Abstract
The relation between structure and dynamics in glasses is not fully understood. A new approach based on machine learning now reveals a correlation between softness—a structural property—and glassy dynamics. In contrast with crystallization, there is no noticeable structural change at the glass transition. Characteristic features of glassy dynamics that appear below an onset temperature, T0 (refs 1,2,3), are qualitatively captured by mean field theory4,5,6, which assumes uniform local structure. Studies of more realistic systems have found only weak correlations between structure and dynamics7,8,9,10,11. This raises the question: is structure important to glassy dynamics in three dimensions? We answer this question affirmatively, using machine learning to identify a new field, ‘softness’ which characterizes local structure and is strongly correlated with dynamics. We find that the onset of glassy dynamics at T0 corresponds to the onset of correlations between softness (that is, structure) and dynamics. Moreover, we construct a simple model of relaxation that agrees well with our simulation results, showing that a theory of the evolution of softness in time would constitute a theory of glassy dynamics.
Highlights
When a liquid freezes, a change in the local atomic structure marks the transition to the crystal
To look for correlations between structure and dynamics, one typically tries to find a quantity that encapsulates the important physics, such as free volume, bond orientational order, locally preferred structure, etc. In contrast to this approach, we use a machine learning method designed to find a structural quantity that is strongly correlated with dynamics
We describe a particle’s local structural environment with M = 166 “structure functions” [13] that respect the overall isotropic symmetry of the system and include radial density and bond angle information
Summary
A change in the local atomic structure marks the transition to the crystal. To look for correlations between structure and dynamics, one typically tries to find a quantity that encapsulates the important physics, such as free volume, bond orientational order, locally preferred structure, etc In contrast to this approach, we use a machine learning method designed to find a structural quantity that is strongly correlated with dynamics. We can deduce the most important structural features contributing to softness either by training on fewer structure functions or by examining the projection of the hyperplane normal onto each orthogonal structure function axis Both analyses yield a consistent picture (see supplementary information): the most important features are the density of neighbors at the first peaks of the radial distribution functions gAA(r) and gAB(r); these two features alone give 77% prediction accuracy for rearrangements. Soft particles typically have a structure that is more similar to a higher-temperature liquid, where there are more rearrangements, while hard particles whose structure appears closer to a lower-temperature liquid [18]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
