Impact of grain boundaries on thermal conductivity in hexagonal boron nitride sheets: a study based on neural network potentials

Hexagonal boron nitride (h-BN) has been predicted and confirmed to exhibit a high thermal conductivity which makes it a viable candidate for a building block in 2D materials-based electronic devices. As defects and grain-boundary often exist in h-BN, these imperfections could hinder such high thermal conductivity. In this work, we first develop a neural network interatomic potential for h-BN that provides reliable energetics of the grain boundaries and then perform non-equilibrium molecular dynamics simulations to study the effect of grain-boundary on the thermal conductivity. The neural network potential is trained using data from ab initio molecular dynamics simulation of small h-BN sheets with defects and grain-boundaries and validated through comparison of the results of the non-equilibrium MD simulations with prior experimental and theoretical observations. We show that the calculated thermal conductivity of a pristine h-BN sheet agrees with experimental and theoretical data (~ 500 W/m-K) and that the presence of grain-boundary in h-BN sheet increases its thermal resistance.

Work is supported in part by DOE grant DE-FG02-07ER46354.