Lattice dynamics and thermal transport in crystalline solids from machine learning interatomic potentials

Abstract

Thermal transport properties of insulating or semiconducting crystalline solids are dominated by their anharmonic lattice dynamics. Combining Boltzmann transport equation, the lowest-order perturbation theory, and accurate atomic forces calculated from density functional theory, many studies have demonstrated that the lattice thermal conductivity and the phonon coupling effects can be reliably studied from first principles. However, in materials with restricted phase space or extremely strong anharmonicity, more sophisticated models were shown to be needed for accurate computation of phonon lifetime and lattice thermal conductivity. In this presentation, the calculation of anharmonic terms based on machine learning interatomic potentials, and the effects on phonon lifetimes of a range of materials will be discussed. The application of a two-channel lattice thermal transport model, which incorporates the population and coherence contributions, combining with phonon lifetimes obtained from both perturbation theory and molecular dynamics simulation will be discussed.