Physical and Theoretical Chemistry

Machine-Learned Potentials and Data-Driven Simulation at Scale

Additional sessions will consider foundation models for atomistic simulation, delta-learning against classical force fields, and reactive potentials for bond-breaking and transition-metal chemistry. Presentations may cover phonon and thermal transport prediction, mechanical property calculation, and the benchmarking of learned potentials against standard test sets. Discussion will extend to uncertainty propagation in downstream calculations, model provenance, and the reproducibility of training pipelines, as well as community benchmark challenges. Software and hardware ecosystems, including GPU acceleration and mainstream simulation packages, will also be addressed. The integration of machine-learned potentials with high-throughput screening and autonomous platforms will be emphasized, including workflows that feed simulation results back into experimental design for catalysis, energy storage and sustainable materials development.