test. c Stability test for electrochemical CO2-OPDH reaction system (Ni-YSZ/YSZ/Co3O4-LSCF-0.26, 10 mA cm-2 at 600 °C). The rose squares, open orange circles, and blue balls denote the cell voltage, olefin selectivity, and propane conversion, respectively. The cell voltage without iR corrected. Source data for Fig. 3 are provided as a Source Data file. Electrochemical CO2-OPDH performance. a Products selectivity and propane conversion on anodes with different Co3O4 loading at a driven current density of 10 mA cm-2, 600 °C (Cathode: 10% CO2 + 5% N2 + 85% Ar, with a flow rate of 2.5 mL min-1; Anode: 5% C3H8 + 5% N2 + 90% Ar, with a flow rate of 10 mL min-1). b Selectivity and propane conversion on Co3O4-LSCF-0.26 anode at different driven currents. The activity data are derived from a single test. c Stability test for electrochemical CO2-OPDH reaction system (Ni-YSZ/YSZ/Co3O4-LSCF-0.26, 10 mA cm-2 at 600 °C). The rose squares, open orange circles, and blue balls denote the cell voltage, olefin selectivity, and propane conversion, respectively. The cell voltage without iR corrected. Source data for Fig. 3 are provided as a Source Data file. Figure 3b demonstrates that the current has a significant impact on the conversion and selectivity of the electrochemical OPDH process. With open circuit conditions (only thermal catalytic contributions are present), the conversion of propane on the Co3O4-LSCF-0.26 anode is merely 1.1%, with a propylene selectivity of 56.2%. In the absence of electric field assistance, CO2 on the cathode side cannot be readily activated, failing to supply sufficient oxygen species to the anode side, which results in a low propane conversion rate. After applying the working current, the propane conversion increases from 5.8% at 4 mA cm-2 to 16.3% at 13 mA cm-2 (Supplementary Fig. 16). At the optimal current density of 10 mA cm-2, the propylene selectivity reaches a maximum value of 80.6% with a propylene yield of 5.11 mmol gtotal-1 h-1, which is almost 18.5 times higher than that of the open circuit condition (0.28 mmol gtotal-1 h-1). Further increasing the current density to 13 mA cm-2 results in a 5% increase in the proportion of CO2 in the product, indicating excessive oxidation of propane. It should be emphasized that, within the current density range of 4-13 mA cm-2, the anodic oxygen production rate semi-quantitatively measured by mass spectrometry is linearly correlated with the current density (Supplementary Fig. 17). Higher current densities provide more active oxygen species derived from CO2 electrolysis to the anode, thereby increasing the propane conversion rate. Moreover, the oxygen flux at the corresponding current density is significantly higher than the actual oxygen demand required for propane conversion. This indicates that, under the present reaction conditions, oxygen transport is not the limiting factor responsible for the low propane conversion rate. In addition to propylene generation on the anode side, we also successfully reduce CO2 to produce CO, which is the only detected product on the cathode side (Supplementary Fig. 18). The in-situ current-programmed surface reaction (CPSR) experimental results reveal a distinct correlation between work current density and catalytic conversion behavior of propane. Mass spectrometry analysis demonstrates that the signal of C3H6 (m/z = 42) significantly increases with the introduction of current, while the H2 signal (m/z = 2), which mainly originates from propane cracking, simultaneously weakens (Supplementary Fig. 19). The C3H6 signal intensity exhibits a positive correlation with the magnitude of applied current density. These observations suggest that the current-driven migration of oxygen species facilitates the selective oxidative dehydrogenation of propane on the anode while simultaneously suppressing the competing cracking pathway. Further, the constructed Ni-YSZ/YSZ/Co3O4-LSCF-0.26 electrochemical system for CO2-OPDH exhibits good stability. During a 7-hour testing period (600 °C and 10 mA cm-2), the system maintained a stable propane conversion