The Dynamics of Oxygen Ion Exchange in Epitaxial Strontium Cobaltite Bilayers

The exchange of ions across interfaces is key to the field of iontronics, where the properties of the device can be altered by the local ion concentration. This study investigates a complex oxide system where structural and electronic phase transitions can be driven by changes in the concentration of oxygen ions. In situ coherent X‐ray studies are conducted on epitaxial bilayers of insulating SrCoO2.5 and metallic SrCoO3 − δ. The diffusion of oxygen ions across the bilayer is studied with X‐ray photon correlation spectroscopy to capture the dynamical behavior of the interface in reducing and oxidizing environments. The behavior is strongly asymmetric, with much slower dynamics appearing in reducing versus oxidizing environments. According to the correlation functions determined from different points in reciprocal space, this study finds that the dynamics near the center of the SrCoO2.5 crystal are generally similar to those near the heterointerfaces. The results suggest that the interface is stable and reversible, making SrCoOx a model system for the study of iontronic behavior.


Introduction
The variable oxygen stoichiometry of transition metal oxides, which can be responsible for significant changes to their physical and electronic properties, makes these materials attractive for a host of applications in catalysis, energy storage, and microelectronics, [1][2][3][4][5][6][7] as well as the burdgeoning field of iontronics. [8,9] Of these oxides, strontium cobaltite (SrCoO x ), particularly in thin film form, has garnered interest because of two structurally and electrically distinct phases that can transition reversibly between each other via a topotactic pathway. [10,11] The orthorhombic brownmillerite, SrCoO 2.5 (BM-SCO), possesses alternating octahedral and tetrahedral layers and is electrically insulating, differentiating it from the cubic perovskite, SrCoO 3 (PV-SCO), with repeating octahedral layers and metallic conductivity. [12][13][14] The ordered vacancy channels of BM-SCO allow the phase to act like an "oxygen sponge" and take in additional oxygen as the valence state of Co changes. [11] The reversible transition between epitaxial thin-film BM-SCO and PV-SCO has been accomplished through the use of oxidizing environments and elevated temperatures (≈200-500°C) [11,15,16] both with and without an external bias, [17,18] as well as by solid-state and ionic liquid gating. [19][20][21][22] A schematic of an iontronic device utilizing the reversible transition (i.e., moving of the interface) between epitaxial BM-SCO and PV-SCO under oxidizing and reducing conditions is shown in Figure 1.
Interestingly, Mou et al. recently showed that the topotactic transition in SrCoO x could be highly beneficial for memristive applications since the resistive switching process is likely to be better controlled than in most filamentary systems, [18] and other reports have similarly demonstrated interest in this and related systems for memristive applications. [20,23,24] Here, the ordered vacancy channels allow a pre-defined path for oxygen migration without structural breakdown. From a thermodynamic perspective, the energy barrier for switching between the two phases is relatively small -smaller than those of similar systems such as strontium manganite (SrMnO 2.5 and SrMnO 3 − ) [11] -and the application of a bias to a BM-SCO/PV-SCO heterostructure could be used to change the overall resistance of a stack, paving the way for controllable and reversible synaptic memories. However, Cross-section schematic of proposed iontronic device bilayer (BM-SCO/PV-SCO) before and after the reduction/oxidation process. The reversible topotactic transition is depicted with the shifting of the interface between the bilayers during reduction and oxidation. Shown also are the crystal structures of BM-SCO and PV-SCO. The initial structure of PV-SCO is depicted to be more oxygen-deficient than after the reduction/oxidation process.
the reversible exchange of oxygen ions between the PV-SCO and BM-SCO phases requires an interface that remains stable and conducive to ion exchange throughout the lifetime of any device. Currently, little is known regarding the interface between the PV-SCO and BM-SCO cobaltite layers, not only in terms of structural stability but also with respect to oxygen ion dynamics (i.e., the hopping of oxygen ions back and forth across the interface).
