Introduction

Transition metal oxides (TMOs) exhibit diverse emergent properties that have enabled their application in energy storage, catalysis, superconductivity, and electronic devices1,2,3,4,5,6. Among various tunable parameters, oxygen vacancies have emerged as a powerful handle to modulate the valence state, local structure, and associated functionalities of TMOs. In several systems, oxygen vacancy ordering leads to new structural phases, such as brownmillerite and infinite-layer configurations7,8,9. These vacancy-engineered phases, governed by local site-preferential vacancy formation, are promising candidates for oxygen-driven programmable matter. The formation and dynamics of oxygen vacancies under controlled environments—such as hydrogen, nitrogen, or argon—have been shown to critically influence the functional properties of perovskite oxides. Prior studies report reversible redox switching in SrCoOy8, oxygen deficiency-induced phase transitions in SrFeOy10, resistive switching behavior and vacancy ordering in brownmillerite films3, and modulation of vacancy behavior through multi-cation substitution1,11.

Recent efforts have focused on stabilizing oxygen-deficient phases in TMOs, particularly in cobalt- and iron-based perovskites, due to their rich redox chemistry and structural flexibility9,12. In cobalt oxides, phase transitions were typically limited between the perovskite and brownmillerite structures. However, a recent report on CaCoO2 demonstrated the formation of an infinitely layered structure stabilized under strong reducing conditions, where the small ionic radius of Ca2+ may introduce chemical pressure and facilitates planar Co−O bonding13. In contrast, SrCoO2 has been demonstrated to form a square spin-tube structure. This indicates that other factors, like chemical pressure from A-site cations, may hamper the formation of an infinitely layered structure in the Sr-based cobalitite4,14. In contrast, iron oxides can adopt a wider range of oxygen stoichiometries without considering the chemical pressure. The associated transitions among perovskite (AFeO3), brownmillerite (AFeO2.5), and infinite-layer (AFeO2) phases are often driven by apical oxygen removal and accompanied by notable structural contraction along the c-axis9,15. However, the redox reactions in mixed B-site systems such as SrFe1−xCoxO2.5 remain underexplored, particularly in terms of whether cooperative or competing interactions between Fe and Co can stabilize unique oxygen-deficient phases. In this study, we demonstrate that the reduction pathway of SrFe0.5Co0.5O2.5 (SFCO) deviates from typical infinite-layer transitions. Instead, it leads to the formation of a defective perovskite phase stabilized by selective Co reduction and oxygen vacancy formation at tetrahedral sites, while Fe remains chemically inert. This behavior is distinct from both SrCoO2.5, which decomposes under reducing conditions, and SrFeO2.5, which retains its structure without substantial reduction. We attribute this structural stability to Fe incorporation, which modifies the local coordination environment and suppresses Co-induced structural collapse16. This behavior highlights the site- and element-specific nature of redox control in multi-cation oxide systems and opens up a new phase space for programmable oxygen-deficient materials.

To systematically investigate the redox behavior of SFCO, we performed annealing experiments under reducing conditions using diluted hydrogen gas. A series of time- and temperature-dependent studies revealed the formation of a distinct structure in SFCO, which is different from the brownmillerite or infinite-layer phases. To assess the compositional dependence, we extended our study to brownmillerite SrFe1−xCoxO2.5 with x = 0, 0.2, and 1.0. Notably, the Fe-rich sample (x = 0) retained its brownmillerite structure, showing only a crystallographic axis rotation consistent with previous studies17. In contrast, the Co-rich sample (x = 1.0) underwent structural decomposition, accompanied by the formation of CoO impurity. The intermediate composition (x = 0.2) exhibited no sign of structural change. These observations highlight the unique redox stability of the x = 0.5 composition, where the incorporation of Fe plays a crucial role in maintaining structural coherence and preventing decomposition11,13,18. The cooperative yet asymmetric redox response of Fe and Co in the SFCO system stabilizes an oxygen-deficient defective perovskite phase that cannot be accessed in single-cation systems.

