A public database of thermoelectric materials and system-identified material representation for data-driven discovery

This paper introduces a public dataset, ESTM, containing experimentally synthesized thermoelectric materials and their properties. Prediction models built on this dataset achieve high accuracy (R² > 0.9) in predicting thermoelectric properties from chemical compositions. A novel material descriptor, SIMD, improves extrapolation capabilities, significantly boosting prediction accuracy in unexplored material groups (R² from 0.13 to 0.71). SIMD enhances high-throughput screening, reducing false positives and aiding the discovery of high-performance thermoelectric materials.

Gyoung S. Na, Hyunju Chang