The electrified graphene-water interface is a vital component in many energy storage applications. However, understanding the interfacial properties is challenging due to the requirement of a high-quality atomic interfacial model. Recently, the hybrid solvation model, including the computationally affordable implicit solvation model and a thin layer of explicit water solvent slab next to the solid, has become a promising approach to address this issue. The identification of the rational explicit water slab thickness holds the key to the computational results by using this hybrid solvation model. In this study, we present a framework combining ab initio molecular dynamics (AIMD) and supervised machine learning (ML) to address this challenge. Based on the database from the AIMD simulations, the relationship between the total energy of the system and the distance from the oxygen in water molecules to the graphene was successfully identified through supervised ML. Our results further demonstrate that the first few layers of water next to the graphene play the decisive role in the change of the total energy. The cutoff thickness of 7 Å can reproduce the majority of the impact of the solvent on the total energy change of the water-graphene system. The success of this ML-assisted platform suggests it can also be used as a protocol to build the hybrid solvation model for understanding other electrified solid-liquid interfaces.
Wen et al. (Wed,) studied this question.