Towards tailored thermoelectric materials: An artificial intelligence-powered approach to material design

Abstract

The pursuit of novel thermoelectric (TE) materials with exceptional properties stands as a critical frontier in material science. This transformative endeavor demands a paradigm shift in materials discovery, necessitating the integration of cutting-edge computational approaches. In this groundbreaking study, we employ the power of artificial neural networks (ANNs) to accelerate the identification of promising TE candidates. By meticulously training an ANN model on a comprehensive dataset of TE compounds and their corresponding properties, we have established a powerful tool capable of accurately predicting atoms that can be substituted into a TE material's structure to achieve desired TE properties. Our findings, validated the Materials Research Laboratory (MRL) dataset, demonstrate the exceptional accuracy of the ANN model, with an r2 detection coefficient of 0.97 in the best architectural case. This underscores the transformative potential of ANNs in propelling materials discovery forward, offering a promising avenue for the development of next-generation TE devices.