Fermi Level Equilibration and Charge Transfer at the Exsolved Metal-Oxide Interface

S upported metal nanoparticles are essential for industrial catalysis. 1,2 A recent advance in their fabrication is exsolution, [3][4][5] a method that uses redox-driven metal precipitation to spontaneously disperse metal nanocatalysts across oxide surfaces [6][7][8] (Figure 1a). Exsolved nanoparticles exhibit a highly oriented 9 and "socketed" 10,11 metal-oxide interface, providing remarkable structural stability in (electro)catalysis. 12 While these structural benefits are well recognized, the electronic interactions at the exsolved interface, specifically electronic metal-support interactions (EMSI), 13,14 remain underexplored. EMSI involves orbital rehybridization and charge transfer at the metal-support interface, 15,16 and proper tuning of these interactions has been shown to lead to orders of magnitude enhancement 13 in catalytic performance (e.g., C 1 chemistry, [17][18][19] oxygen and hydrogen electrocatalysis). 20,21 Despite its importance, the nature and even the existence of EMSI at exsolved interfaces remain unclear, [22][23][24] hindering the optimization of exsolved catalysts.

The key challenge is that the lattice redox processes 25,26 occurring during exsolution complicate the identification of EMSI. In supported metal nanocatalysts, changes in oxide valence states before and after metal deposition are commonly used to infer charge transfer at the interface. 27,28 However, this method is insufficient for exsolved systems. During exsolution, host oxides continuously release oxygen both before and during metal precipitation, 29 making it challenging to differentiate whether the observed valence changes are due to lattice redox effects or charge transfer at the exsolved interface.

In this study, we present in situ Fermi level (E F ) monitoring as an effective method to probe EMSI at the redox-active exsolved interface. EMSI can be detected because charge transfer across the metal-oxide interface causes E F to equilibrate between the two phases. [30][31][32] In contrast, lattice redox processes change the E F position as oxides develop nonstoichiometry. [33][34][35] Therefore, by observing whether and how E F equilibrates during exsolution, we gain direct insights into the presence and nature of EMSI at the exsolved metaloxide interface.

Ambient pressure X-ray photoelectron spectroscopy (APXPS) 36 is ideal for in situ E F monitoring due to its compatibility with high-temperature and gaseous exsolution environments. 8,29,[37][38][39] In APXPS, E F variations are detected through electrostatic shifts in the photoemission spectra. [39][40][41][42] To demonstrate this approach, we selected epitaxial (001)oriented SrTi 0.65 Fe 0.35 O 3 (STF) thin films as the model system (Figure S1), due to their relevance as electrode materials in solid oxide cells 43,44 and their well-established point defect chemistry. 45,46 By continuously tracking the evolution of E F during metallic iron (Fe 0 ) precipitation on STF, we provide direct evidence of E F equilibration, confirming the presence of EMSI at the exsolved interface. First-principles calculations and thermodynamic defect modeling further elucidate point defect reactions and charge transfer during Fe 0 exsolution.

Representative APXPS spectra are shown in Figure 1b. With prolonged H 2 reduction, metallic Fe 0 species emerged in the Fe 2p spectra, indicating metal exsolution on STF. 29,37 The formation of Fe 0 nanoparticles was further validated by a comprehensive suite of ex situ characterizations (Figures 1c, S2, and S3). Meanwhile, spectra of nonexsolved elements (Ti 2p, O 1s, and Sr 3d) exhibited a rigid shift to higher binding energies throughout the reduction process (Figures 1b and S4). This binding energy shift (ΔBE) in XPS could result from several factors, such as Fermi level variation 39,40 (ΔE F ), changes in valence 47 and chemical 48 states, changes in Madelung potential, 49,50 surface band bending, 51,52 or a combination of these factors. 53 We now discuss how the observed ΔBE in our experiment is primarily driven by ΔE F .

First, the Ti 2p, O 1s, and Sr 3d spectra, normalized for peak intensity, remain nearly unchanged after adjusting for the binding energy shift (Figures 1d and S4), indicating that the ΔBE is not driven by changes in the valence 47 or chemical state 48 of these nonexsolved elements. Second, the consistent ΔBE for both anion (O) and cation (Sr, Ti) spectra suggest a negligible contribution from Madelung potential variations, which would otherwise induce opposite shift directions. 49,50 Finally, the stable core-level peak widths during the binding energy shift (Figures 1d and S4) suggest minimal contribution from band bending, which is typically accompanied by peak broadening due to the built-in potential. [54][55][56] Therefore, as illustrated in Figure 1e, we conclude that the ΔBE in nonexsolvable elements mostly reflects ΔE F within STF.

We then quantified the time-dependent surface chemistry and E F evolution in STF during H 2 reduction and Fe 0 exsolution. As shown in Figure 2a, the sample was first preconditioned in an oxidizing environment (0.02 Torr O 2 , 300 °C) and then switched to a reductive condition (0.1 Torr H 2 , 400 °C). After switching to H 2 atmosphere, the Sr 3dnormalized O 1s intensity gradually decreased (Figure 2b, top) while the normalized Ti 3d remained stable (Figure S5). This decrease in normalized O 1s intensity reflects the formation of oxygen vacancies 57 in the STF lattice. Concurrent with lattice reduction, the binding energies of Sr 3d, Ti 2p, and O 1s spectra increased by approximately 1 eV before stabilizing. ΔE F was then calculated as the average shift in these spectra (Figure 2b, middle). Finally, by deconvoluting the Fe 2p 3/2 spectra into lattice cation and metallic iron components (Figure S6), we quantified the time-dependent Fe 0 concentration on STF (Figure 2b, bottom). Plotting the surface Fe 0 concentration and ΔE F against the normalized O 1s intensity reveals a twostage evolution of E F (Figure 2c). Initially, E F rises in singlephase STF during oxygen release, prior to Fe 0 phase formation. Second, after Fe 0 precipitation, E F stabilizes in the dual-phase Fe/STF system and becomes insensitive to further oxygen release. In the following, we employ thermodynamic analysis and first-principles calculations to elucidate the origin of the observed two-stage E F evolution in STF (Figure 2d) during Fe 0 exsolution.

