The flowchart illustrating schematically the preparation of atomically flat STF50 thin film is shown in Fig. 1a. Corresponding AFM micrographs for the annealed as-supplied STO substrate, atomically flat STO substrate, and atomically flat STF50 thin film, are reported in Fig. 1b-d, respectively. Compared to the as-supplied STO substrate (Fig. 1b), BOE treatment and annealing resulted in the flat step structure, as shown in Fig. 1c. The width of the steps was at the level of several hundred nanometers, but it varied slightly depending on the cutting angle of the supplied substrate. The surface topography of STF50 thin film (Fig. 1d) also exhibits a stepped shape matching that of the substrate indicating that it underwent epitaxial growth. The HR-XRD patterns in Fig. 1e and magnified (002) peak in Fig. 1f also support that the thin film was epitaxially grown with the lattice constant of 0.3956 nm, which is consistent with the reference data (0.3944 nm)39. The AFM line scan along the solid blue line in Fig. 1d is shown in Fig. 1g with the step height of 0.39 nm, which matches the unit cell parameter of STF50. The consistent step height uniformity observed may suggest that the STF50 thin films are atomically flat and possess a single termination. Further discussion on this aspect will follow.
A detailed investigation of the structure and composition of the STF50 thin film was conducted using STEM-EDS, APT, XPS, and LEIS (Fig. 2). In Fig. 2a, the cross-sectional STEM image clearly shows the epitaxial growth of STF50 thin film with a thickness of approximately 10 nm. A magnified image of the gas-solid interface is shown in Fig. 2b, with the brighter and bigger Sr in contrast to the darker and smaller B-site cations. From the STEM image, it can be distinguished that the thin film was terminated with a SrO layer, which was employed in the DFT calculations. Moreover, the STEM image and corresponding EDS maps of the film-substrate interface showed the perfect epitaxial growth of STF50 thin film on the substrate (Fig. 2c and Fig. 2d).
Next, we utilized APT, which is the state-of-the-art material characterization technique that enables 3D atomic imaging with sub-nanometer resolution and ppm-level chemical sensitivity58, for precise investigation of the chemical distribution of the STF50 thin film. The three-dimensional atom map (Fig. 2e) and cross-section map (Fig. 2f) demonstrate the absence of any elemental segregation of Sr, Ti, and Fe throughout the thin film, also clearly shown in the 2D concentration plots (Fig. 2g). The existence of a Ni protection layer at the top confirms that the APT analysis was conducted from the topmost layer of the STF50 thin film (Fig. S4a). The atomic concentration profile (Fig. 2g) along the vertical direction of the atom map further supports the uniform distribution of each element in the as-grown STF50 film. In addition, the Sr/(Ti + Fe) ratio was nearly 1 from the entire film (Fig. S4c), and the bulk composition of the film was SrTi0.51Fe0.49O3−δ, consistent with the nominal composition and the composition measured through STEM-EDS (Fig. S5).
The Sr 3d spectrum shown in Fig. 2h was deconvoluted to a single 3d doublet with an intensity ratio of 2:3, indicating the absence of non-lattice Sr species and perfect surface homogeneity. In addition, LEIS data acquired from the as-supplied STO substrate and the STF50 thin film (Fig. 2i) showed that the as-supplied STO displayed a combination of Sr and Ti signals, indicating a randomly terminated surface, whereas the STF50 thin film displayed a single Sr peak, indicating singly-terminated SrO layers. It is important to emphasize the single termination of the STF50 thin film, as the termination of perovskite oxides can greatly affect surface activity29.
The STF50 thin films were annealed at 300°C for 0-600 minutes and then air-quenched. AFM observations (Fig. 3a) showed that the surface topography remained unchanged during 30 minutes of annealing. However, after 60 minutes, significant morphological changes were observed, including the formation of second-phase particles on the step edge or terrace. Here, the first question from our introduction section (i.e., Can we measure the reactivity of an ideal perovskite oxide surface in the absence of initial Sr segregation?) can be answered using the first data point (\(t\) = 30 min) in Fig. 3b. The surface oxygen exchange coefficient, \({k}_{s}\), of an atomically flat and perfectly homogeneous STF50 (001) surface was measured to be approximately 7.4 × 10− 9 cm·sec− 1 at 300°C without any effect from surface Sr segregation. The segregation occurred noticeably after 60 minutes. This coefficient value is nearly seven times higher than the previously reported value of 1.1 × 10− 9 cm·sec− 1 of an epitaxially grown STF50 (001) surface39, even though the two studies utilized the same PLD facility. This deviation is attributed to the effect of surface Sr segregation that occurred in the sample of the previous study, which was conducted without controlling the surface termination and surface deterioration during deposition. This again emphasizes the importance of precise preparation of the perovskite oxide surface for the reproducible measurement of intrinsic properties.
