(190e) Elucidating Adsorbate Interactions at Platinum-Water Interface Using Machine Learning Potentials.

Electrocatalysis plays a pivotal role in the development of a sustainable energy ecosystem. These electrochemical reactions take place at electrified solid-liquid interfaces, which pose challenges for precise characterization using experimental methods. Therefore, there is a need for computational tools that can offer accurate atomic-scale descriptions of these interfaces. Presently, due to the lack of resources the computational approaches only implicitly account for the liquid interactions at the solid-liquid interface. The effects arising from explicit solvent molecules are therefore ignored. To address this gap, machine learning potentials (MLPs) have emerged as a hybrid solution. MLPs offer a means to integrate the information from explicit solvent calculations without incurring the high computational costs, thus achieving required computational efficiency and chemical accuracy.

In this study, we developed MLPs to model the behavior of oxygenated adsorbates on various platinum-water interfaces, aiming to accurately characterize the water structure surrounding the adsorbates. To assess the accuracy of these MLPs, we used adsorbate solvation energies as an evaluation metric. Solvation, which refers to the stabilization of solute by solvent, can significantly impact the thermodynamics and kinetics of reactions. Our analysis reveals that a reduction in force/energy errors does not necessarily ensure accurate solvation energies. Additionally, contrary to prevailing assumptions, we observed limited transferability of MLPs across different platinum terrace and stepped surfaces. Nonetheless, MLPs offer substantial advantages in accessing longer time and length scales for a given Pt surface, leading to statistically more accurate solvation energy computations. Finally, we compared site-specific solvation energies for OH* and OOH* adsorbates on Pt (111), Pt (221), and Pt(322) surfaces within the context of the oxygen reduction reaction, a crucial industrial electrochemical process. Our findings suggest that solvation energies can influence the rates of sites near steps and provide insights into discrepancies between computational and experimental results.