Neural network approach to diffusion of B and N adatoms on the Pt(111) surface

Artificial neural network boasts an outstanding class in terms of its performance and efficiency and is being spread widely in most fields of physics. We report our investigation of the diffusions of boron and nitrogen adatoms on the Pt(111) surface by performing molecular dynamics simulations equipped with machine-learned potentials. Platinum is commonly used as a substrate for the growth of hexagonal boron nitride (h-BN) thin films, and the diffusion of B and N atoms on the substrate, which are decomposed from the precursor molecules, plays important roles in the initial stages of h-BN growth. The two-dimensional potential energy surfaces and the trajectories of the B and N adatoms are consistent with the DFT calculation. The Arrhenius plots of the diffusion coefficients produce the diffusion barriers of the B and N adatoms on Pt(111), which agree well with the DFT barriers.