Machine learning for predicting thermal transport properties of solids

Abstract

Quantitative descriptions of the structure-thermal property correlation have always been a challenging bottleneck in designing functional materials with superb thermal properties. In the past decade, the first-principles-based modeling of phonon properties using density functional theory and the Boltzmann transport equation has become a common practice for predicting the thermal conductivity of new materials. However, first-principles calculations of thermal properties are too costly for high-throughput material screening and multi-scale structural design. First-principles calculations also face several fundamental challenges in modeling thermal transport properties, for example, of crystalline materials with defects, of amorphous materials, and for materials at high temperatures. In the past five years or so, machine learning started to play a role in solving the aforementioned challenges. This review provides a comprehensive summary and discussion on the state-of-the-art, future opportunities, and the remaining challenges in implementing machine learning techniques for studying thermal conductivity. After a brief introduction to the working principles of machine learning algorithms and descriptors for characterizing material structures, recent research using machine learning to study nanoscale thermal transport is discussed. Three major applications of machine learning techniques for predicting thermal properties are discussed. First, machine learning is applied to solve the challenges in modeling phonon transport of crystals with defects, in amorphous materials, and at high temperatures. In particular, machine learning is used to build high-fidelity interatomic potentials to bridge the gap between first-principles calculations and empirical molecular dynamics simulations. Second, machine learning can be used to study the correlation between thermal conductivity and other relevant properties for the high-throughput screening of functional materials. Finally, machine learning is a powerful tool for structural design to achieve target thermal conductance or thermal conductivity. This review concludes with a summary and outlook for future directions for implementing machine learning in thermal sciences.