The clean perovskite oxide surface without any second-phase particles is advantageous for observing the changes of \({k}_{s}\) according to the Sr segregation due to its atomic-scale flatness, and absence of grain boundaries and as-segregated particles. The observation of nanoscale chemical changes on this surface allows us to quantify Sr segregation and the degree of inhomogeneity during the annealing. On this well-controlled surface, we confirmed that there were no changes in crystal structure or in the termination layer during the annealing. The annealing only changed the exposed reactive surface area, which is not blocked by inert particles, and the composition of this exposed surface. Thus, these factors can be compared with surface activity to determine the representative factors that define how much Sr segregation has occurred and how much of this surface deviated from ideality.
As indicated by AFM images after 120 and 600 minutes of annealing, some particles became noticeably larger than others (see Fig. 3a). The particles continued to emerge and agglomerate, increasing the average particle size while causing fluctuations in particle density or coverage. The exposed surface area was obtained using XEI software and further confirmed with SEM images to compensate for the small scan area of AFM (4 µm2) (see Fig. S6). The changes in the percentage of exposed surface (solid blue line in Fig. 3b) shows notable divergence from the measured surface oxygen exchange coefficient over time. This implies a minor correlation between reaction site blockage and surface activity.
Finally, to verify that the degree of Sr segregation is highly correlated to the compositional change of the exposed surface area, and to answer the remaining question concerning the effect of Sr segregation on the surface activity degradation, a deeper investigation of the exposed surface was conducted. As depicted in Fig. 3c, the formation of second-phase particles on the surface is expected to deplete the Sr content at the exposed surface, and this was confirmed by the APT experiments (Fig. 3d-g). The 3-dimensional reconstruction (Fig. 3d) of the 180 min-annealed STF50 thin film and the cross-sectional map (Fig. 3e) show Sr segregation at the topmost surface, which was confirmed by the presence of a Cr protection layer (Fig. S4b).This distinctive Sr segregation was further clarified by the 2D concentration plots (Fig. 3f). Moreover, the atomic concentration profile (Fig. 3g) demonstrated the length scale of the Sr-segregated region is about 3 nm with a much higher Sr/(Ti + Fe) ratio than the below region (Fig. S4a). In contrast, below this region lies an area with a Sr-deficient composition of Sr0.95Ti0.5Fe0.5O3−δ, as expected from the schematic figure (Fig. 3c).
The SrO-free surface on the 180 min-annealed STF50 thin film was also analyzed by APT to clarify the Sr-deficiency of all-exposed oxide surfaces as well as the region below the SrO particles. The 3D atom map and corresponding 2D concentration plots showed the absence of any elemental segregation at the near-surface region (Fig. S7a). Moreover, the atomic concentration profile and Sr/(Ti + Fe) ratio profile revealed the uniform distribution of Sr, Ti, and Fe, the same as the as-grown STF50 film (Fig. S7b and c). However, their one distinctive difference was the Sr bulk composition; while the as-grown STF50 film showed the stoichiometric composition (SrTi0.51Fe0.49O3−δ), the SrO-free surface of 180 min-annealed film represented the Sr-deficient composition, Sr0.95Ti0.52Fe0.48O3−δ. These results imply that the formation of SrO particles decreases the Sr contents from the entire oxide surface, providing the fidelity of the schematic shown in Fig. 3c.
To investigate why surface activity degradation occurs due to the Sr segregation, a parallel experiment was conducted using the same platform with Sr0.95Ti0.5Fe0.5O3−δ (Sr95TF50). The initial surface oxygen exchange coefficient of Sr95TF50 was approximately 4 × 10− 9 cm·sec− 1 at 300°C, which is lower than that of the pristine STF50 thin film but comparable to that of Sr-segregated-STF50 after 150 minutes (see Fig. S8a), proving that the surface activity degradation originated from the resulting Sr-deficiency of the exposed oxide surface due to the surface Sr segregation. Additionally, the surface activity of Sr95TF50 (Fig. S8b) underwent almost no degradation for 600 minutes and showed no changes in surface morphology (Fig. S 8c). This corresponds to the postponed change in surface activity of STF50 after a degree of Sr segregation has occurred (solid red line after 150 minutes in Fig. 3b.)
