Crystallization of h-BN by molecular dynamics simulation using a machine learning interatomic potential

Abstract

This study employs machine learning-driven molecular dynamics simulations to investigate the structure and physical properties of hexagonal boron nitride (h-BN) across a wide temperature range. A novel machine learning potential, neuroevolution potential, trained using first-principles calculations, is utilized for high-accuracy and efficient atomic-level simulations. Results reveal that slower cooling rates enhance the crystallinity and mechanical properties of h-BN. The unique temperature dependence of the elastic modulus and anisotropic thermal conductivity of h-BN are also elucidated. Furthermore, a predictive model is developed to estimate the tensile strength of low-density porous h-BN, providing a theoretical foundation for its potential applications in various fields such as electronics, aerospace, and energy storage.