of ~ 13.0% without obvious degradation. Besides, the olefin selectivity remains consistently above 90%, with a propylene selectivity exceeding 78% throughout the whole experiment. In contrast, we observe a sudden decline in the propylene selectivity after 4.5 hours of reaction with open circuit (Supplementary Fig. 20). The observed behavior is likely due to the progressive consumption of surface-active oxygen without timely replenishment, leading to the reaction pathway of propane switching from oxidative dehydrogenation to cracking. During a 160-hour durability test, the CO2-OPDH continuous chemical looping strategy system exhibited good stability. During the first 125 h, no decay in propane conversion was observed (100% retention), and olefin selectivity decreased by only 2% (Fig. 3c). The average propane conversion was 13.6% over this period, and the average olefin selectivity reached 82.6%. The propane conversion began to drop to 11% at 130 h, and further decreased to 8% after 160 h of operation. Notably, the olefin selectivity remained above 80% throughout the entire test, whereas the propane conversion only decreased at the final stage. This is likely attributable to carbon coking on the Ni-YSZ cathode45,46, which leads to an insufficient supply of active oxygen. Under the optimal operation conditions (10 mA cm-2, 0.26 mg cm-2 Co3O4 loading amount), our work reaches a propylene selectively of 80.6%, which is comparable to the reported performance for the electrochemical CO2-OPDH system (Supplementary Fig. 21a). It is important to note that while previous literature has reported propane conversion exceeding 70%, these achievements typically require elevated reaction temperature (over 750 °C) and the usage of stronger oxidants (such as O2), which result in significantly compromised propylene selectivity, typically below 40% (Supplementary Table 1). Although the overall propane conversion remains relatively low, the propylene space-time yield normalized to the total catalyst mass is already comparable to that reported for thermal catalytic chemical looping systems. Our work indicates that the electrically-driven continuous chemical looping CO2-OPDH system demonstrates significant advancements in operational efficiency over the conventional CL-ODH technique, particularly by eliminating the necessity for periodic regeneration cycles to sustain catalytic activity (Supplementary Fig. 21b). In-situ characterizations of Co3O4-LSCF anode and oxygen migration behavior during CO2-OPDH process The quasi-operando XPS analysis was conducted to investigate the evolution of catalyst states during the electrochemical CO2-OPDH process. Analysis of O 1s XPS spectra demonstrates that the lattice oxygen (LO) content in Co3O4 decreases from 35.9% of the fresh electrode to 27.1% after 7 hours of open circuit operation. This indicates that lattice oxygen was consumed to compensate for the insufficient supply of reactive oxygen species, resulting in the reduction of Co3O4. Notably, with 10 mA cm-2 current density applied, the LO content of Co3O4 increases to 37.7%, implying the replenishment of lattice oxygen in Co3O4 by current-driven (Fig. 4a and Supplementary Table 2). The evolution in cobalt species further corroborates this inference. As shown in Supplementary Fig. 22, the proportion of Co2+ species increased from 64.1% in the fresh anode to 73.4% after 7 h reaction of under open circuit conditions (Supplementary Table 3), indicating the reduction of Co3O4 during propane dehydrogenation. In contrast, the distribution of cobalt species was consistent with that of the fresh electrode after 7 h of current-driven propane dehydrogenation. The quasi-operando XANES reveals that the absorption edge of Co shifts towards low energy compared to the fresh anode after 2 hours open circuit operation, and prolonging the reaction times further intensifies this trend (Supplementary Fig. 23). This observation confirms the reduction of Co3O4 is caused by lattice oxygen consumption during the OPDH process, which is only driven by thermal catalysis. Conversely, applying