DFT calculations were performed to confirm that Sr-deficiency can cause surface activity deterioration. The STF50 bulk was predicted to be in its cubic phase with a lattice constant of 3.93 Å, consistent with XRD results. The calculated \({E}_{Sr, vac}\) for Sr1 and Sr2 are 9.01 and 9.20 eV, respectively, implying that removing the Sr1 surrounded by 8 Ti atoms is more favorable. The lower energy may be a result of the stronger coulombic interaction between \({V}_{Sr}{\prime }{\prime }\) at the Sr1 site with more positively charged Ti4+ cation rather than Fe2+/3+. Hence, the Sr95TF50 was simulated using the 3×3×3 supercell with one Sr1 removed. The electronic densities of states of STF50 and Sr95TF50 were computed (Fig. 4a) with the electronic band structure shown in Fig. 4b. The Sr vacancy made Sr95TF50 more p-type than STF50 with a large band gap, causing a lower electron density. The lowered electron density and the resulting difficulty in charge transfer (\(2{e}^{{\prime }}+{V}_{O}^{\bullet \bullet }+{O}_{ads}={O}_{O}^{\times }\)) yields the lower surface activity of Sr95TF50, as previously highlighted in our work15 and the literature44,59–61. In addition, creating one Sr vacancy moves the O 2p band center from − 1.27 eV to -1.39 eV relative to the Fermi level, for which surface oxygen exchange and O 2p band center distance from the Fermi level have been positively correlated62. In other words, as a smaller O 2p band center implies lower surface activity, Sr95TF50 is expected to show decreased oxygen reduction reaction catalytic activity relative to STF50.
To estimate the Sr segregation in both STF50 and Sr95TF50, we further calculated \({E}_{\text{Schottky}}\) and \({E}_{\text{Sr, vac}}\) with corresponding values shown in Table 1. We observe that \({E}_{\text{Schottky}}\) is larger for Sr95TF50 than STF50, implying that forming a Schottky defect pair of \({\text{V}}_{\text{Sr}}^{{\prime }{\prime }}-{\text{V}}_{\text{O}}^{}\) in Sr95TF50 is more difficult. Considering that the Sr segregation takes place by creating a Schottky defect in the bulk of the slab, a higher value of \({E}_{\text{Schottky}}\) is associated with Sr segregation hindrance. This may be attributed to the negative charge of \({\text{V}}_{\text{Sr}}^{{\prime }{\prime }}\) that hinders the further formation of the \({\text{V}}_{\text{Sr}}^{{\prime }{\prime }}-{\text{V}}_{\text{O}}^{}\) defect pair. Moreover, we observe that removing one Sr from the slab induces the locally structural distortion by reducing the Ti-O-Fe angle from ~ 178° (layer 5 in Fig. S2c counting from the bottom) to ~ 169° (layer 5 in Fig. S3c counting from the bottom). This distortion may lift the total energy of the slab in Fig. S3c, thus increasing \({E}_{\text{SrO, form}}\), the energy required to form one monolayer of SrO on the top surface of the SrO-terminated and (001)-oriented slabs. Consistently, it was found that \({E}_{\text{SrO, form}}\) is larger for Sr95TF50 than STF50. This means that the formation of surface inactive SrO is less favorable in Sr95TF50 relative to STF50. The conclusions drawn here are consistent with the experimental observations that Sr-deficient hinders surface Sr segregation, thus improving the surface activity.
Table 1
Calculated formation energies for a pair of Schottky defect \({\varvec{V}}_{\text{Sr}}^{\varvec{{\prime }}\varvec{{\prime }}}-{\varvec{V}}_{\text{O}}^{\text{∙∙}}\) and a monolayer of SrO at the top surface, in eV.
| \({E}_{\text{Schottky}}\) | \({E}_{\text{SrO, form}}\) |
STF50 | 3.58 | 0.91 |
Sr0.95TF50 | 3.85 | 0.98 |
It can be inferred that the deterioration of oxygen activation due to Sr segregation on the STF50 surface can be highly correlated to the compositional change, especially the Sr deficiency of the exposed surface. This plays a major role in the degradation of surface activity. Several studies have observed a similar surface Sr deficiency on the surface of annealed perovskite oxides33,37, but the previous studies could not offer fully confident conclusions because they measured the average composition of the near-surface (~ 10 nm), without accounting for nm-scale inhomogeneities. In contrast, this study has been able to provide a clear and comprehensive answer to the degradation mechanism caused by Sr segregation by preparing atomic-level well-defined surfaces and integrating state-of-the-art surface composition analysis technique with computational results.
Sr-deficient compositions have been widely proposed as a strategy to enhance surface activity and stability for high-temperature electrocatalysts.48–51 This consensus, however, appears to conflict with our observations, potentially stemming from degradation during high-temperature calcination or prior to activity characterization. Notably, we observed a substantial degradation in surface oxygen exchange rate (approximately 7-fold in our case) even at 300°C, well below conventional characterization conditions (> 500°C).
In this case study, we focused on the representative target material STF50, but further application to other materials is possible since various single-crystal substrates with lattice constants ranging from ~ 3.7 Å (YAlO3) to ~ 4.2 Å (MgO) can be termination-controlled52, and thereby used to grow atomically flat thin films. This work can serve as a starting point for designing high-performance electrochemical devices using conductive perovskite oxide electrodes.