Accelerating the exploration of material space for hydride double perovskite superconductors under ambient pressure through machine learning.

In the field of hydride superconductors, a great challenge is to achieve superconducting states under ambient pressure conditions rather than the extreme high-pressure environments that have been required in experiments. Achieving this goal is crucial for advancing the practical applications of high-temperature superconducting materials. We discover a family of compounds (hydride double perovskite superconductors with space group Fm3̄m and chemical formula A2MM'H6) to achieves this goal. A machine-learning-accelerated approach is utilized to search for hydride double perovskite superconductors under ambient pressure within an extensive dataset comprising over 106 535 hypothetical compounds. 15 stable hydride double perovskite superconductors are discovered under ambient pressure, with the highest superconducting transition temperature (Tc) reaching 18.7 K. The structural stability, electronic properties, and superconducting behavior of these materials have been comprehensively analyzed. Phonon dispersion analysis has highlighted the critical role of lattice vibrations in electron-phonon coupling (EPC), where the contribution of H atom vibrations is essential for facilitating electron pairing and the onset of superconductivity. This demonstrates that the machine-learning-accelerated approach is a highly effective method and can be easily extended to other compounds.