We first use analytical defect modeling to show that the initial E F upshift, transitioning from an oxidized state to a highly reduced state prior to Fe 0 precipitation, can be quantitatively explained by electron doping. 46 The defect model used in this work is based on Rothschild et al. 46 (see Methods), which accurately reproduces the experimentally obtained oxygen nonstoichiometry (δ) in bulk STF 45 across a wide range of oxygen partial pressures (pO 2 ) and temperatures (Figure 3a). By linear extrapolation (Figure S7), we predicted the defect chemistry in STF at the experimental condition of 400 °C, with the calculated electron concentration in the conduction band (n) shown in Figure 3b. The power-law dependencies in the three distinct regions can be explained by Brouwer approximations, 58 with key defect reactions 59 being anion Frenkel disorder at high pO 2 , electron-hole pair generation at intermediate pO 2 , and oxygen vacancy formation at low pO 2 . For a nondegenerate semiconductor like STF, 46 its E F position relative to the conduction band edge (E C ) is given by 33

, where N C is the effective density of states in the conduction band, T represents temperature, and k B being the Boltzmann constant. Assuming E C and N C are constant, we can calculate ΔE F = k B T ln(n) using the calculated n-pO 2 relationship in Figure 3b. By analyzing the Fe 2p spectra prior to the exsolution onset (Figure S8), we determined that Fe 0 precipitation occurs when surface oxygen nonstoichiometry δ exceeds ∼0.1 (shaded region in Figure 3a), which corresponds to an equilibrium pO 2 below ∼10 -43 Torr at 400 °C (shaded region in Figure 3b). Thus, the first-stage E F upshift in our experiment can be estimated as the difference in equilibrium E F between pO 2 ∼ 10 -43 and 0.02 Torr (i.e., the oxidation environment in APXPS experiment) in Figure 3b, which is around 1 eV. The excellent agreement between analytical defect modeling and experimental observations (Figure 2c) indicates that the initial E F upshift is due to reduction-induced electron doping in single-phase STF.

Following Fe 0 precipitation, while STF continues to release oxygen (Figure 2), two key changes occur. First, Fe vacancies begin to form in the lattice as Fe cations are expelled. 37,60 Second, the surface transforms into a dual-phase system due to the formation of Fe 0 nanoparticles. Since our experiment focuses on the early stage of exsolution, the low concentration of Fe vacancies (Figure S2) has a negligible impact on E F , as confirmed by our previous study 38 (Figure S9). We thus attribute the experimentally observed E F stabilization to the E F equilibration and charge transfer 30,31 between the exsolved metal and the host oxide. This is further supported by density functional theory (DFT) calculations presented below.

To accurately model the exsolved STF surface, we first examine its atomic structures using high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM). As shown in Figure 4a, exsolved Fe nanoparticles epitaxially grow on STF with the orientation relationship (001) STF || (001) Fe and [100] STF || [110] Fe . We then used these epitaxial structures in our DFT calculations (Figure 4b). To reduce computational complexity, we used a composition of SrTi 0.75 Fe 0.25 O 2.875-δ (see Methods). This variation is expected to have minimal impact on the DFT results, as the SrTi 1-x Fe x O 3-x/2-δ solid solution family exhibits similar defect chemistry. 46 We will also refer to the computational composition as STF for simplicity in the following discussion.

The DFT calculated E F in both single-phase STF and dualphase Fe/STF is summarized in Figure 4c. In single-phase STF, the calculated E F increases with δ, aligning with our analytical defect model (Figure 3). In contrast, in dual-phase Fe/STF, the calculated E F closely matches that of single-phase Fe and remains insensitive to δ. This stabilization results from E F equilibration and charge transfer between STF and the exsolved Fe (Figure S10). The direction of charge transfer in dual-phase Fe/STF at a given stoichiometry can be understood by comparing the E F values of single-phase Fe and STF. At δ = 0.125, where Fe has a higher E F , electrons transfer from Fe to STF (Figure 4d). Accordingly, Bader charge analysis shows that the Fe layer in the dual-phase system is positively charged, while the valence state of Fe in the STF layer decreases compared to their single-phase counterparts. Conversely, at δ = 0.25, since STF has a higher E F , electrons transfer from STF to Fe (Figure 4e). Therefore, our DFT calculations indicate that charge transfer at the exsolved metal-oxide interface accounts for the observed E F stabilization in Figure 2.

Overall, we demonstrate that in situ E F monitoring via APXPS effectively probes EMSI at the exsolution interface. We observed that, once Fe 0 precipitation begins, the E F level remains stable even as STF continuously forms O and Fe vacancies within it. This is in contrast to the initial stage before Fe 0 precipitation, where E F steadily increases as STF reduces. This stabilization of E F during the exsolution process indicates dynamic charge transfer between the exsolved metal nanoparticles and the host oxide. As host oxides continuously develop nonstoichiometry and lattice defects throughout exsolution, we speculate that the extent and direction of charge transfer at an exsolved interface may depend on its precise preparation history. The potential for tunable EMSI at the exsolved metal-oxide interface could offer promising avenues for optimizing the catalytic performance of exsolved nanocatalysts.