Results and discussion

Structural transformation via thermal reduction

To reduce the oxygen content in the as-grown brownmillerite (BM) SFCO thin films, we performed post-annealing in a 3% H2/Ar forming gas (FG), which facilitates the removal of lattice oxygen via reaction with hydrogen (Fig. 1a). Films were annealed at various temperatures for 1 hour to investigate temperature-dependent structural evolution. X-ray diffraction (XRD) patterns (Fig. 1b and Supplementary Fig. 1) show that the BM structure is preserved up to 300 °C, with progressive out-of-plane lattice expansion indicating gradual oxygen vacancy formation19. Upon annealing at 400 °C and 500 °C, we observed an abrupt shift in the 008 diffraction peak to lower angles and the disappearance of the 006 diffraction peak, suggesting the collapse of BM ordering and the emergence of a new reduced phase (Fig. 1c). Annealing at 600 °C led to complete loss of diffraction peaks, indicating a breakdown of epitaxial order.

Fig. 1: Structural evolution of SrFe0.5Co0.5O2.5 (SFCO) during thermal reduction.
Fig. 1: Structural evolution of SrFe0.5Co0.5O2.5 (SFCO) during thermal reduction.
Full size image

a Schematic of the reduction process in brownmillerite SFCO thin films on (001) SrTiO3 via annealing in 3% H2/Ar forming gas. Green, red, blue, and grey spheres represent Sr, Fe/Co, Ti, and O atoms, respectively. b X-ray diffraction patterns of as-grown and reduced SFCO films. c Out-of-plane lattice constants as a function of annealing temperature (1 h fixed). d Out-of-plane lattice constants as a function of annealing time at 400 °C. Blue and yellow background regions indicate the brownmillerite phase and the oxygen-deficient perovskite phase, respectively, emphasising the temperature range (300–400 °C) where the structural transition occurs.

To further investigate the time-dependent stability of the reduced phase, we fixed the annealing temperature at 400 °C and varied the annealing time. As shown in Fig. 1d and Supplementary Fig. 2, out-of-plane lattice expansion continued up to 5 hours, reaching ~4.020 Å, but remained unchanged thereafter up to 100 hours, indicating that the reduced phase is thermodynamically stabilized after extended reduction. Importantly, in-plane XRD measurements (Supplementary Fig. 3) confirm that the in-plane lattice parameters are locked by substrate-induced strain from SrTiO3 substrate throughout the reduction process. This anisotropic lattice evolution strongly contrasts with the c-axis contraction, which is typically found in an infinite-layer structure, suggesting that the reduced phase is structurally distinct. Further structural analyses using reciprocal space mapping (RSM, Supplementary Fig. 4) and HAADF-STEM (Supplementary Fig. 5) reveal out-of-plane peak shifts and a Sr−Sr lattice expansion in the vertical direction, while the in-plane periodicity is preserved. The extracted lattice constants from these measurements are summarized in Supplementary Table 1 and show excellent agreement between the XRD and STEM results. Taken together, these observations indicate that the reduced SFCO does not transform into a planar infinite-layer phase, but instead evolves into an oxygen-deficient defective perovskite4,14,16. The lack of long-range superstructure peaks suggests that oxygen vacancies are not ordered.

To understand the formation and stability of the reduced SFCO phase, we performed systematic annealing experiments on BM-SrFe1−xCoxO2.5 films with various compositions (x = 0, 0.2, 0.5, and 1.0), grown under identical conditions and annealed at 400 °C for 5 hours in FG. As summarized in Supplementary Fig. 6, the annealed BM-SrFeO2.5 (x = 0) retained its BM structure, with a minor crystallographic rotation similar to prior observations under vacuum annealing17. In contrast, BM-SrCoO2.5 (x = 1.0) decomposed into a CoO impurity phase, as evidenced by XRD20,21. For BM-SrFe0.8Co0.2O2.5 with x = 0.2, no structural transformation was observed, indicating limited reduction. Notably, only the x = 0.5 composition led to a distinct structural change upon reduction, highlighting the critical role of Fe/Co ratio in determining redox behavior and phase stability.

