Phonons at Scale: High-Throughput Lattice Dynamics for Data-Driven Materials Discovery

Phonons govern many of the most important phenomena in condensed-matter physics, including superconductivity, ferroelectricity, and lattice thermal transport. In this talk, I will present a new high-throughput ab initio database of harmonic phonon properties for over 27,000 inorganic crystals (at the time of abstract submission), computed using automated workflows built within the atomate2 and pheasy frameworks. By coupling density-functional theory with a LASSO-regularized compressive-sensing approach for force-constant extraction, we achieve order-of-magnitude speedups over conventional finite-displacement or DFPT methods while maintaining excellent agreement with benchmark DFPT datasets.

I will discuss both the technical advances (robust enforcement of symmetry and sum-rule constraints, cross-validation protocols for interatomic force-constant fitting, and convergence strategies across low-symmetry lattices) as well as insights from large-scale statistical analyses of phonon trends, dynamical stability, and chemical diversity spanning 86 elements. The resulting database, now integrated into the Materials Project, provides phonon band structures, densities of states, and thermodynamic quantities for public access. Beyond serving as a resource for screening and phase-stability modeling, it establishes a high-fidelity reference for benchmarking machine-learned interatomic potentials and developing next-generation phonon prediction models.