Impact of surface vacancies on the dynamics of metal atoms on reducible oxides: an enhanced sampling study with machine-learned potentials for Pt1/TiO2

Abstract

Surface vacancies play an important role in the chemistry of reducible oxides such as TiO₂. It has been shown through experiments and quantum chemistry calculations that surface vacancies can trap precious metal atoms more effectively than pristine surfaces. In this study, we perform nanosecond timescale molecular dynamics (MD) and metadynamics simulations at 1000 K to probe the evolution of platinum metal atoms adsorbed at or adjacent to cationic and anionic surface vacancies on rutile titania (110). To this end, we employ machine learned interatomic potentials (MLIPs) that are iteratively trained on density functional theory (DFT) data. MD simulations show that Pt atom motion is subdiffusive across both anionic and cationic vacancies, indicating that these sites are thermally stable. Near-linear O–Pt–O motifs, found in earlier studies to be stable on the pristine surface of TiO₂, are no longer stable in the vicinity of a vacancy. We utilize well-tempered, multiple walker metadynamics to map the free energy landscape of Pt migration out of the vacancy sites. The free energy barriers to Pt migration are highest for the cationic vacancy, followed by bridging oxygen and basal oxygen vacancies. The fact that Pt at the cationic vacancy site is the most stable is encouraging because it has also been shown to be one of the most active sites for reactions such as CO oxidation.