of 7 mA cm-2 reaction current density maintains the Co absorption edge at the same position as that of the fresh anode. When the current density is increased to 10 mA cm-2, the absorption edge shifts toward higher energy, demonstrating the maintenance of high-valence cobalt species in an oxygen-enriched environment. EXAFS analysis shows a significant decrease in Co-O coordination scattering intensity during open circuit OPDH, while the coordination structure remains stable with current applying (Fig. 4b). Wavelet transform analysis further corroborates these findings, proving the degradation of Co-O coordination due to lattice oxygen consumption with open circuit conditions and the stabilization of the coordination structure under current-driven operations through oxygen replenishment (Fig. 4c).Fig. 4In-situ characterization and oxygen migration mechanism.a O 1s spectra of Co3O4-LSCF-0.26 anode after 7 hours of propane dehydrogenation under different conditions. The shaded region is attributed to the lattice oxygen of Co3O4. b Fourier transforms of the Co K-edge EXAFS spectra under different reaction currents. c The corresponding Wavelet transforms of the k3-weighted EXAFS data. d Photograph of an operating cell with the Raman detector. e–g In-situ Raman spectra of Co3O4-LSCF-0.26 under (e) open circuit, (f) a drive current of 5 mA, and (g) return to open circuit. (550 °C, cathode: 5% CO2, 5% N2, 90% Ar; anode: 5% C3H8, 5% N2, 90% Ar). h Schematic illustration of oxygen migration behavior and reaction mechanism in electrochemical CO2-OPDH process. Source data for Fig. 4a–c, e are provided as a Source Data file. In-situ characterization and oxygen migration mechanism. a O 1s spectra of Co3O4-LSCF-0.26 anode after 7 hours of propane dehydrogenation under different conditions. The shaded region is attributed to the lattice oxygen of Co3O4. b Fourier transforms of the Co K-edge EXAFS spectra under different reaction currents. c The corresponding Wavelet transforms of the k3-weighted EXAFS data. d Photograph of an operating cell with the Raman detector. e–g In-situ Raman spectra of Co3O4-LSCF-0.26 under (e) open circuit, (f) a drive current of 5 mA, and (g) return to open circuit. (550 °C, cathode: 5% CO2, 5% N2, 90% Ar; anode: 5% C3H8, 5% N2, 90% Ar). h Schematic illustration of oxygen migration behavior and reaction mechanism in electrochemical CO2-OPDH process. Source data for Fig. 4a–c, e are provided as a Source Data file. In-situ Raman spectroscopy was employed to further characterize the dynamic structural evolution of catalysts during the OPDH process (Fig. 4d and Supplementary Fig. 24). In the continuous Raman spectra, the Co3O4-LSCF electrode exhibits two characteristic bands at approximately 448 cm-1 and 678 cm-1, which correspond to the Eg and A1g vibrational modes, respectively. Under open circuit conditions, the A1g band of Co3O4 at ~ 678 cm-1 gradually redshifts to 650 cm-1 as propane dehydrogenation progresses (Fig. 4e), which should be attributed to the formation of oxygen vacancies37,47. However, upon current introduced, the A1g band exhibited a gradually blueshift, suggesting a replenishment of oxygen vacancies within the Co3O4 lattice (Fig. 4f). More importantly, when external current is interrupted, the A1g band redshifts again to 650 cm-1 (Fig. 4g). The observed reversible behavior highlights a dynamic equilibrium of lattice oxygen in Co3O4 during CO2-OPDH process. This equilibrium is sustained by the potential-driven migration of O2- from the cathode, which continuously compensates for the consumed lattice oxygen at the anode, thereby ensuring structural stability. Based on the above findings, we propose that the current-driven CO2-OPDH process follows an electrochemical looping mechanism, which involves a dynamic equilibrium between the consumption of surface-active oxygen and the continuous replenishment by lattice oxygen (Fig. 4h). Under open-circuit conditions, the surface-active oxygen on the Co3O4-LSCF anode is initially consumed for propane. Driven by the oxygen chemical potential, lattice