Title:
Bring Your Own Models: Catalysis from Scratch, Friendsgiving Style
Abstract:
Catalysis is central to many chemical production processes that supply society with fertilizers, medicines, fuels, and materials. Improving these processes through better catalysts is key to using resources more efficiently and reducing their environmental impact. Yet a computational description of catalytic behavior remains particularly challenging because working catalysts are complex and constantly changing: atoms rearrange, molecules compete for reactive sites, and reactions can proceed along multiple pathways.
Molecular simulations provide a microscopic foundation for understanding this complexity, linking the statistical properties of molecular configurations and trajectories to thermodynamic stability and reaction kinetics. In this talk, I will discuss how we combine generative machine learning with statistical mechanics to compute free energies, and how we extract coarse-grained dynamical models from simulation trajectories to identify reaction pathways and determine reaction rates.