While Fe remains chemically stable under these reducing conditions, its presence appears to suppress structural collapse by resisting apical oxygen removal, enabling the formation of a stable oxygen-deficient phase. This observation is consistent with prior reports showing that Fe-based perovskites can maintain structural coherence under reducing conditions9, whereas Co-based systems often decompose4,21,22. Note that X-ray reflection revealed an increase in film thickness after reduction (23.40 nm for the as-grown SFCO and 24.30 nm for the reduced SFCO; Supplementary Fig. 7), consistent with the out-of-plane lattice expansion. When converted to unit cell layers, both films show comparable values (19.7 for the as-grown SFCO and 20.1 for the reduced SFCO), confirming that the reduction proceeds without significant structural degradation, while maintaining volumetric integrity. Atomic force microscopy images further revealed that the step-terrace morphology was preserved in both the as-grown and reduced films (Supplementary Fig. 8), indicating that the reduction process does not deteriorate surface integrity. The consistent root-mean-square roughness supports the conclusion that the reduced phase maintains high structural quality, without degradation of surface morphology. These results reinforce that the epitaxial strain is maintained in the reduced phase, while oxygen vacancies are accommodated within the lattice. To evaluate the thermal stability of the reduced phase, we conducted additional annealing at 500 °C. As shown in Supplementary Fig. 9, XRD and X-ray absorption spectroscopy (XAS) measurements after prolonged annealing revealed spectral broadening and structural degradation, indicating that the oxygen-deficient phase becomes unstable at the higher temperatures. This suggests an upper thermal limit for the reversible redox functionality in SFCO films.

Element-specific redox responses of Fe and Co

To determine whether the observed structural transformation arises from oxygen vacancy formation, we performed soft XAS at the Fe and Co L-edges. Figure 2a shows the Fe L-edge spectra for the as-grown and reduced SFCO films (annealed for 1, 5, and 100 hours). The spectral shape remains nearly unchanged, closely resembling that of LaFeO3 and SrFe0.8Co0.2O2.5, indicating that the Fe valence state is largely unaffected by the reduction process12,23. In contrast, the Co L-edge spectra exhibit significant changes upon annealing, as shown in Fig. 2b20. An absorption edge shift of 1.65 eV toward lower energies is observed for the Co L3-edges, consistent with a reduction in oxidation state24. To quantify the Co valence states, we performed multiplet simulations using the charge transfer multiplet for X-ray absorption spectroscopy (CTM4XAS)25 code, a ligand-field multiplet framework that enables detailed modeling of transition metal L-edge spectra. The experimental spectrum was fitted by a linear combination of Co4+ and Co2+ configurations, revealing a decrease in the average Co valence from Co2.91+ of the as-grown to Co2.00+ of the reduced (Fig. 2c and Supplementary Figs. 10). These results indicate that the structural transformation is driven by selective Co reduction, while Fe remains chemically inert.

Fig. 2: Element-specific electronic structure evolution upon reduction.
Fig. 2: Element-specific electronic structure evolution upon reduction.
Full size image

X-ray absorption spectra (XAS) of (a) Fe L-edge and (b) Co L-edge for as-grown and reduced SrFe0.5Co0.5O2.5 (SFCO) films. c Estimated Fe and Co valence states from multiplet fitting. Lists relative proportions of Fe4+, Fe3+, Fe2+ and Co4+, Co2+ in as-grown and 5 h reduced films. d O K-edge XAS spectra for SFCO films before and after reduction, showing changes in hybridization. e O K-edge XAS spectra for SrFe1−xCoxO2.5 with different x values.

