Identification of long-lived clusters and their link to slow dynamics in a model glass former

TL;DR

This study links long-lived icosahedral clusters in a model glass to slow dynamics and heterogeneities, revealing their growth and influence on particle mobility as the system cools towards the glass transition.

Contribution

It introduces a novel cluster-based method to identify and analyze the stability of local structural motifs and their connection to slow dynamics in glass-forming liquids.

Findings

Long-lived icosahedral clusters correlate with slow regions.

Cluster domains grow and become more stable upon cooling.

Structural correlation lengths increase with cooling, but differently from dynamic length scales.

Abstract

We study the relationship between local structural ordering and dynamical heterogeneities in a model glass-forming liquid, the Wahnstrom mixture. A novel cluster-based approach is used to detect local energy minimum polyhedral clusters and local crystalline environments. A structure-specific time correlation function is then devised to determine their temporal stability. For our system, the lifetime correlation function for icosahedral clusters decays far slower than for those of similarly sized but topologically distinct clusters. Upon cooling, the icosahedra form domains of increasing size and their lifetime increases with the size of the domains. Furthermore, these long-lived domains lower the mobility of neighboring particles. These structured domains show correlations with the slow regions of the dynamical heterogeneities that form on cooling towards the glass transition. Although icosahedral clusters with a particular composition and arrangement of large and small particles are structural elements of the crystal, we find that most icosahedral clusters lack such order in composition and arrangement and thus local crystalline ordering makes only a limited contribution to this process. Finally, we characterize the spatial correlation of the domains of icosahedra by two structural correlation lengths and compare them with the four-point dynamic correlation length. All the length scales increase upon cooling, but in different ways.