Accelerating materials property discovery in uncharted domains through the integration of high-throughput computation and machine learning

Abstract

An integrated approach that combines high-throughput (HT) computations with machine learning (ML) is proposed to accelerate the discovery of novel materials with optimized electrical conductivity, carrier mobility, and thermal conductivity. By systematically varying crystal structures, temperature conditions, and energy parameters, extensive datasets were generated and analyzed to extract meaningful features that establish structure–property relationships. Feature engineering and selection enabled the development of accurate ML models, facilitating efficient material screening and prediction without the need for extensive domain knowledge. The proposed framework automates key processes such as data preprocessing, feature extraction, and model training, ensuring scalability and reproducibility. Model validation against experimental data demonstrates the reliability of the predictions, while iterative improvements further enhance accuracy. This data-driven strategy offers a powerful tool for advancing materials discovery in diverse applications, including energy storage, electronics, and thermal management, providing a foundation for future innovations in materials science.