Accelerating Thermodynamic Simulations of Electrochemical Interfaces at the Atomic Scale

Surfaces and interfaces play a critical role in diverse applications, including catalysis, energy storage, and electronics. Traditional thermodynamic studies of material surfaces relied on a limited set of guess structures validated by costly first-principles calculations. This approach is insufficient for exploring the vast compositional and configurational spaces required by complex materials in use today. Recent advancements, such as the Virtual Surface Site Relaxation-Monte Carlo (VSSR-MC) algorithm developed in our group [1], leverage machine learning force fields (MLFF) and various sampling strategies to accelerate surface reconstruction studies under vacuum and gas conditions. In this work, we extend the VSSR-MC algorithm to investigate aqueous electrochemical interfaces by developing a framework to describe thermodynamic equilibria under the Pourbaix grand potential. We demonstrate that a fine-tuned foundational MLFF can reveal surface reconstructions of perovskite materials relevant to electrocatalysis. Finally, we construct surface Pourbaix diagrams that enhance our understanding of electrochemical interfaces compared to previous studies.<br/><br/>[1] Du, X. <i>et al.</i> Machine-learning-accelerated simulations to enable automatic surface reconstruction. <i>Nat Comput Sci</i> 1–11 (2023)