Talk as part of a colloquium in Data-Driven Materials Discovery

Summary: Discovery in soft matter is limited by the cost of one quantity: the free energy that links a molecule's chemistry to its function. Permeation through a membrane, selective sorption in a polymer, and the assembly of a responsive material are the same quantity under different names, and computing it fast enough to search chemical space is the shared bottleneck. Coarse-grained simulation makes that search tractable, returning a free energy from a transferable force field, not a fitted score, and so holding beyond the chemistry it was built on. Compounds picked out this way have been assayed by experimental collaborators. Generative models carrying physics as an inductive bias rather than post-hoc validation now bridge resolutions and score candidates quantitatively. I will show how these parts assemble into a discovery engine for soft and bio-inspired matter