Machine-Learned Force Fields for Investigating the Interfacial Properties of HOPG Structures: Implications for Supercapacitors and Adsorption

This study investigates the interfacial properties of highly oriented pyrolytic graphite (HOPG) surfaces using ab initio molecular dynamics (AIMD) combined with on-the-fly training of machine-learned force fields (MLFF) for improved accuracy and experimental consistency. The model’s accuracy was validated through extensive AIMD simulations, dipole moment, electrostatic potential calculations, and force–distance evaluations. We observed that introducing a vacuum surface or defects, such as atom removal or dopants, increases the total energy, reducing system stability compared to the pristine structure. Defects and dopants also alter interlayer binding energy and electronic conductivity. The radial distribution function shows changes due to defects, with nitrogen doping causing strong C–N bonding, oxygen incorporation leading to altered local bonding environments highlighting the electronic redistribution, and the C–S bond stretching beyond the hexagonal lattice, leading to noticeable distortions. These changes result in shifts in the vibrational density of states, indicating modifications in electron–phonon interactions that can affect electrical conductivity and superconductivity. Additionally, localized dipoles are created by dangling bonds, edge states, or dopants, with oxygen doping causing the most significant potential fluctuations, nitrogen inducing moderate changes, and sulfur having the smallest effect, offering stability but less tunability. These findings highlight that even minor microstructural deviations can significantly influence the behavior of carbon-based materials, especially at solid–liquid interfaces. This is particularly relevant for the development of advanced energy storage systems, such as supercapacitors, where the performance is critically dependent on the properties of the electrode–electrolyte interface.