Electrocatalytic oxygen evolution reaction (OER) has been extensively studied because it is of fundamental importance in renewable energy technologies, such as water splitting, CO₂ reduction, rechargeable metal-air batteries, and regenerative fuel cells [1-3]. OER is known for its complexity and slow kinetics owing to four concerted proton-electron transfer steps as well as oxygen-oxygen bond formation [4, 5]. Therefore, developing efficient OER catalysts and understanding the mechanism involved are imperative and are the main focus of the research [2, 6].

Recently, numerous experimental results have pointed out that OER catalysts can undergo a drastic self-reconstruction during OER treatment [7-11]. For instance, Fabbri et al. [12] reported that dynamic surface self-reconstruction was the main driving force for highly OER-active perovskite-based catalysts. In our previous work, we found that the surface of the catalyst was transformed into a definite coordination-unsaturated structure under electrochemical conditions, which is responsible for the outstanding performance of the material [13]. Unfortunately, the complexity of the multi-component structure impedes the precise structure identification, thus greatly limiting the understanding of the OER process over these catalysts, despite employing the operando techniques [14-16]. Thus far, probing the reconstruction and identifying the true catalytically active components at an atomic resolution has remained highly challenging. Therefore, it is necessary to develop a series of catalysts with well-defined structures to explore the origin of the surface reconstruction and the relation between the surface structure and OER activity.

Perovskite oxides (ABO₃) have been regarded as promising electrocatalysts toward the OER [17-20]. In this work, we prepared perovskite LaCo_0.8Fe_0.2O_3-δ with different Sr doping levels. The electrocatalytic performance indicated that La_0.2Sr_0.8Co_0.8Fe_0.2O_3-δ (LSCF28) exhibited a low overpotential and small Tafel slope. Surface analysis indicated that the thickness of an amorphous layer of the material after reaction increases with increasing Sr content. The amorphous layer was considered as an activity center, which had also been found previously [10, 21, 22]. We found that as the Sr content increased, no noticeable change in the valence states of Co and Fe ions was observed, whereas doped holes at the O 2p state were clearly observed.

The La_1-ySr_yCo_0.8Fe_0.2O_3-δ (y = 0, 0.4, 0.8) denoted as LCF, LSCF64, and LSCF28 were readily synthesized by a molten salt method (see Supporting Information for details). In the precipitation reaction, a salt mixture acted as a solvent during the reaction leading to a short reaction time and significantly small crystal sizes [23, 24]. The structural information of the as-prepared powder is shown in Fig. 1(a). It was found that the perovskite-type structure was obtained, as shown the Fig. 1(b). In addition, a notable shift of the diffraction line toward lower 2θ angles was observed with increasing the Sr content. This can be ascribed to the relatively large ionic radius of the Sr(Ⅱ) ion (1.44 Å) as compared with that of the La(Ⅲ) ion (1.36 Å) [25]. Scanning electron microscopy (SEM) was employed to determine the morphologies of the samples. In previous work, it has been found that the Fe content and the molten-salt approach are essential for producing a specific porous structure in LSCF perovskite oxides [26]. Here, we found that with increasing the Sr content, the porous structure and nanosize of the particles remained unchanged, as shown in Fig. S1. Furthermore, BET experiments indicated that all the particles had almost the same specific surface areas, as plotted in Fig. S2. Conclusively, the Sr content slightly affects the morphology of the sample.

To further study the electronic and geometric structures, we employed X-ray absorption spectroscopy (XAS) measurements. Generally, XAS is a powerful technique for investigating the oxidation states of transition metals with element-selective characters. The Co K-edge X-ray absorption near-edge structure (XANES) for LCF, LSCF64, and LSCF28 as well as for the reference samples are shown in Fig. 1(c). It could be clearly observed that the line shapes of the Co K-edge XANES spectra for the catalysts are very similar. We can conclude that the electronic structures and the local structures of the Co ions did not change upon Sr doping. Generally, the position of the adsorption edge reflects the oxidation state of the Co ion [27]. The absorption edges in LCF, LSCF64, and LSCF28 are located at the same energy as that of the Co³⁺ reference of CoOOH, suggesting that the Co ions exist in the Co(Ⅲ) oxidation state. The spectral evolution and absorption peaks can be discerned more clearly in the derivative curves of the XANES spectra, as confirmed in Fig. S3(a), which also revealed that the oxidation state was unchanged [28-31]. Moreover, the Fe K-edge XANES spectra in Fig. 1(d) and Fig. S3(b) revealed that the Fe ion has a similar trivalent state.

Considering that the XANES at the transition metal K-edge has a typical probing depth in micrometers, we performed soft XAS measurements at the transition metal L_{2, 3}-edge to study the valence states of Co and Fe ions at the sample surface using a total electron yield (TEY) mode, which has a typical exploring depth of 2-5 nm. The room-temperature Co L_{2, 3}-edge XAS spectra of the LCF, LSCF64, and LSCF28 as well as LaCoO₃, serving as a trivalent reference, are shown in Fig. 2(a). It was found that LCF has a Co³⁺ state, which remained unchanged with increasing Sr content. Similarly, the Fe L_{2, 3}-edge XAS spectra show that the Fe ion has a trivalent state, as illustrated in Fig. 2(b).