To support this estimation, we performed multiplet simulations using CTM4XAS to fit the Fe and Co L-edge spectra. To refine the CTM4XAS simulations, we included multiple coordination environments for each oxidation state. For Fe, four components were considered: Fe4+ (octahedral), Fe3+ (octahedral and tetrahedral), and Fe2+ (tetrahedral), based on prior reports identifying Fe3+ species in both sites under strain or oxygen-deficient conditions26,27. Co spectra were simulated using Co4+ (octahedral) and Co2+ (tetrahedral), consistent with known valence distributions in cobaltite24,28. The Fe L-edge spectrum showed a dominant Fe3+ character (Fe4+: Fe3+: Fe2+ = 2.80: 93.74: 3.46), corresponding to an estimated oxygen stoichiometry of SrFe0.5Co0.5O2.497. The Co L-edge fitting yielded a Co4+: Co2+ ratio of 45.34: 54.66, corresponding to SrFe0.5Co0.5O2.453. The combined analysis supports an overall oxygen content of SrFe0.5Co0.5O2.475 ± 0.022 for the as-grown film, consistent with the BM phase. The simulated spectral weights and corresponding oxygen stoichiometry estimates for both Fe and Co L-edges are summarized in Fig. 2c. These results stand in contrast to previous reports on mixed B-site perovskites, such as SrFe1−xCuxO3−δ and SrFe1−xCoxO3−δ, where both B-site cations exhibited simultaneous redox behavior, yet no structurally distinct oxygen-deficient phase was stabilized under comparable reducing conditions29,30.

We further investigated the element-specific magnetic response using X-ray magnetic circular dichroism (XMCD) at the Fe and Co L-edges at 130 K. As shown in Supplementary Fig. 11, only the O2-annealed (O2, 400 °C, 5 h anneal) sample exhibited discernible XMCD signals, confirming the presence of finite magnetic moments in the oxidized phase (SrFe0.5Co0.5O2.5+δ). In contrast, no XMCD signal was detected in either the as-grown or reduced samples, suggesting the absence of net magnetic moments. These valence changes correlate with the anisotropic lattice evolution observed in RSM and STEM analyses, supporting the formation of an oxygen-deficient defective perovskite phase.

To further verify the element-specific valence changes associated with reduction, we examined the O K-edge XAS of SFCO films before and after annealing (Fig. 2d). In the reduced film, the feature within the 527.8–531.1 eV range, typically associated with Co 3 d − O 2p hybridization31, is significantly suppressed, consistent with the removal of oxygen atoms near Co sites. In contrast, a pronounced intensity enhancement is observed near 533 eV. This spectral feature, measured in total electron yield (TEY) mode, has previously been attributed to the σ* resonance of surface-stabilized O2− superoxide species in oxygen-deficient cobalt oxides32. The formation of these species is facilitated by oxygen vacancies and the reduced transition metal states, particularly near the film surface33. To further elucidate the element-specific hybridization, we measured the O K-edge spectra of BM-SrFe1−xCoxO2.5 with varying x values (x = 0.0, 0.2, 0.5, and 1.0) (Fig. 2e). The spectral region from 530 to 535 eV reflects hybridization between transition metal 3 d and O 2p orbitals, with contributions from both Fe 3 d − O 2p (near 529.5–530.5 eV) and Co 3 d − O 2p (within 527.8–531.1 eV) interactions. These compositional spectra serve as fingerprints to distinguish local oxygen environments in mixed B-site perovskites. Quantitative multiplet fitting, summarized in Fig. 2c, confirms that the selective reduction of Co is accompanied by oxygen deficiency, further validating the proposed redox mechanism. Additionally, electron energy loss spectroscopy (EELS) measurements (Supplementary Fig. 12) show a discernible chemical shift in the Co L3-edge toward lower energy after annealing, confirming the reduction of cobalt ions. This local electronic structure change is consistent with the observed O K-edge spectral evolution and supports our interpretation of Co-selective reduction. In contrast, no significant energy shift was observed for the Fe L-edge in the EELS spectra, further supporting the conclusion that iron remains largely chemically inert during the reduction process.

Role of the apical oxygen sites and Co-selective reduction pathway

To gain atomistic insight into the site preference of oxygen vacancies and the mechanism underlying Co-selective reduction, we performed density functional theory (DFT) calculations (Fig. 3). We evaluated the oxygen vacancy formation enthalpy for three distinct oxygen sites in the BM-SFCO structure34,35,36,37,38,39: equatorial oxygen within the octahedral layers, tetrahedral oxygen within the tetrahedral layers, and apical oxygen connecting tetrahedral and octahedral layers, as illustrated in Fig. 3a.

