Structural entropy of glassy systems from graph isomorphism.

Abstract

Configurational entropy plays a central role in thermodynamic scenarios of the glass transition. As a measure of the number of basins in the potential energy landscape, configurational entropy for a glass-forming liquid can be evaluated by explicitly counting distinct inherent structures. In this work, we propose a graph-theory based method to examine local structure and obtain the corresponding entropy of hard-particle systems. Voronoi diagrams of associated clusters are classified using a graph isomorphism algorithm. The statistics of these clusters reveal structural motifs such as icosahedron-like order, and also allow us to calculate the structural entropy SG. We find the structural entropy of an n-particle subsystem grows linearly with n. Thus the structural entropy per particle can be obtained from the slope dSG/dn. Our results are consistent with previous values for configurational entropy obtained via thermodynamic integration. Structural entropies per particle are measured for hard-disk and hard-sphere polydisperse systems, and extrapolated for monodisperse hard disks, all of which are nonzero at the dynamic glass transition.