Abstract

Geometric packing is an oft-used causal mechanism for structure formation acrossmany length scales – from the entropic ordering of colloidal nanoparticles to molecularco-crystallization of drug-like molecules. However, its exact role is hard-to-quantify,particularly in regimes where many competing forces can motivate nucleation, andevidently, crystallization. In this talk, I will first discuss the impact of geometric packing insystems where its effect should be most pronounced – hard, faceted nanoparticles thatself-assemble based on volume exclusion alone. Using Maxwell relations, I will showthat markers for “packing” behavior are absent in the regimes where self-assemblyoccurs, pointing to packing as a correlative, rather than causal, force in the emergenceof spontaneous order. I will then shift focus to crystallization in small-molecule systems,where the role of packing is still an open question and hard to pinpoint in analyses.Using physics-based machine learning representations and hybridsupervised-unsupervised models, I show how we can identify the role of enthalpic andgeometric components in stabilizing (or destabilizing) these systems. I will end with anoutlook on the future of physics-informed machine learning for understanding molecularpacking, including future work directions.This talk will pull from the following works:● Cersonsky, R. K., Pakhnova, M., Engel, E. A., Ceriotti, M., A data-driveninterpretation of the stability of organic molecular crystals. Chemical Science 14,1272–1285.● Helfrecht, B. A., Cersonsky, R. K., Fraux, G., Ceriotti, M., Structure-propertymaps with Kernel principal covariates regression. Machine Learning: Scienceand Technology 1, 045021.● Cersonsky, R. K., Van Anders, G., Dodd, P. M., Glotzer, S. C., Relevance ofpacking to colloidal self-assembly. Proceedings of the National Academy ofSciences● 115, 1439–1444