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

Abstract

Thermoelectric materials have received much attention as energy harvesting devices and power generators. However, discovering novel high-performance thermoelectric materials is challenging due to the structural diversity and complexity of the thermoelectric materials containing alloys and dopants. For the efficient data-driven discovery of novel thermoelectric materials, we constructed a public dataset that contains experimentally synthesized thermoelectric materials and their experimental thermoelectric properties. For the collected dataset, we were able to construct prediction models that achieved R2-scores greater than 0.9 in the regression problems to predict the experimentally measured thermoelectric properties from the chemical compositions of the materials. Furthermore, we devised a material descriptor for the chemical compositions of the materials to improve the extrapolation capabilities of machine learning methods. Based on transfer learning with the proposed material descriptor, we significantly improved the R2-score from 0.13 to 0.71 in predicting experimental ZTs of the materials from completely unexplored material groups.

Artificial Intelligence for Learning Material Synthesis Processes of Thermoelectric Materials

“…One of the fundamental goals in thermoelectric materials is to discover novel materials of high ZT . To this end, various calculation and machine learning methods have been proposed to predict ZTs of the materials from their chemical compositions. ,, Although existing methods successfully predicted ZTs from the chemical compositions of the thermoelectric materials, they overlooked the effect of the synthesis process on the thermoelectric properties of the materials. Various experimental observations showed that the energy conversion efficiency changes by different synthesis processes even if product materials have the same chemical composition. , …”

“…For data-driven research on material synthesis, we constructed a benchmark dataset that contains 771 unique thermoelectric materials and their synthesis processes by extracting experimentally validated synthesis recipes from the scientific literature in the ESTM dataset. 35 We collected the materials synthesis recipes of the thermoelectric materials based on ChemDataExtractor 26 with human supervision. In this paper, we refer to the collected dataset as the thermoelectric materials synthesis recipes (TMSR) dataset.…”

Interpretable Machine Learning Workflow for Evaluating and Analyzing the Performance of High-Entropy GeTe-Based Thermoelectric Materials

“…(3) Rapidly hot-pressing and sintering the obtained powder into samples. In this study, a dataset of GeTe-based thermoelectric materials, as reported in previous literature, 33 was selected. This dataset consists of 458 samples, with each sample comprising the chemical formula, temperature, and ZT value.…”

Machine learning based feature engineering for thermoelectric materials by design

“…19 Along the lines of StarryData2, Na and Chang constructed a dataset containing 5205 chemical compositions of the experimentally synthesized thermoelectric materials and their experimental thermoelectric properties. 20 All these approaches rely on manual or semi-manual extraction from the literature. Sierepeklis and Cole used a combination of web-scrapping and natural language processing to develop the first automatically generated database of thermoelectric materials and their properties from the existing literature, containing 22 805 data records, automatically generated from the scientific literature, spanning 10 641 unique extracted chemical names.…”