2021/02/27 by Peng Song, Zhufeng Hou, Song, Peng +11
Engineering · Materials Science · Physics and Astronomy · #Applied Physics (physics.app-ph) #Computational Physics (physics.comp-ph) #Data Analysis #FOS: Physical sciences #Machine Learning in Materials Science #Rare-earth and actinide compounds #Statistics and Probability (physics.data-an) #Superconducting Materials and Applications #Superconductivity (cond-mat.supr-con)
paper · pdf · doi:10.48550/arxiv.2103.00193
openalex publication_date 2021/02/27 · openalex created_date 2021/03/15 · openalex updated_date 2026/07/28
The search for hydride compounds that exhibit high Tc superconductivity has been extensively studied. Within the range of binary hydride compounds, the studies have been developed well including data-driven searches as a topic of interest. Toward the search for the ternary systems, the number of possible combinations grows rapidly, and hence the power of data-driven search gets more prominent. In this study, we constructed various regression models to predict Tc for ternary hydride compounds and found the extreme gradient boosting (XGBoost) regression giving the best performance. The best performed regression predicts new promising candidates realizing higher Tc, for which we further identified their possible crystal structures. Confirming their lattice and thermodynamical stabilities, we finally predicted new ternary hydride superconductors, YKH12 [C2/m (No.12), Tc=143.2 K at 240 GPa] and LaKH12 [R3m (No.166), Tc=99.2 K at 140 GPa] from first principles.