An automated and accelerated computational framework to compute finite temperature properties including anharmonic effects in different property domains.

Accurately predicting the properties of solids at finite temperatures is a computationally intensive task. Calculating thermodynamic, mechanical, and thermal transport properties simultaneously can greatly increase the computational cost. Traditionally, different computational setups are required to predict these three types of properties or domains. Moreover, the information generated while computing one domain is generally not utilized in the other two. Besides the high computational cost, traditional methodologies often overlook the need for incorporating high-temperature corrections that stem from the anharmonicity of the material. To tackle these challenges, an automated and accelerated computational framework has been developed to enable the computation of finite properties of the three above-mentioned domains simultaneously. This new approach utilizes interdomain data efficiently and combines accelerated methodologies such as machine learning regression for the extraction of high-order force constants and the quasi-harmonic three-phonon method to reduce the computational cost without compromising accuracy [1-3]. Temperature-dependent phonons are included in the calculation to include strong anharmonic effects. The methodology can be used with either classical interatomic potential or DFT-based codes.