The soft X-ray absorption spectrum at the O-K edge is known to be very sensitive to the doped hole at the O 2p state [32]. Fig. S4 shows the O-K XAS spectra of the samples studied. Noticeably, Sr doping leads to a new pre-edge peak at 528 eV, which represents the doped hole at the O 2p state [33, 34]. However, the doped holes cannot totally compensate the Sr doping level, because, at Sr = 0.8, the spectral weight of the energy at 528 eV is much lower than that of Li_0.66CoO₂ [34, 35]. Therefore we can conclude that the high Sr content would lead to oxygen vacancies.

The electrocatalytic activities of all the obtained samples and IrO₂ reference for OER were evaluated by linear sweep voltammetry (LSV) measurements conducted in an O₂-saturated 0.1 mol L^-1 KOH at a scan rate of 5 mV s⁻¹ and a rotation speed of 1600 rpm, as plotted in Fig. 3(a). All potential values were IR-corrected during each LSV measurement to compensate for the resistance of the solution. The LSV curve measurements were performed on the fully activated sample that underwent cyclic voltammetry (CV) scans. It is important to study the overpotential at a current density of 10 mA cm⁻², which matches the sunlight flux (based on 10% solar-to-fuel conversion efficiency) [36]. We found that LSCF28 had approximately the same overpotential, which achieved a current density of 10 mA cm⁻² at a comparably small η value of 0.35 V, which is more negative than those of LSCF64 (η = 0.42 V) and LCF (η = 0.53 V). Moreover, as displayed in Fig. 2(b), LSCF28 presents a smaller Tafel slope of 87 mV dec⁻¹ compared to those of the others. These results clearly demonstrate the potential of LSCF28 as an oxygen evolution catalyst.

To gain more information on the kinetics of the OER, electrochemical impedance spectroscopy (EIS) measurements were performed to obtain the charge transfer resistances shown in Fig. 3(c). LSCF28 possessed a good charge transfer resistance (R_ct) with the value of 58 Ω, while those of the LSCF64 and LCF samples were 83 and 125 Ω, respectively. Long term testing is essential for determining the duration of application. As shown in Fig. 2(d), we also conducted chronopotentiometric measurements to evaluate the catalytic durability of LSCF28, and a nearly constant potential was observed for over 48 h, as well as good electrochemical stability.

To further compare with the specific activity, we calculated by normalizing with the real oxide surface areas, using BET theory. The results are plotted in Fig. S5. Furthermore, according to the SEM analysis, the samples exhibited similar porous morphologies and specific surface areas. However, it could be observed that the OER activity increased with the Sr content. Therefore, we considered that the additional increase in the OER activity implies that there is an underlying effect factor.

Many studies have reported that the electronic and crystal structures of catalysts change after OER treatment [37]. Therefore, it was highly desirable to investigate the surface structures of the samples after electrochemical treatment. High-resolution transmission electron microscopy (HRTEM) can provide direct confirmation of the change in the morphology at the near-surface regions of the materials after the electrochemical process, as shown in Fig. 4(a) and 4(b). Noticeably, after the reaction, the surface of LCF experienced a slight change. While, a visible amorphous surface layer with a thickness of about 3 nm could be observed in LSCF64, as plotted in Figs. S6(a) and S6(b). Furthermore, the amorphous surface layer of LSCF28 reached a thickness of about 6 nm in Fig. 4(c) and 4(d). It was previously mentioned that the amorphous layer is the actual origin of the OER activity. Here, from the XAS study, we found that the valence states of Co and Fe ions have almost no effect on the formation of the amorphous layer. Although the O 2p holes appear upon Sr doping, it cannot fully compensate the Sr doping level. Therefore we can safely conclude that the surface amorphous layer, which is the true origin of the OER activity, is related to the occurrence of the oxygen vacancies upon heavily increasing the Sr doping level.

In conclusion, we reported a system of Sr-doped perovskite prepared by the molten salt method. The LSCF28 component (Sr = 0.8) in the system exhibited the best OER activity with a low overpotential of 0.35 V and small Tafel slope of 87 mV dec⁻¹ in 0.1 mol L^-1 KOH. More importantly, we found that: (1) Sr-doping leads to the formation of an amorphous layer, whose thickness increases with the Sr content; (2) upon Sr-doping, the valence states of the Co and Fe ions remain nearly unchanged, whereas O 2p holes appear, which cannot compensate the Sr doping level. From the above observation, we can conclude that the Sr doping results in oxygen vacancies and the formation of an amorphous layer, which is responsible for the outstanding OER activity. Our finding provides new insights into the design of active OER catalysts.

Acknowledgments

We thank Renduo Liu and Yu Wang for help with TEM and XAFS measurements, respectively.