Controlling the oxygen reduction reaction (ORR) is a pivotal challenge at the heart of energy conversion science. It's an area of extensive research across various scientific disciplines, including catalysis and biological reactions. Developing materials that maximize the efficiency of the 4-electron oxygen reduction reaction $\left(\frac{1}{2}\mathrm{O}_{2} + 4\mathrm{e}^{-} + 4\mathrm{H}^{+} \to 2\mathrm{H}_{2}\mathrm{O}\right)$ through their electrocatalytic properties is crucial. Industrially, platinum surfaces are commonly used. However, their high cost and limited durability necessitate the search for novel materials. This has led to the development of acid-resistant oxides. Doped TiO₂ and ZrO_$x$N_$y$, for instance, have shown activity comparable to platinum. Nevertheless, understanding their reaction mechanisms remains a pressing challenge.

First-principles calculations based on Density Functional Theory (DFT) are the most effective way to investigate reaction activity. However, we face several hurdles related to DFT accuracy and electrode-interface modeling when it comes to materials like TiO₂ and ZrO_$x$N_$y$. The ORR begins when oxygen molecules adsorb weakly. The accuracy of DFT for these weakly adsorbed oxygen systems must be carefully validated. Furthermore, oxygen must compete successfully with water molecules for absorption. To examine the content of competition, sampling of adsorption structures of water molecules is necessary. Accurately modeling the surface also demands extensive sampling of various dopant configurations. It is clear that these challenges have hindered theoretical research in this area. This report will explain how we overcame these obstacles and provide crucial insights into the activity mechanisms of these promising materials.

Accuracy of O₂ adsorption [1]: To validate the accuracy of oxygen adsorption, we compared experimental Temperature-Programmed Desorption (TPD) spectra with van der Waals-corrected $\mathrm{DFT} + U$ calculations on the anatase-TiO₂(101) surface. The results showed excellent agreement between the experiment-based and DFT-based simulations, both indicating weak adsorption of approximately 0.2 eV. This comparison between TPD and DFT is a novel approach.

Applying this method to the Pt(111) surface, we observed agreement for dissociative adsorption, albeit with a certain degree of correction. However, for molecular adsorption, the adsorption energy was significantly overestimated. This suggests that more advanced DFT functionals are necessary for accurate predictions on this system.

Adsorption Competition on ZrO_$x$N_$y$ Surfaces [2]: To understand the competition between oxygen and water molecules for adsorption sites on the ZrO_$x$N_$y$(101) surface, we exhaustively sampled water molecule structures. By employing a machine learning force field, which allowed us to maintain DFT accuracy while achieving the computational efficiency of classical force fields, we performed approximately 1000 molecular dynamics simulations, each on the nanosecond scale, to determine average distributions. We found that on defect-free ZrO₂ (where $x=$2, $y=$0), the water molecule layer was weakly adsorbed and maintained a distance from the surface. In contrast, on the defective surface (where $x=$8/7, $y=$4/7), the distance shortened, indicating stronger adsorption. Notably, water molecules exhibited a strong tendency to adsorb away from the oxygen vacancies present on the surface (which were previously shown to be the O₂ adsorption site [3]), suggesting they do not compete with oxygen molecules for these specific adsorption sites. This finding supports the scenario where oxygen vacancies are stabilized by nitrogen impurities, thereby avoiding site competition.

Role of Surface Oxygen Vacancies: While our previous first-principles Monte Carlo calculations have shown that nitrogen impurities stabilize surface oxygen vacancies, it is not immediately obvious whether these vacancies act as active sites. This is because a simple comparison of adsorption energies for oxygen and water molecules under vacuum conditions suggested that water molecules would adsorb preferentially. Therefore, it is surprising that water molecules are stable in a vacuum yet become unstable at interfaces surrounded by other water molecules. The exact mechanism of this instability is unknown, but our simulations clearly demonstrate this tendency. The significance of our simulations lies in their ability to explain observational results by thoroughly sampling heterogeneous systems. This is a promising step forward for future research on heterogeneous catalytic reactions.

References

• [1] S. Muhammady et al., arXiv:2506.18225.
• [2] A. Nakanishi et al., J. Phys. Chem. C. 129, 2403 (2025).
• [3] S. Muhammady et al., J. Phys. Chem. C 126 15662 (2022).

Authors

• O. Sugino, S. Kasamatsu^a, S. Muhammady^b, A. Nakanishi^c, and  J. Haruyama^b
• ^aYamagata University
• ^bRIKEN
• ^cTsukuba University