Here, we investigate the BM-SCO/PV-SCO interface and oxygen motion across it using in situ synchrotron X-ray scattering. The dynamics of oxygen motion were studied by X-ray photon correlation spectroscopy (XPCS) under both reducing and oxidizing conditions -conditions under which the average motion of oxygen ions across the heterointerface is expected to be in opposite directions. We find that the behavior is highly asymmetrical: heating in reducing environments (flowing N 2 ) yields slow dynamics in a 15-nm-BM-SCO/15-nm-PV-SCO heterostructure, while the dynamics are more rapid under oxidizing conditions (flowing O 2 ), accompanied by distinct changes to the PV-SCO phase due to the accommodation of oxygen. From results measured at different regions in reciprocal space, we observe growth of the BM-SCO phase into the PV-SCO crystal during reduction and growth of PV-SCO into the BM-SCO crystal during oxidation. Our results provide important insights into the processes taking place during this reversible phase transition, key to understanding interfacial behavior in these topotactic systems.

Results
To focus on the stability and dynamic behavior of the BM-PV interface, we grew 15-nm-BM-SCO/15-nm-PV-SCO heterostructures on SrTiO 3 (001) (STO) substrates by pulsed laser deposition (PLD). Structural and dynamical characterization with Xrays were performed with a laboratory diffractometer (8.05 keV) and at the Advanced Photon Source (8-ID-E), with 7.37 keV X-rays. Additional details on synthesis and characterization can be found in the Materials and Methods and Supporting Information (SI) sections.
PV-SCO is a cubic perovskite (Pm3m) with a lattice parameter of a PV = 0.3835 nm at room temperature. [14] BM-SCO is orthorhombic (Ima2) with the lattice parameters a BM = 1.56376 nm, b BM = 0.55644 nm, and c BM = 0.5458 nm in the bulk. [13] The BM-SCO structure can also be described as pseudo-tetragonal, with a t = 0.3905 nm and c t /2 = 0.3936 nm. [10] While the ideal oxygen stoichiometries for the PV-and BM-SCO phases are x = 3 and x = 2.5, respectively, significant deviations to the oxygen concentrations of either phase are possible, which has been shown to affect the conductivity. [18] Pseudo-tetragonal BM-SCO has an ideal lattice match with the STO (001) substrate (a STO = 0.3905 nm) and grows epitaxially with its c-axis parallel to the STO [001], while cubic PV-SCO grows epitaxially on STO but is strained under biaxial tension.
Results from specular X-ray diffraction (XRD) for the asdeposted bilayer are shown in Figure 2a (red curve), plotted as a function of L -the out-of-plane direction -in reciprocal lattice units (rlu) based on the STO lattice parameter. [16] The PV-SCO layer of the heterostructure exhibits Bragg peaks that overlap with those from the STO substrate, but rocking curves clearly reveal the 001 of the PV-SCO ( Figure S1a, Supporting Information). The overlap is due to a reduced state of the PV-SCO (SrCoO 3 − ) caused by the subsequent, high temperature deposition of BM-SCO in reducing conditions; reduction is known to lead to a larger overall lattice parameter. [11,25] Evidence of the BM-SCO phase is provided by the 00 1 2 , 00 3 2 , and 00 5 2 reflections that arise from the alternating octahedral and tetrahedral layers.
X-ray reflectivity (XRR) indicates the total thickness of the as-deposited heterostructure is ≈ 30 nm, as determined by the thickness fringes ( Figure S2, Supporting Information). We note that we were not able to well differentiate the BM-SCO from the PV-SCO layers from XRR owing to their similar scattering length densities (SLDs), which provide a measure of the scattering power of the material dependent on the physical density and intrinsic scattering power. Furthermore, the SLD of the STO substrate is similar to those of the strontium cobalt oxides, leading to very shallow thickness fringes. The thickness of the BM-SCO layer was therefore determined by applying the Scherrer equation to the 00 1 2 reflection, resulting in a BM layer thickness of 14.0 nm ± 0.6 nm ( Figure S3, Supporting Information). The spacing of the thickness fringes adjacent to the reflection corroborates this finding, yielding a BM-SCO layer thickness of 15 ± 1 nm. Together with the thickness of the bilayer determined from XRR, we conclude that the as-deposited heterostructure roughly corresponds to the desired 15-nm-BM-SCO/15-nm-PV-SCO bilayer. Atomic force microscopy (AFM) reveals the as-deposited films to be smooth (< 1 nm root-mean-squared roughness), displaying a step-terrace structure (Figure 2b).
