As future alternatives to Pt-based materials, group-IV oxides have been proposed for electrocatalysts against the oxygen reduction reaction (ORR) in acid solution [1]. Those oxides are abundant and electrochemically stable while Pt is scarce and soluble in acid solution. Although pristine oxides are nonreactive, they can be made reactive by introducing defects. For example, in the case of TiO₂, doping Nb atoms and introducing oxygen vacancies (V_Os) makes the oxide reactive [2]. However, the microscopic mechanisms of the reaction at defective oxide surfaces are still not well understood. In order to improve our understandings on the ORR at defective oxides, we have performed density functional theory simulations. We have studied defective TiO₂ as an example of defective oxides.

A challenge in theoretical study of the reaction on defective oxides is high complexities of the surfaces. At defective oxide surfaces, there are many possible active sites, defect configurations and adsorbed structures of reaction intermediates. In order to elucidate the reaction on the surface, extensive investigations are mandatory. For an extensive investigation, we have studied two major polymorphs of TiO₂ (rutile and anatase), several defect configurations, several surface orientations and many adsorbed structures of intermediates. As for surface orientations, we have studied the rutile (110) surface, the anatase (101), (100), (001) and (001)-(1x4) reconstructed surfaces. In total, we investigated the total energies of about 500 configurations.

By calculating the free energies of reaction intermediates using PBE+U functional with computational hydrogen electrode model, we found the followings. (1) At Nb-doped TiO₂ surfaces, calculated free energy of OH* intermediate is very low. This indicates that this surface is poisoned by OH* intermediate and nonreactive against ORR. The strong adsorption is because the adsorbed OH captures an electron from Ti³⁺ generated by a Nb_Ti defect and thus produced OH^- is very stable. (2) At oxygen-deficient surfaces (V_O surfaces), the calculated free energy of O* intermediate at V_O site is very low, so the V_O surface seems to be nonreactive and poisoned by O* intermediate. This is because the O* intermediate can capture two electrons from two Ti³⁺, which arise by making V_O on the surface, and the resulting surface are very stable because they are pristine surfaces. Our calculations indicate that both the Nb-doped surface and Vo surface are likely to be nonreactive. (3) If we take the OH poison of Ti³⁺ existing surface into consideration, it is natural that there exists the OH poisoned V_O surface (V_O+2OH* surface), where the Vo site is unoccupied and the two surface Ti sites are adsorbed by OH* intermediates. On this surface, there are no Ti³⁺ because OH* intermediates capture electrons from Ti³⁺s, and so the V_O site may be reactive. Indeed, at the V_O+2OH* surface, the calculated adsorption strength of O* intermediate is moderate, this means the surface is likely to be reactive. The calculated free energies of the entire reaction intermediates suggest that the anatase (001) surface has much higher reactivity than other surfaces. These results indicate that the V_O site without Ti³⁺ is the most likely reactive site of the ORR on defective TiO₂ surfaces.

1. K. Ota, Y. Ohgi, K. Matsuzawa, S. Mitsushima and A. Ishihara, ECS Transaction 45(2), 27 (2012).
2. A. Ishihara, M. Hamazaki, M. Arao, M. Matsumoto, H. Imai, Y. Kohno, K. Matsuzawa, S. Mitsushima and K. Ota, J. Electrochem. Soc. 163, F603 (2016).