| Abstract Scope |
Chemically disordered materials offer new opportunities for energy technologies, but their vast configurational complexity makes predictive design difficult. We develop data-driven, first-principles workflows to model, validate, and discover complex energy materials. These workflows use machine-learned interatomic potentials to model configurational disorder efficiently and descriptor-rich databases to connect composition, local structure, and synthesizability. We also benchmark electronic fidelity across oxides, iodides, and alloys, showing that structural accuracy alone does not guarantee reliable electronic properties. To better connect computation with experiment, we calculate Debye-Waller factors for more realistic finite-temperature diffraction references. Together, these efforts link machine learning, first-principles modeling, and structure-informed validation, with applications spanning high-entropy oxide discovery, thermochemical hydrogen production, and finite-temperature structural characterization. |