Understanding the behavior of the interface between conducting and insulating regions is essential to managing stochastic properties in iontronic devices. Nevertheless, the transition and migration of the interface is driven by defects, the behavior of which is fundamentally stochastic. We investigate the properties of the BM-PV interface in situ, as one phase grows at the expense of the other under reducing/oxidizing environments (Figure 1), here N 2 versus O 2 at elevated temperatures. With coherent Xrays, scattering from the sample appears as speckle on an area detector (Figure 3a), and the variation of the speckle pattern can be analyzed with time-dependent correlation functions. [26] In our measurement, the regions of interest include the 00 1 2 and the thickness fringe: these are displayed in the upper left of Figure 3b. [16,27] (We note that the shoulder of the 00 1 2 was chosen in order to avoid issues with the high intensity and sharpness of the Bragg peak on the detector).  regions of interest are similar, their origins in terms of scattering differ substantially. The correlation functions determined from these different regions allow one to distinguish between the dynamics of the BM-SCO crystal (00 1 2 ) versus the dynamics of the interfaces (thickness fringe).
The bilayer was heated under either flowing N 2 or O 2 for the conditions shown at the bottom of Figure 4a and allowed to reach the target temperature and stabilize for at least 10 min before X-ray measurements; the annealing times were chosen to provide statistically significant decorrelation times based on the scattered intensities. We employ the autocorrelation function g 2 (q, dt), which correlates the individual pixels with themselves as a function of time: here the scattering vector q is related to L by q = 2 L/a STO . The numerator corresponds to the correlation component, and the denominator normalizes the function such that g 2 decays to unity over long time scales. The g 2 (q, dt) measured while in N 2 or O 2 environments were fit to the equation where g 2 (∞) is the baseline (fixed at or close to 1), is the contrast (fixed to a few percent for a specific data set), t is time, and is the characteristic decorrelation time; [28] the fits, are shown in Figures S5 and S6 (Supporting Information). The decorrelation times determined during reduction and oxidation, are shown in Figure 4a for both the 00 1 2 (red squares) and thickness fringe (blue circles). The time dependence of g 2 for the 00 1 2 and thickness fringe are depicted in Figures 4b and  4c, respectively, for the conditions listed in their respective insets. Under N 2 flow at 250°C, the time extracted is over 10 000 s, corresponding to slow but not static dynamics, where the "static" dynamics were determined from a reference sample giving a decorrelation time of over 75 000 s ( Figure S7, Supporting Information). As heating continues under N 2 flow, does not decrease, as would be expected if the BM-SCO thickness is changing more quickly than at lower temperatures, but inspection of the 00 1 2 reflection measured between XPCS scans does reveal that there are changes to the BM-SCO thickness over time ( Figure S8, Supporting Information). After annealing at 250°C for 11 h, the thickness of the BM-SCO film increases from 15 nm to ≈ 19 ± 1 nm; the thickness eventually stabilizes at ≈17 ± 1 nm. While further reduction of the PV-SCO phase is possible in the nitrogen environment, this requires oxygen diffusion and residency within the BM-SCO overlayer. Hyperstoichiometric brownmillerite has been observed, [29] but the incorporation of a significant amount of oxygen is unlikely due to the observed stability of the BM-SCO crystal structure as evidenced by the half-order reflections in the blue profile of Figure 2a (post-treated bilayer). It has also been shown that the activation energy for oxygen ion diffusion along the [001] is high relative to diffusion within the plane of the film; [18] this, and the mobility necessary for oxygen vacancy ordering throughout the brownmillerite phase, implies that the conversion of PV-SCO to the BM-SCO phase can be slow at these temperatures, even in reducing environments.
Upon switching to the oxidizing environment via the flow of O 2 (right side of Figure 4a), the decorrelation times remain relatively long at 200°C and below (≳ 10, 000 s): begins to decrease at 250°C with regard to scattering at the thickness fringe. The shortest decorrelation times (≈ 2, 000 s) were observed for both the peak and thickness fringe after increasing the O 2 flow to 0.8 lpm. The BM-SCO film thickness returns to 15 nm after completion of O 2 exposure, decreasing in thickness by ≈ 13% from the start to the end of oxidation ( Figure S8, Supporting Information). The out-of-plane lattice parameter of PV-SCO decreases to the expected value of 0.380 nm for stoichiometric PV-SrCoO 3 on STO. [25,30] The reduced stability of BM-SCO in oxygen environments therefore facilitates the migration of oxygen ions through BM-SCO and into PV-SCO. Further growth of PV-SCO at the expense of BM-SCO can be expected at higher temperatures.
