| Abstract Scope |
Hydrogen-rich compounds provide a compelling route toward high-temperature superconductivity, but the pressures required for current record-Tc hydrides remain a major barrier to practical use. I will discuss our ongoing efforts to develop a physics-informed computational framework for identifying hydride superconductors with high Tc at reduced pressures. The approach integrates first-principles calculations, superconductivity theory, generative structure models such as CDVAE- and LLM-based workflows, and machine-learning models trained on literature and ab initio data to map composition–structure–property relationships, with emphasis on hydrogen sublattice motifs. Machine-learned interatomic potentials are used to accelerate structure relaxation and stability screening. Candidate materials are evaluated using electronic-structure calculations, phonon stability, electron–phonon coupling, and estimates of superconducting Tc, with ongoing extensions to transition-state searches and proton dynamics. This framework aims to accelerate discovery of experimentally realizable hydrides and clarify microscopic design principles governing high-Tc superconductivity. |