Publication date: 22nd July 2026
Electrified metal–electrolyte interfaces are central to electrochemical processes such as electrocatalysis, corrosion, and energy storage, yet resolving how solvated ions behave under an applied potential at the atomistic level remains challenging. Here we simulate charged Au(111)–water interfaces containing solvated Na+ and F- ions, as a model system for such electrified interfaces. Reference calculations are performed in VASP [5-8] using the Coulomb-kernel-truncation method [1] to correctly treat the electrostatics of charged periodic simulation cells. The ab initio reference calculations are used to train two-layer GRACE [4] machine-learning force fields, enabling long-timescale, ML-accelerated molecular dynamics. We apply the weighted histogram analysis method (WHAM) [2] to obtain free energy profiles of ion adsorption as a function of distance from the surface, and metadynamics [3] to explore the ion-pairing/dissociation free energy landscape of the combined Na⁺/F⁻ system. The free energy profiles from the single-ion simulations closely reproduce those from the independent Na⁺/F⁻ metadynamics run, validating this efficient single-ion approach as a reliable substitute for the more costly explicit ion-pair sampling. This DFT-trained, ML-accelerated workflow offers a general, transferable route for studying ions at charged, electrified interfaces, of relevance to electrocatalysis, electrolyzers, and batteries.
