Presentation Information
[9p-S301-6]Exploration of High-Pressure Superconducting Materials in Ternary Hydrides Using Machine Learning
〇Souta Miyamoto1, Kazuaki Tokuyama1, Taichi Masuda1, Katsuaki Tanabe1 (1.Kyoto Univ.)
Keywords:
superconductor,hydrides,materials informatics
We explored superconducting materials by applying machine-learning techniques to ternary hydrides. A dataset comprising roughly 2,000 records of binary and ternary hydrides was compiled, from which we trained an ensemble of 30 XGBoost models. Using the 95 % lower confidence bound, we ranked potential high-temperature superconductors at 100, 200, and 300 GPa, identifying promising systems outside the training data, such as Ca–Ti–H and Na–Mg–H. Preliminary validation with density-functional theory calculations is currently underway.