oxygen from Co3O4 migrates to the surface to replenish the consumed active oxygen. However, the limited reservoir of Co3O4 lattice oxygen, coupled with the absence of current-driven O2- replenishment, leads to reaction slowdown or termination once the lattice oxygen is depleted. This explains the very low propane conversion observed under open-circuit conditions (Fig. 3b). Upon applying an electrical bias, O2- generated from CO2 electrolysis migrates continuously and efficiently to the anode, significantly enhancing propane conversion. Nevertheless, the unmodified LSCF anode tends to favor propane cracking due to the intrinsic activity of its surface oxygen species. The introduction of Co3O4 effectively addresses this issue: it not only provides highly active lattice and surface oxygen species, but also serves as a rapid ion-conducting pathway for O2- migration. Meanwhile, the highly active surface oxygen significantly promotes the oxidative propane dehydrogenation pathway, thereby synergistically improving the production efficiency of propylene (Fig. 3a). Mechanism of propane dehydrogenation over Co3O4-LSCF catalyst The atomic-scale details of surface oxygen activity and surface reactions over the Co3O4-LSCF catalyst were investigated by density functional theory (DFT) calculations. Based on the characterization results of XRD and TEM, a Co3O4 (220) slab, an exposed LSCF (110) slab, and a LSCF (110) slab with single-layer Co3O4 composite were chosen to simulate the large-sized Co3O4 particles for Co3O4-LSCF-0.39, the Co3O4-LSCF-0, and the Co3O4-LSCF-0.26 anodes, respectively (Supplementary Data 1-3). Despite the challenges in modeling realistic materials under catalytic conditions, DFT results from simplified models can still offer useful references for mechanistic interpretation. The structures of Co3O4 (220), LSCF (110) and Co3O4-LSCF (110) are fully optimized, as shown in Supplementary Figs. 25 and 26. DFT results show that the VO formation energy (Fig. 5a and Supplementary Table 4) increases as Co3O4 (0.12 eV) < Co3O4-LSCF (0.78 eV) < LSCF (1.14 eV) (Supplementary Fig. 27). LSCF exhibits the highest oxygen vacancy formation energy, indicating poor oxygen mobility, whereas Co3O4 shows the lowest oxygen vacancy formation energy, suggesting that its surface oxygen species are highly reactive. The moderate VO formation energy in the Co3O4-LSCF interfaces implies that the interaction between LSCF and Co3O4 suppresses the mobility of interfacial oxygen. These results are consistent with the O2-TPD and XPS data, which may explain the observed suppression of overoxidation. The moderate oxygen activity at the Co3O4-LSCF interface may enhance the oxidative dehydrogenation performance without inducing over-oxidation.Fig. 5DFT calculations and propane dehydrogenation mechanism.a The structural model and corresponding formation energy of oxygen vacancy over three systems: Co3O4 surface, LSCF surface, and Co3O4-LSCF interface. Orange, white, and purple spheres and red-circled grey spheres indicate Co, O, Fe, and oxygen vacancy sites, respectively. b The energy diagrams of H2O formation and desorption. c The energy diagrams of the adsorption, dehydrogenation, and dissociation of propane. d Schematic illustration of the dehydrogenation mechanism of propane on three catalyst interfaces. Grey, pink, rose-red, red, and aqua circles indicate metal atoms in LSCF, lattice oxygen, interfacial reactive oxygen species, adsorbed O2, and Co3O4 particles, respectively. Source data for Fig. 5 are provided as a Source Data file, and DFT computational models for Fig. 5a–c are provided as a Supplementary Data file. DFT calculations and propane dehydrogenation mechanism. a The structural model and corresponding formation energy of oxygen vacancy over three systems: Co3O4 surface, LSCF surface, and Co3O4-LSCF interface. Orange, white, and purple spheres and red-circled grey spheres indicate Co, O, Fe, and oxygen vacancy sites, respectively. b The energy diagrams of H2O formation and desorption. c The energy diagrams of the adsorption, dehydrogenation, and