Machine Learning Approaches for Accelerating the Discovery of Thermoelectric Materials
Luis M. Antunes, Vikram Vikram, José J. Plata, Anthony V. Powell, Keith T. Butler, Ricardo Grau‐Crespo
Abstract
We provide here a summary of how machine learning techniques are being employed for the investigation of thermoelectric behaviour and for identifying new candidate thermoelectric materials. We show that while physics-based computational methods allow increasingly reliable prediction of the electron and phonon transport coefficients that determine the thermoelectric efficiency of materials, such methods are generally too expensive for high-throughput applications. Modern machine learning techniques, which make predictions based on existing data rather than on physical principles, can dramatically accelerate the computational design of thermoelectric materials. Using examples from recent literature, we provide an overview of the approaches that can be used for this purpose, identify the main trends and remaining challenges, and give our outlook for the future development of this field.