Prediction of phonon properties of cubic boron nitride with vacancy defects and isotopic disorders by using a neural network potential

Abstract

Cubic boron nitride (c-BN) is a promising ultra-wide bandgap semiconductor for high-power electronic devices. Its thermal conductivity can be substantially modified by controlling the isotope abundance and by the quality of a single crystal. Consequently, an understanding of the phonon transport in c-BN crystals, with both vacancy defects and isotopic disorders at near-ambient temperatures, is of practical importance. In the present study, a neural network potential (NNP) for c-BN has been developed, which has facilitated the investigation of phonon properties under these circumstances. As a result, the phonon dispersion and the three- and four-phonon scattering rates that were predicted with this NNP were in close agreement with those obtained from density-functional theory (DFT) calculations. The thermal conductivities of the c-BN crystals were also investigated, with boron (B) vacancies ranging from 0.0% to 0.6%, by using equilibrium molecular dynamics simulations based on the Green-Kubo formula. These simulations accurately capture vacancy-induced phonon softening, localized vibration modes, and phonon localization effects. As has previously been experimentally prepared, four isotope-modified c-BN samples were selected for analyses in the evaluation of the impact of isotopic disorders. The calculated thermal conductivities aligned well with the DFT benchmarks. In addition, the present study was extended to include a c-BN crystal with a natural abundance of B atoms, which also contained B vacancies. Reasonable thermal conductivities and vibrational characteristics, within the temperature range of 250–500 K, were then obtained.