dissociation of propane. d Schematic illustration of the dehydrogenation mechanism of propane on three catalyst interfaces. Grey, pink, rose-red, red, and aqua circles indicate metal atoms in LSCF, lattice oxygen, interfacial reactive oxygen species, adsorbed O2, and Co3O4 particles, respectively. Source data for Fig. 5 are provided as a Source Data file, and DFT computational models for Fig. 5a–c are provided as a Supplementary Data file. The formation and desorption of H2O on the catalyst surface are critical steps in the oxidative dehydrogenation of alkanes48. As shown in Fig. 5b, the energy required for the conversion of *OH to H2O on LSCF (110) is the highest, at 0.74 eV, whereas on Co3O4-LSCF (110) and Co3O4 (220), the corresponding values are 0.14 eV and 0.18 eV, respectively. Notably, although the H2O formation energies on Co3O4 and Co3O4-LSCF are comparable, the energy required for H2O desorption from the Co3O4 surface is as high as 1.05 eV, higher than that from Co3O4-LSCF (0.85 eV). During alkane oxidative dehydrogenation, the difficulty in forming H2O or its hindered desorption from the catalyst surface can both constrain the reaction activity. Co3O4‑modified LSCF exhibits moderate energy requirements for both steps, which is conducive to promoting propane conversion. This finding is consistent with the experimental data, where Co3O4-LSCF-0.26 shows the highest propane conversion after anodic current was applied, rather than Co3O4-LSCF-0.39, which possesses higher oxygen activity. The energy diagrams of propane adsorption and dissociation are shown in Fig. 5c and Supplementary Fig. 28. The adsorption energy in the three systems follows the order of Co3O4 (220) (-0.52 eV) = Co3O4-LSCF (110) (-0.52 eV) < LSCF (110) (-0.23 eV). In propane dehydrogenation, the initial cleavage of the C-H bond is widely identified as the rate-limiting step49,50. The first dehydrogenation energy barrier in the three systems follow the order of LSCF (110) (0.25 eV) > Co3O4 (220) (0.23 eV) > Co3O4-LSCF (110) (0.20 eV) (Supplementary Fig. 29). Owing to the favorable C3H8 adsorption and significantly lower energy barrier for the first C-H bond cleavage, the dehydrogenation of C3H8 is thermodynamically more facile on Co3O4-LSCF interface compared to the Co3O4 and LSCF surface. Furthermore, the energy barrier for C-C bond cleavage of C3H7* on Co3O4 (220) is the lowest (2.06 eV), lower than that on LSCF (110) (2.34 eV) and on Co3O4-LSCF (110) (2.32 eV) (Supplementary Fig. 29), indicating that propane cracking occurs more readily on Co3O4. Considering the highly active oxygen species on its surface, the reaction intermediates are prone to be deeply oxidized to COx. LSCF exhibits inhibited oxidative dehydrogenation activity owing to its weak oxygen activity and unfavorable C-H bond breakage step, favoring propane cracking instead. This is consistent with the highest ethylene and methane yields observed with unmodified LSCF anode (Fig. 3a). Additional evidence comes from Raman spectra, where distinct carbon deposition peaks appear on the blank LSCF electrode under open-circuit conditions but not on Co3O4-LSCF-0.26 electrode (Supplementary Fig. 30). In contrast, the intrinsically high oxygen reactivity of Co3O4, combined with the favorable C-C bond breakage step, promotes over-oxidation of propane cracking products, while its strong adsorption of H2O limits the propane conversion rate. The synergistic interaction between Co3O4 and LSCF endows the Co3O4-LSCF composite with a balanced catalytic profile, featuring thermodynamically favorable propane adsorption/dehydrogenation, suppressed cracking propensity, and efficient H2O formation/desorption. The activity of oxygen species can be effectively optimized by regulating the interaction between Co3O4 and LSCF, thereby achieving a balance between propane conversion and propylene selectivity (Fig. 5d). Although Co3O4-LSCF-0.39 has the highest Co3O4 loading, it exposes the fewest highly active heterointerfaces (Supplementary Fig. 31); Co3O4-LSCF-0.13 possesses the most heterointerfaces, but suffers from