Title: Exploring thermal transport properties of solids by machine learning

Abstract

Micro/nanoscale heat conduction is crucial for a broad range of applications such as thermal management of electronic devices, thermal insulation, and thermoelectrics. Understanding and designing of thermal transport properties in solid materials largely depends on atomistic simulations based on density functional theory (DFT) or empirical potentials, which however suffer either low computational efficiency or accuracy. In recent years, machine learning is emerging as a powerful tool to bridge the gap between DFT and empirical simulations. The applications of machine learning in exploring thermal transport properties of solids mainly include constructing interatomic potentials and predicting thermophysical properties of materials. In this talk, we will introduce our recent progess in building machine learning interatomic potentials and predicting interfacial thermal resistance using machine learning. On one hand, we have developed machine learning interatomic potentials that can accurately describe phonon transport in materials containing point defect, grain boundary structures, and layered materials. The machine learning interatomic potentials achieve DFT-level accuracy and 3 to 5 orders of magnitude higher efficiency than DFT. On the other hand, we have developed machine learning models to accurately predict the thermal resistance of non-metallic/non metallic and metal/non-metallic interfaces.