Seiichiro Ten-no

Oxygen-vacancy diffusion in SrTiO$_3$ is accurately predicted over long timescales using neural network potentials, resolving a key limitation of ab initio methods and outperforming classical MD.