ľ¹Ï¸£ÀûÓ°ÊÓ

Title: Autonomously revealing hidden local structures in supercooled liquids

Abstract: Unraveling the interplay between structure and dynamics in glassy materials is a major challenge in condensed matter science. When a liquid is rapidly cooled down or compressed to the point where it almost turns into a glass, its dynamics slow down by several orders of magnitude, while its structure typically stays largely unchanged. The dynamics of such glassy fluids are heterogeneous, with some regions of particles rearranging much more rapidly than others. Key to understanding this phenomenon is identifying structural characteristics that are associated with these heterogeneities. To explore this link, much research has been devoted to pinpointing local structures and order parameters that correlate strongly with dynamics. Here I will present a new unsupervised machine learning algorithm that can autonomously reveal structural heterogeneities, and show its application to three archetypical glass formers: hard spheres, Wahnström, and Kob-Andersen. Despite being purely based on structure, the structural heterogeneities identified by the algorithm turn out to be  strongly connected to the dynamical heterogeneities, and provide a new way forward for unraveling the structural origins of the slow dynamics of glassy materials.

Based on published paper: E. Boattini, S. Marín-Aguilar, S. Mitra, G. Foffi, F. Smallenburg and L. Filion, Autonomously revealing hidden local structures in supercooled liquids, Nat. Commun. 11, 5479 (2020)