Non-Pt catalysts can be broadly classified into oxide-based and carbon-based catalysts, but considering the start-up and shutdown potentials of PEFCs, oxide-based non-Pt catalysts are expected to show higher durability. It is known that the introduction of oxygen-deficient sites in the oxides of Group 4 and Group 5 elements in the oxide-based non-Pt catalysts results in a high oxygen reduction reaction rate (ORR). Therefore, oxygen-deficient sites are considered to be active sites for ORR, and the introduction of oxygen-deficient sites in the crystal lattice is a guideline for catalyst design in non-Pt catalysts based on oxides. On the other hand, it has been reported that noble metal oxides show high ORR activity even without clear oxygen-deficient sites. It is easy to imagine the hypothesis that the ORR proceeds by the adsorption of oxygen molecules on the oxygen-deficient sites, but the same hypothesis holds true for low-coordinated metal atoms on the oxide surface without the introduction of oxygen-deficient sites. Furthermore, it has been reported that in alkaline solution, oxygen atoms in the perovskite structure are the active sites. These discrepancies are due to the fact that the ORR mechanism on oxides is still unknown.

In this study, in-situ XAFS measurements were performed to identify the active sites of ORR on oxides, and changes in the chemical states of metal elements during the ORR were investigated from changes in XANES spectra. A noble metal oxide (Rh₂O₃), whose valence changes with potential, was used as the target oxide.

The target oxide, Rh₂O₃, was prepared by the Adams method. The prepared sample was placed on a glassy carbon plate at 100 g cm^-2 and placed in a tripolar electrolytic cell for in-situ XAFS. The in-situ XAFS cell was designed to detect X-ray fluorescence by irradiating X-rays from the back of the sample. Energy of the incident X-ray was set near the Rh K edge. In-situ XAFS measurements were performed at five points: 0.4 V, where the ORR is in progress; 0.6 V and 0.8 V, near the ORR starting potential; 1.0 V and 1.2 V, where the ORR is not in progress. To investigate the effect of oxygen partial pressure, the oxygen partial pressure varied between 0%, 25%, 50%, 75%, and 100%.

Comparing the height of the white line in nitrogen at 0.4 V and at 1.2 V, which is the most reducing state, the height of the white line was lower when the applied voltage was 0.4 V. This is because the average valence of Rh was higher at 0.4 V than at 1.2 V, which is the most reducing state. This indicates that the average valence of Rh changes with applied voltage. In other words, at 0.4 V, when the ORR is fully advanced, the oxidation number of Rh is considered to be in a reduced state. This behavior is consistent with the electrochemical behavior and with previous reports.

In contrast, when the oxygen partial pressure was set to 25%, this difference in the applied voltage was hardly observed. The difference in the white lines between the high and low potential sides became smaller as the oxygen partial pressure increased, and under 100% oxygen, the XANES spectra obtained at the applied voltage of 0.4 V and 1.2 V almost overlapped. Therefore, unlike under a nitrogen atmosphere, the oxygen gas adsorbs on the Rh₂O₃ surface at 0.4 V, where the ORR proceeds under oxygen distribution. As a result, it is considered that the oxidation number of Rh increased. From the above, we predict that in Rh₂O₃, the ORR progresses in the state where the valence of Rh is decreased.