Long timescale molecular dynamics simulations of carboxylic acid-modified anatase TiO2(101)-water interfaces using ab-initio deep neural network potentials

Abstract

Carboxylic acid-modified anatase TiO${}_2$-water interfaces are widely relevant, yet understanding of their molecular scale structure is limited. To help improve this understanding, we here construct a deep neural network potential (DP) that accurately represents the potential energy surface of the formic (FA) and acetic acid (AA)-covered anatase TiO${}_2$(101) (A101) interfaces with water predicted by Density Functional Theory (DFT) with the SCAN exchange–correlation functional. Long time-scale (ns) Molecular Dynamics simulations employing such DP provide insight into the hydration structure at the interface, showing how the water density profile and radial distribution functions depend on the coverage and adsorption configurations of the acids. The developed model sets the stage for estimating the adsorption energetics of these small carboxylic acids on the A101 surface in an aqueous environment.