(678c) A Theoretical Study of Halide Adsorption on Metal Surface: Grand-Canonical Molecular Dynamics Based on a Machine Learning Force Field

Understanding the liquid-solid interface from first principles has been a long-standing challenge that has implications for electrocatalysis and the solution-phase synthesis of metal nanocrystals. Traditionally, theoretical studies have relied on ab initio thermodynamics. However, these calculations are hampered by small unit cells, which practically necessitate studies of highly ordered interfaces, when this may not be the case in experiments. In this work, we introduce techniques based on Ab Initio Grand-Canonical Monte Carlo (AIGCMC) and we demonstrate them in a study of chloride adsorption on Pt and Pd surfaces. Halide adsorption has been demonstrated to affect the growth and assembly of metal nanocrystals in solution-phase synthesis and experiments. However, it has been a challenge to understand the atomic behavior in current experiments. Using the VASP code, we performed AIGCMC studies, and we fit a machine learning force field (MLFF) to perform GCMC. These methods can explore various configurations on the metal surfaces, and we can explore much larger unit cells using GCMC than we can using traditional ab initio thermodynamics. We obtain surface phase diagrams as a function of the halide chemical potential, and we observe disordered halide configurations that cannot be captured in conventional DFT total energy calculations. Our phase diagrams exhibit experimentally observed halide adsorption configurations. We show that this computational technique can result in more realistic surface description of metal nanocrystals that could lead to qualitative changes in predicted shapes.