poor interfacial oxygen activity due to size effects. These integrated merits collectively contribute to the highest propane conversion and olefin selectivity observed on the Co3O4-LSCF-0.26 electrode. Discussion In summary, this work presents an approach to electrochemical OPDH reaction within a solid oxide electrolysis cell. The developed system establishes a continuous chemical looping process for propane oxidative dehydrogenation by effectively integrating anodic propane oxidation with cathodic CO2 electrolysis. Under current-driven conditions, the Co3O4-LSCF electrode can provide sufficient active oxygen species with appropriate oxidation ability, thereby significantly suppressing coking and achieving a mutually beneficial solution for the conversion efficiency and selectivity of the OPDH process. In-situ characterization techniques combined with DFT calculations reveal that maintaining sufficient oxygen species at the reaction interface and precisely regulating appropriate activity are critical factors in achieving high conversion and olefin selectivity in propane oxidative dehydrogenation. Although the current density employed in this study remains below industrial standards and challenges such as the performance degradation of Ni-YSZ electrodes persist, these issues can potentially be addressed through rational electrode design and structural optimization. Overall, this work demonstrates a continuous electrochemical looping strategy for converting propane into valuable olefins, which also potentially inspires the design of other selective oxidation systems. Methods Chemicals Lanthanum nitrate hexahydrate (La(NO3)3·6H2O, AR), Strontium nitrate (Sr(NO3)2, AR), Cobalt nitrate hexahydrate (Co(NO3)2·6H2O, AR), Iron nitrate nonahydrate (Fe(NO3)3·9H2O, AR), Ethylenediaminetetraacetic acid (EDTA, AR), Citric acid (CA, 99.5%), Ethyl cellulose (AR), terpinol (AR, 94 wt.%), ammonia (NH3·H2O, 25-28%) were purchased from Kelong Chemical Reagent. All the chemicals in the experiments were used without further treatment. The integrated NiO-YSZ/YSZ electrodes were purchased from Zhejiang Zhentai Energy Technology Co., Ltd. Preparation of LSCF powder La0.6Sr0.4Co0.2Fe0.8O3-δ (LSCF) powder was prepared by the sol-gel method. Stoichiometric amounts of La(NO3)3·6H2O, Sr(NO3)2, Fe(NO3)3·9H2O and Co(NO3)2·6H2O were dissolved in deionized water. Then, a certain amount of CA and EDTA was added into the above solution with a molar ratio of CA, EDTA, and total metal cations of 1.5: 1: 1. Add an appropriate amount of ammonia to adjust the solution pH to 7-8, and heat the solution to 85 °C under stirring to obtain a viscous gel. The formed gel was first heated at 240 °C for 1 h and then calcined at 950 °C in an air atmosphere for 5 h with 5 °C min-1. The collected black solid powder is LSCF. Fabrication of the integrated electrode LSCF slurry was prepared by mixing LSCF powder and 0.5 wt.% carbon black in a mortar for 30 minutes with the addition of the solution of 5 wt.% ethyl cellulose and 95 wt.% terpinol. The integrated electrode (NiO-YSZ/YSZ/LSCF) were fabricated by printing the obtained LSCF slurry onto the YSZ side of the integrated NiO-YSZ/YSZ electrode and sintering the electrode at 1100 °C for 2 hours in the muffle furnace. For the preparation of Co3O4-modified LSCF anodes, 180, 320, and 480 μL of Co(NO3)2 solution (0.05 mol L-1) were infiltrated onto the LSCF side of the integrated electrode, respectively, and the obtained electrode was calcinated at 650 °C for 2 hours in an air atmosphere. According to the dosage of Co(NO3)2 and stoichiometric relationship, the Co3O4 loading is determined to be 0.13, 0.26, and 0.39 mg cm-2, respectively. The final electrodes were denoted as Co3O4-LSCF-0.13, Co3O4-LSCF-0.26, and Co3O4-LSCF-0.39, respectively. Physicochemical characterizations XRD patterns were acquired on Bruker D8 Advance with Cu Kα radiation as the resource. TEM images were obtained on JEOL-2011 microscope with an accelerating voltage of 200 kV. SEM images were collected from ZEISS Gemini SEM 300 with an accelerating voltage of 30 