Aside from the 250°C measurement in oxygen at 0.4 lpm, the decorrelation times for the 00 1 2 and thickness fringe are similar, exhibiting the same temperature dependence whether in reduc-ing or oxidizing environments. Although an oversimplication, the 00 1 2 reflects the "bulk" of the BM-SCO layer while the thickness fringe yields information on the BM-SCO interfaces. The similar trends indicate that the dynamics of oxygen ion migration into, through, and out of the BM-SCO crystal are similar for both reduction and oxidation processes. The much faster dynamics at 250°C/0.4 lpm for the BM-SCO interfaces suggests that the passage of oxygen through the interfaces becomes significant at 250°C, potentially indicating greater reactivity at the BM surface. Jeen et al. [11] and Folkman et al. [31] have indeed found that the BM-SCO surface can be reactive at moderate temperatures.
Remaining at this temperature and increasing the oxygen flow rate by a factor of two shows a drop in decorrelation time for both the 00 1 2 and thickness fringe although this may be more related to the additional hour at 250°C than to the change in flow rate.

Discussion
The behavior observed in the bilayer sample during reduction and oxidation is summarized in Figure 5. Initially, the SrCoO 2.5 with ordered oxygen vacancies resides atop the SrCoO 3 − with disordered vacancies. During reduction in N 2 from 100-250°C (top of Figure 5), additional vacancies form in the PV-SCO layer (step 1), implying O 2 − migration from the PV-SCO to the BM-SCO crystal. The vacancies re-order in BM-SCO, shifting the boundary between the two layers, as shown in step 2. During oxidation in O 2 from 250-300°C (bottom of Figure 5), O 2 − ions enter the surface of the BM-SCO crystal (step 1). The ions subsequently diffuse from the BM to the PV layer (step 2), leading to an increase in thickness of the PV-SCO crystal.
The BM-SCO phase is more stable than PV-SCO when grown on STO (001) due to smaller biaxial strain. [11] The brownmillerite becomes even more stable under highly reducing conditions, in general agreement with N 2 results shown in Figure 4a, and the BM phase is expected to grow at the expense of PV-SCO, as was also observed. However, full breakdown of the PV-SCO phase is prevented by the inability of the ordered BM-SCO to accept significant amounts of excess oxygen. Furthermore, Zhang et al. have shown that growth of BM-SCO at the expense of PV-SCO occurs by a 3D mechanism, requiring a significant amount of time at moderate temperatures due to the need for oxygen vacancy ordering. [16] Under oxidizing conditions where the BM-SCO is less stable, we find that excess oxygen ions within the BM-SCO structure can migrate to the PV-SCO such that it eventually reaches stoichiometric SrCoO 3 , as indicated by the lattice parameter. Zhang et al. [16] showed that the growth of PV-SCO into BM-SCO occurs by a 2D mechanism and is comparatively fast. This asymmetry in the reduction and oxidation mechanisms appears in the dynamical properties shown in Figure 4a.
We note that in addition to effects caused by strain and the growth mechanism, the observed behavior is highly dependent on the configuration of the heterostructure (i.e., with BM-SCO exposed to the environment and PV-SCO adjacent to the STO (001) substrate). As noted, the migration of oxygen ions in BM-SCO (001) is anisotropic, with the diffusivity being much higher within the tetrahedral planes parallel to the heterointerface than along the out-of-plane axis. [11,29,32] Successful attempts to grow BM-SCO on different STO orientations, such as STO (110), have resulted in oxygen diffusion channels with a larger out-of-plane component, [18,31] which would likely lead to faster dynamics in a heterostructure. Furthermore, the rates of oxygen incorporation and evaporation at the surface of BM-SCO versus PV-SCO are likely to be very different in reducing and oxidizing environments, as noted by Zhang et al. [16] In this work, in situ XANES revealed that oxygen incorporation in the BM-SCO phase in oxidizing environments occurred readily as compared to oxygen evaporation in PV-SCO in reducing environments which required an incubation period. [16] This difference in rates implies that a PV-SCO/BM-SCO/STO heterostructure would likely exhibit a very different temperature dependence with regard to phase dynamics.