Fig. 3: DFT calculations of oxygen vacancy formation.
Fig. 3: DFT calculations of oxygen vacancy formation.
Full size image

a Atomic structure of brownmillerite SFCO indicates three inequivalent oxygen sites. In the schematic, green, blue, and red spheres represent Sr, Fe/Co, and O atoms, respectively. b Computed oxygen vacancy formation enthalpies at different oxygen sites and B-site configurations.

Since the experimental results indicate a selective reduction of Co ions while Fe remains largely unchanged, we considered different cationic environments surrounding each oxygen site, including Fe−Fe, Fe−Co, and Co−Co configurations. As shown in Fig. 3b, the formation enthalpy for oxygen vacancies is consistently lower when the oxygen is coordinated by Co-containing pairs. For instance, while Fe−O−Fe configurations at equatorial sites yield vacancy formation enthalpies close to 3 eV, Co−O−Co and Co−O−Fe configurations result in significantly lower values around 2 eV. This suggests that cobalt incorporation into the local environment promotes easier vacancy formation. Notably, even in Fe−O−Fe geometries, the enthalpy decreases as more Co atoms are introduced into the simulation cell, implying that the second-nearest neighbour also affects vacancy energetics.

When comparing different geometric sites for the same Co−Fe composition (x = 0.5), the tetrahedral oxygen sites exhibit the lowest average vacancy formation enthalpy (1.97 eV), followed by equatorial sites (2.25 eV), while apical sites show the highest average value (>3 eV). These trends confirm that oxygen vacancies preferentially form at tetrahedral sites in SFCO. This pathway contrasts sharply with the infinite-layer phase formation, which requires removal of apical oxygen atoms to induce square-planar coordination and c-axis contraction. Our DFT calculations demonstrate that, in general, the removal of oxygen ions from apical sites is relatively difficult when compared to equatorial sites. Combined with the random distribution of Co and Fe ions, it is expected that there will be randomized oxygen coordination in the reduced SFCO. In addition, XRD and STEM do not show superlattice peaks or electron diffraction patterns associated with long-range vacancy ordering but indicate the expansion of the lattice along the out-of-plane direction. This suggests that the oxygen vacancies, while site-selective, are likely disordered at the macroscopic scale, resulting in a defective perovskite phase without long-range structural modulation.

Therefore, we interpret the reduced phase as an oxygen vacancy-stabilized defective perovskite with local site selectivity, rather than a vacancy-ordered phase. These findings have important implications for the reduction mechanism in SFCO. Unlike infinite-layer phases, where they are developed via removal of apical oxygen and lattice contraction, our data suggest that the reduction proceeds mainly via tetrahedral oxygen removal, leading to a structurally stable, anisotropic, and defective perovskite phase. This interpretation aligns with our experimental observations of c-axis expansion and preserved in-plane coherence during reduction (Supplementary Fig. 3). Furthermore, our results support a pathway distinct from both SrFeOy and SrCoOy, where SrFeO2.5 typically maintains its BM structure under mild reducing conditions and SrCoO2.5 decomposes into CoO-like phases9,15,40,41. In contrast, our SFCO films exhibit a cooperative but asymmetric redox behavior, selective reduction of Co accompanied by retention of the Fe valence state, leading to the emergence of a stable oxygen-deficient phase (Supplementary Fig. 6). This site-specific redox pathway has not been observed in conventional single-cation perovskites and underscores the potential for achieving programmable oxygen-vacancy configurations in multi-cation systems9,15,40,42. Selective Co reduction in SFCO, particularly at tetrahedral sites, enables a structural and electronic landscape that is chemically and functionally distinct from previously reported perovskite and infinite-layer oxides.