kV. FTIR spectra were recorded on ThermoScientific Nicolet iS50. XPS data were collected from PHI Versa Probe II instrument equipped with a monochromatic Al Kα source, and the binding energy calibration was based on C1s at 284.8 eV. Raman spectra were obtained using a Horiba LabRAM HR Evolution Raman spectrometer (532 nm). The AC impedance was performed on Princeton Parstat 4000 A with an amplitude of 10 mV and a frequency of 1 Hz to 10 kHz. H2-TPR profiles were collected on the Auto Chem II 2920 instrument. The sample was pretreated with He at 200 °C for 1 h followed by cooling to 50 °C, then 10% H2 + 90% Ar was switched into the system and the sample was heated to 800 °C at a rate of 10 °C min−1. O2-TPD profiles were collected on the Auto Chem II 2920 instrument equipped with a mass spectrometer (OmmistarTM). The sample was first pretreated in a He atmosphere at 250 °C for 1 h and cooled to 50 °C. The sample was pre-adsorbed for 1 h in 3% O2 + 97% He at 50 °C and then switched to pure He for 30 min to remove oxygen in the gas phase. Finally, the sample was gradually heated from 25 to 1000 °C with a heating rate of 10 °C min−1 in the pure He atmosphere. XAFS spectra of Co K-edge were measured at the 1W1B beamline of Beijing Synchrotron Radiation Facility (BSRF, operated at 2.5 GeV with a maximum current of 250 mA). The X-ray was monochromatized by a double-crystal Si (311) monochromator, and the energy was calibrated using a Co metal foil for Co K-edge. Quasi-operando XAFS were collected at anodes under different reaction currents and reaction times. The in-situ Raman experiment The in-situ Raman spectra were recorded on Renishaw in Via-Qontor (532 nm) with a homemade chamber. The cell was heated to 100 °C, and the cathode was reduced in a H2 atmosphere for 1 hour after assembly. The temperature was then raised and stabilized at the operating temperature of 550 °C. During this period, gas mixtures were introduced to the electrodes (cathode: 5% CO2, 5% N2, 90% Ar; anode: 5% C3H8, 5% N2, 90% Ar). Subsequently, Raman spectra (200–1000 cm-1) were continuously collected for 40 minutes under open-circuit conditions. A constant current of 5 mA was then applied, and spectral acquisition continued for another 40 min. Finally, the current was interrupted, and spectra were recorded for a further 40 minutes under restored open-circuit conditions. Electrochemical and catalytic performance tests The electrically-driven continuous chemical looping CO2-OPDH experiments were implemented on a home-made testing instrument at 600 °C. Before testing, the channels on both sides of the anode and cathode were purged with 100 mL min-1 Ar at 600 °C for 30 min. Then, the cathode atmosphere was switched to pure H2 with 40 mL min-1 and maintained for 1 h to reduce the NiO-YSZ to Ni-YSZ. After replacing the cathode atmosphere with Ar, a mixture gas of 10% CO2 + 5% N2 + 85% Ar was fed to the Ni-YSZ cathode with a flow rate of 2.5 mL min-1, while 5% C3H8 + 5% N2 + 90% Ar was supplied to the Co3O4-LSCF anode with a flow rate of 10 mL min-1. Ar and N2 serve as the reaction equilibrium gas and the chromatographic internal standard, respectively. The current was applied using a workstation (CHI 660E). Quantification of the gas products was performed on gas chromatograph (GC2060, Shanghai Ruimin Co., Ltd) equipped with TCD and FID detectors. The conversion of propane (XC3H8) and the selectivity of products (Si) were calculated following the equations:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${[C]}_{{total},{in}}={3[{C}_{3}{H}_{8}]}_{{in}}$$\end{document}[C]total,in=3[C3H8]in2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${[C]}_{{total},{out}}= {[{CH}4]}_{{out}}+{[{CO}]}_{{out}}+{[{CO}2]}_{{out}}+{2[C2H4]}_{{out}}+{2[C2H6]}_{{out}}\\ +{3[C3H6]}_{{out}}+{3[C3H8]}_{{out}}$$\end{document}[C]total,out=[CH4]out+[CO]out+[CO2]out+2[C2H4]out+2[C2H6]out+3[C3H6]out+3[C3H8]out3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{C3H8}={1-3[C3H8]}_{{out}}/{3[C3H8]}_{{in}}$$\end{document}XC3H8=1−3[C3H8]out/3[C3H8]in4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${S}_{C3H6}={3[C3H6]}_{out}/({[C]}_{{total},{out}}-{3[C3H8]}_{out})$$\end{document}SC3H6=3[C3H6]out/([C]total,out−3[C3H8]out)5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${S}_{{CO}2}={[{CO}2]}_{{out}}/({[C]}_{{total},{out}}-{3[C3H8]}_{{out}})$$\end{document}SCO2=[CO2]out/([C]total,out−3[C3H8]out)6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C\,balance}={[C]}_{{total},{out}}/{[C]}_{{total},{in}}\ge 92\%$$\end{document}Cbalance=[C]total,out/[C]total,in≥92%where [C]total, in and [C]total, out are the total amount of C in the feed and exit gas. [C3H8]in is the propane content in the feed, and [CH4]out, [CO]out, [CO2]out, [C2H4]out, [C2H6]out, [C3H6]out are the respective amounts of methane, carbon monoxide, carbon dioxide, ethylene, ethane, and propane and propylene in the exit gas. The current-programmed surface reaction (CPSR) experiment The C3H8 CPSR profiles were collected through a combination of a homemade reactor and mass spectrometer (Hidden HPR-20 EGA). The Co3O4-modified LSCF anode was exposed to a 5% C3H8 + 5% N2 + 90% Ar atmosphere at 600 °C, and the reaction was controlled using a workstation for current modulation. The appearance of C3H6 and H2 was analyzed using m/z = 42 and m/z = 2 to signal. DFT calculation details All calculations were performed based on DFT calculations, as implemented in the Vienna Ab-initio Simulation Package (VASP)51,52. The generalized gradient approximation (GGA) of the Perdew-Burke-Ernzerhof (PBE) functional53 and the projected augmented wave (PAW) method were employed to describe the exchange-correlation interactions54. The weak interaction was described by DFT + D3 method using empirical correction in Grimme’s scheme55. A cut-off energy of 450 eV for plain-wave basis sets was adopted, and the convergence threshold was 10-5 eV, and 0.01 eV/Å for energy and force, respectively. The vacuum space was set to be more than 15 Å, which was enough to avoid the interaction between periodical images. Spin-polarization was included for all surfaces. Four different sites (Supplementary Fig. 26), including La-top site, Sr-top site, Fe-top site and Co-top site, have been tested to screen out the thermodynamically most stable Co3O4-LSCF heterointerface model according to the following equation:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${E}_{{\rm{f}}}={E}_{{\rm{h}}{\rm{eterostructure}}}\,\mbox{--}{E}_{{LSCF}}\,\mbox{--}{E}_{{Co}3O4}$$\end{document}Ef=Eheterostructure–ELSCF–ECo3O4Where, Eheterostructure, ELSCF, and ECo3O4 refer to the calculated energy for Co3O4 cluster on LSCF surface, pure LSCF surface and isolated Co3O4 cluster, respectively. In order to correct the on-site Coulomb correlation of occupied Co and Fe 3 d orbitals, the Hubbard U corrections were adopted for both Co and Fe elements56. The Ueff(Ueff=U-J) values are 3.5 eV for Co57 and 4.0 eV for Fe58. The formation energy of oxygen vacancy on the surface was calculated using the following equation:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varDelta E}_{v}={E}_{{\rm{surface\; wit}}{\rm{h}}\,{\rm{Ov}}}+\mu O-{\rm{corrected}}-{E}_{{\rm{surface}}}$$\end{document}ΔEv=Esurface withOv+μO−corrected−EsurfaceWhere, Esurface with Ov and Esurface refer to total energy with oxygen vacancy and clean surface total energy, respectively. The chemical potential of oxygen atoms (μO-corrected = EO + b) has been corrected according to literature reports59, in which experimental thermodynamic references were used for the oxygen chemical potential, EO is single O energy calculated from O2 molecule, and b is energy correction for O. All barrier calculations were performed by using the climbing-image nudged elastic band method (NEB)60 and further improved by using the dimmer method61. The transition state structure is deemed converged when the forces acting on the atoms are all <0.05 eV/Å for the various degrees of freedom set in the calculation. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-3 Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1-3 Transparent Peer Review file Source data Source Data Source Data