Conclusion
In situ coherent X-ray studies were performed on epitaxial bilayers of BM-SCO/PV-SCO (001) to investigate their potential in iontronics. Structural characterization with specular XRD and XRR reveals that the total thickness of the heterostructure does not change after exposure to reducing and oxidizing environments, but relative thickness changes to the different phases were found due to the flexibility of the buried PV-SCO layer in losing or accommodating oxygen. The dynamics of the heterostructure in reducing environments are slow ( ≈ 10 000 s and larger), but subsequent heating to 250°C in oxidizing environments leads to measurably faster dynamics ( ≈ 2000 s). The trends were comparable between the detector regions corresponding to the BM-SCO Bragg reflection and thickness fringe, indicating that the dynamics of oxygen ions within the 15-nm-thick BM-SCO crystal are generally similar to those near the interfaces at moderate temperatures. The results indicate that transition metal oxides with highly variable and reversible oxygen concentrations, [8] such as SrCoO x , [18] can serve as model systems for studying iontronic behavior and the development of future devices.

Experimental Section
Bilayer Preparation and Characterization: Pulsed laser deposition (PVD Products nanoPLD, KrF ( = 248 nm)) was used to deposit 15-nm-SrCoO 2.5 /15-nm-SrCoO 3 − bilayers on single-side-polished STO (001) crystals from a SrCoO 2.5 target (99.99% purity) with all materials from Adv. Mater. Interfaces 2023, 10, 2300127 MTI Corporation. The growth rate was ≈0.006 nm/pulse with a laser power of 225 mJ (≈100 mJ at target due to losses along the optical path), spot size of 2 mm 2 , repetition rate of 1 Hz, target rotation of 14 rpm, substrate rotation of 9 rpm, and target substrate-distance of 65 mm. Films were deposited with a background of 100 mTorr of O 2 and substrate temperature between 790-800°C. The heating rate for the first layer was 30°C min -1 . Upon deposition of ≈15 nm of SrCoO 2.5 , the film was cooled at 30°C min -1 to 50°C. The chamber was flooded with pure O 2 to a pressure of ≈330 Torr, and then heated back to 350°C at 10°Cmin -1 and held for 20 min in order to form the bottom SrCoO 3 − layer. High O 2 pressures, greater than 300 Torr, were needed to convert the as-deposited SrCoO x phase to the SrCoO 3 − phase, and depositing such films directly with high background O 2 was not practical because of the increased scatter of material ablated by the laser. [10,25] The O 2 pressure was reduced to 1 Torr for continued heating at 30°Cmin -1 to the deposition temperature for the films, 790-800°C. The O 2 pressure was returned to 100 mTorr for deposition of the next ≈15 nm of SrCoO 2.5 , which formed the top layer. The bilayer film was cooled rapidly to room temperature at 50°Cmin -1 to 30°C before removing the sample from the chamber. Specular XRD, XRR, and rocking curves of SrCoO 2.5 /SrCoO 3 − bilayers were measured using a Rigaku Smartlab system (Cu K-, = 0.15406 nm). XRR data were analyzed using Motofit software. [33] AFM topography scans were taken on a Bruker ICON system.
X-ray Photon Correlation Spectroscopy: X-ray photon correlation spectroscopy (XPCS) was performed at beamline 8-ID-E of the Advanced Photon Source, Argonne National Laboratory. The X-ray beam was generated by tandem 33 mm period 2.4 m-long undulators. The white beam slit defined the horizontal source size with a horizontal opening of ≈40 μm. [34] The X-rays were then diffracted by a silicon (220) single-bounce monochromator to select a longitudinally coherent beam with a photon energy of 7.35 keV and a bandpass of ΔE/E = 4 × 10 −5 . The X-ray beam passed through a 150 μm × 150 μm slit to select a partially coherent X-ray beam and then focused down to a 5 μm spot on the sample in both horizontal and vertical direction. The X-ray flux on the sample was 5.5 × 10 9 photon per second. The coherent X-ray scattering intensity was collected using a XSpectrum Lambda 250k, a 512 × 512 pixel array, single-photon-counting detector with a pixel size of 55 μm. The detector was placed at a distance of 1.1 m from the sample. Details on the optics and X-ray coherence could be found in previous studies. [34] The high-temperature stage used during X-ray studies was an Anton Paar DHS 1100 with a graphite dome.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.