Suppression of inter-band transition and bandgap tuning via selective reduction

To further investigate the electronic consequences of selective cobalt reduction and oxygen vacancy formation, we examined the optical properties of SrFe1−xCoxO2.5 films using spectroscopic ellipsometry. Figure 4a shows the optical conductivity spectra of BM-SrFe1−xCoxO2.5 films with varying Fe/Co compositions. For BM-SrCoO2.5 (x = 1.0), two distinct absorption peaks appear near 1.5 eV and 2.5 eV, corresponding to d − d and p − d transitions, respectively, consistent with prior reports43,44. In contrast, BM-SrFeO2.5 (x = 0.0) displays peaks near 2.7 eV and 4.0 eV, attributed to p − d transitions involving Fe−O hybridization10,45. The SrFe1−xCoxO2.5 films with x = 0.2 and 0.5 exhibit intermediate spectral features, reflecting a combination of Fe- and Co-related optical transitions. Notably, incorporation of Co leads to suppression of the 4 eV peak and enhanced spectral weight below 3 eV.

Fig. 4: Optical properties of brownmillerite and reduced SFCO films.
Fig. 4: Optical properties of brownmillerite and reduced SFCO films.
Full size image

a Optical conductivity spectra of SrFe1−xCoxO2.5 films (x = 0, 0.2, 0.5, and 1.0), showing composition-dependent absorption features. b Optical conductivity spectra of SrFe0.5Co0.5O2.5 before and after reduction, highlighting spectral weight suppression and bandgap widening from 2.47 eV to 3.04 eV. Inset: photographs of as-grown (dark and opaque) and reduced (lighter and partially transparent) films on SrTiO3, visually confirming enhanced transparency upon reduction.

To track the evolution of optical properties upon reduction, we measured the optical conductivity spectra of the x = 0.5 films annealed for various durations. A significant decrease in optical conductivity between 2 eV and 3.7 eV was observed after reduction, while the high-energy peak near 4 eV remained largely unchanged (Fig. 4b). This behavior suggests that the suppressed spectral weight is primarily associated with Co−O transitions, consistent with the selective Co reduction observed by XAS. Additionally, the optical bandgap increased from 2.47 eV (as-grown SFCO) to 3.04 eV (reduced SFCO), indicative of enhanced transparency8,46. This transition is visually confirmed in the inset of Fig. 4, where reduced films appear significantly lighter and more transparent than the as-grown samples. Such tunable optical responses, achieved via chemical reduction, highlight the potential of oxygen-deficient SFCO phases for optoelectronic/electrochromic applications requiring bandgap control and visible transparency.

Reversible redox transitions and formation of multiple distinct states in SFCO thin films

To explore the reversibility of phase transitions in SFCO, we performed simultaneous in situ experiments under controlled gas environments. These experiments revealed the presence of three structurally and electronically distinct states: the BM phase, the oxygen-deficient reduced phase, and the oxygen-rich perovskite-like phase. These states were identified through simultaneous monitoring of lattice parameters and resistance (Fig. 5) and further corroborated by the appearance or disappearance of BM superlattice reflections (Supplementary Fig. 13).

Fig. 5: In situ monitoring of phase reversibility.
Fig. 5: In situ monitoring of phase reversibility.
Full size image

a In situ x-ray diffraction and b resistance measurements during sequential 3% H2/Ar forming gas switching at 400 °C, showing structural reversibility and electronic modulation among three distinct phases.

Initially, the SFCO film was annealed at 400 °C under FG for ~2.5 hours, leading to the formation of the reduced phase. When the atmosphere was switched to high-purity N2 at the same temperature, no noticeable change in either lattice constant or electrical resistance was observed, indicating that the reduced phase is structurally and electronically stable under inert conditions. Subsequent annealing in pure O2 gas at 400 °C induced a rapid decrease in both lattice constant (Fig. 5a) and electrical resistance (Fig. 5b), suggesting re-oxidation toward an oxygen-rich SFCO phase. Notably, this transition occurred without abrupt changes in transport or diffraction signatures, indicating a continuous, diffusion-driven oxidation process. However, full oxidation to an ideal SrFe0.5Co0.5O3 phase was not achieved under these conditions, likely due to limited oxygen partial pressure. Previous studies have shown that such full oxidation requires high-pressure oxygen or ozone treatments40,41,47.

