In a liquid, any constituent particle can rearrange and change its neighbors, but in crystalline solids, only particles near structural defects such as dislocations are prone to rearrange. A decades-old question in the field of supercooled liquids has been whether all particles are equally likely to rearrange, and if not, what are the “defects” that make some particles more likely to rearrange than others? I will show that machine learning provides a powerful and efficient approach for identifying local structural environments that are predisposed to rearrange. We use machine learning to define a quantity, “softness,” that is highly predictive of rearrangements and simplifies our understanding of glassy dynamics considerably.

Machine-learned softness as a tool for understanding glass formation and properties

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.