Structure and dynamics in glass-formers: predictability at large length scales

TL;DR

This study investigates the relationship between static structure and dynamic heterogeneity in glass-formers, finding that particle-level predictability is limited but larger-scale correlations can be predicted from structural propensity.

Contribution

The paper introduces new statistical indicators and demonstrates that static structure predicts spatial correlations of dynamics at larger length scales, not at the individual particle level.

Findings

Particle-level dynamics cannot be predicted from static structure.

Spatial correlations of dynamic heterogeneity are predictable from the propensity field.

Static-dynamic connection exists at larger length scales, not at the particle level.

Abstract

Dynamic heterogeneity in glass-formers has been related to their static structure using the concept of dynamic propensity. We re-examine this relationship by analyzing dynamical fluctuations in two atomistic glass-formers and two theoretical models. We introduce quantitative statistical indicators which show that the dynamics of individual particles cannot be predicted on the basis of the propensity, nor by any structural indicator. However, the spatial structure of the propensity field does have predictive power for the spatial correlations associated with dynamic heterogeneity. Our results suggest that the quest for a connection between static and dynamic properties of glass-formers at the particle level is vain, but they demonstrate that such connection does exist on larger length scales.