The resistance evolution mirrors the redox-induced structural changes: reduction led to charge localization and increased resistivity, while oxidation enhances carrier delocalization and reduces resistivity. This interplay reflects an intrinsic trade-off between optical transparency and electrical conductivity. While reduction enhances transparency and widens the bandgap, it simultaneously drives the material toward an insulating state. The ability to modulate both optical transparency and electrical conductivity via controlled redox tuning suggests that SFCO thin films could be promising candidates for multifunctional devices, including resistive switching memory and optoelectronic modulators48,49,50,51,52,53,54.

Upon switching the atmosphere to high-purity N2 at the same temperature, we observed a slight increase in the lattice constant and a corresponding increase in resistance. Expanded-range in situ XRD patterns revealed the reappearance of superlattice peaks such as 006BM and 0010BM, indicating partial reversion to the BM structure. This transformation is attributed to oxygen loss from the oxygen-rich perovskite-like state under inert conditions. To rule out the potential effect of residual oxygen during N2 annealing, we performed control experiments under high-purity N255. The absence of structural degradation or phase decomposition under N2 annealing is confirmed by XRD in Supplementary Fig. 14. As shown in Supplementary Fig. 15, no structural degradation was observed even after multiple redox cycles (up to six cycles of alternating FG and O2 annealing), confirming that the reduced phase is a structurally stable defective perovskite. The transient appearance of BM peaks upon switching from O2 to N2 (Supplementary Fig. 13) further supports an oxygen-mediated transformation pathway. After prolonged annealing in N2, we observed a saturation in both resistance and lattice contraction. These values, however, did not fully return to the initial BM levels observed in the reduced state12. Switching back to FG restored the reduced phase lattice constant and electrical resistance, confirming the repeatability of the redox transition. Additional redox cycling experiments (Supplementary Fig. 15) further demonstrated stable and reversible switching between the three phases without structural degradation, underscoring the robustness of the oxygen-deficient phase under cyclic operation. Note that it is important to clarify that while we identify three structurally distinct states, BM, reduced defective perovskite, and oxygen-rich SFCO, these classifications are based primarily on their transport and diffraction signatures.

In this study, we demonstrated the selective reduction of Co ions in epitaxial SFCO thin films while maintaining the chemical inertness of Fe ions. Through a combination of soft X-ray absorption spectroscopy, density functional theory calculations, and spectroscopic ellipsometry, we revealed that oxygen vacancies preferentially form near Co sites, leading to the formation of an oxygen-deficient defective perovskite phase. This phase exhibits a widened optical bandgap and enhanced transparency compared to the BM phase, without significant structural degradation. Comparative experiments on SrFe1−xCoxO2.5 with varying x compositions confirmed that the presence of Fe plays a crucial role in stabilizing the reduced phases. While Co-rich samples (x = 1.0) decomposed into impurity phases, Fe-rich and intermediate compositions (x = 0.0, 0.2, and 0.5) preserved structural coherence upon reduction, with x = 0.5 showing distinct defective perovskite formation. The incorporation of Fe modifies the local coordination environment, suppressing Co-induced collapse and promoting the stability of oxygen-deficient phases under reducing conditions. Our findings highlight the cooperative yet asymmetric redox behavior in multi-cation oxide systems, offering a new avenue for programmable oxygen-vacancy engineering and multifunctional oxide design. The selective control of cation-specific reduction not only enables tunable optical and electronic properties but also offers insights into the design of robust, multifunctional perovskite oxides for future device applications.

Methods

Material synthesis

Epitaxial thin films of SrFe1−xCoxO2.5, with compositions x = 0.0, 0.2, 0.5, and 1.0, were grown on single-crystalline SrTiO3 (001) substrates by pulsed laser deposition (PLD). The deposition was carried out at 600 °C under an oxygen partial pressure of 100 mTorr. A pulsed Nd:YAG laser (energy density: 0.75 J cm−2, repetition rate: 4 Hz) was used as the ablation source. All substrates were pre-treated via standard thermal annealing to ensure atomically flat step-and-terrace morphology (Supplementary Fig. 8)56. After deposition, the films were cooled to room temperature under the same oxygen partial pressure. Film thicknesses were measured using X-ray reflectivity, and structural characterization was conducted with a high-resolution X-ray diffractometer (D8 Discover, Bruker). Reciprocal space mapping confirmed that all films remained fully strained to the SrTiO3 substrate. Post-deposition annealing was performed in a tube furnace under a reducing atmosphere of FG. The system was initially purged with FG at a flow rate of 60 sccm for one hour, followed by a reduction to 20 sccm prior to temperature ramping. The samples were heated at a rate of 200 °C hr−1 to the target annealing temperature, followed by isothermal annealing for durations of up to 100 hours.

Materials characterizations

XAS measurements were carried out at the 2 A beamline of the Pohang Accelerator Laboratory using TEY mode at 300 K57. For accurate comparison, the Fe and Co L-edge spectra were normalized to the intensity of their respective L3-edges, while the O K-edge spectra were normalized to the background and the absorption edge near 565 eV. Multiplet fitting of the Fe and Co L-edge XAS spectra was performed using CTM4XAS, a ligand-field multiplet simulation code for TMOs, to estimate oxidation states from linear combinations of reference spectra. Optical conductivity spectra were acquired by spectroscopic ellipsometry in the photon energy range of 0.75–6.0 eV near the Brewster angle at room temperature58. To examine the feasibility of redox-driven structural reversibility in SrFe0.5Co0.5Ox films, in situ X-ray diffraction and resistance measurements were performed simultaneously using a dome stage (Nextron XRDMPS-CH). To evaluate the redox reversibility, in situ cycling experiments were conducted by alternately switching the gas environment between FG and O2 at 400 °C. Each gas cycle was maintained for several hours, and both structural and electrical responses were continuously monitored. These redox cycling experiments were repeated for up to six full cycles, confirming the phase reversibility and stability of the reduced and oxidized states (Fig. 5 and Supplementary Fig. 15). The dome chamber was interfaced with a high-resolution diffractometer and equipped with mass flow controllers to precisely regulate gas atmosphere and a PID controller for temperature stability. HAADF-STEM images and EDS element maps were obtained using a Cs-corrected STEM (Thermo Fisher Themis Z) operated at 300 kV. The convergence semi-angle and inner collection angle were set to 30 mrad and 50 mrad, respectively. Interatomic distances were quantified using custom python scripts. EEL spectral imaging was performed using a monochromated STEM (Thermo Fisher G2 Cube) operated at 80 kV. The convergence semi-angle was 13.5 mrad, and the full width at half maximum of the zero-loss peak was ~250 meV during spectral imaging.

Theoretical analysis

All DFT calculations of SrBOx (B = Co, Fe) were carried out using the plane-wave-based Vienna Ab initio Simulation Package VASP59,60 version 5.4.4, within the generalized gradient approximation (GGA) using the Perdew−Burke−Ernzerhof for solids (PBEsol) exchange–correlation functional61. The energy cutoff for the plane-wave basis set was 600 eV, employing projected augmented wave potentials62,63. A 6 × 6 × 2 k-point mesh was utilized for sampling the Brillouin zone for a 36-atom unit cell and scaled linearly with the supercell size. We used a 2 × 2 × 1 supercell for all calculations to accommodate Fe-Co solid solution on B-site and decrease periodic effects. The bulk geometry was optimised with a force convergence criterion of 1 meV/Å, and the individual components of the stress tensor were converged to ≤ 0.1 kB. Magnetism of B-site cations, Co, and Fe was treated with the PBEsol collinear spin density approximation in the GGA with an onsite Hubbard U (GGA + U) scheme64. An onsite Coulomb parameter U = 4 eV was applied for all B-site cations (Co and Fe) to account for the increased Coulomb repulsion between the semi-filled 3 d states65. The magnetic moments of B-site cations were initiated in the G–type antiferromagnetic state spins antiferromagnetically aligned within (001) planes between adjacent (001) planes66.