Quantifying Local Point-Group-Symmetry Order in Complex Particle Systems
Domagoj Fijan, Maria R. Ward Rashidi, Sharon C. Glotzer
TL;DR
The paper addresses the challenge of directly quantifying local point-group symmetry during crystallization by introducing $PGOP$, a continuous, bounded order parameter computed from Gaussian representations of neighboring particles and overlaps under symmetry operations. It implements $PGOP$ in the open-source SPATULA toolkit and demonstrates superior sensitivity to local symmetry across simple and complex crystals and Lennard–Jones nucleation, outperforming conventional bond-orientational order parameters. The framework supports extensions to other symmetry operations and to local BOODs, enabling interpretable, per-point symmetry metrics in both disordered and crystalline states. It also outlines pathways toward differentiable formulations and use as a collective variable for enhanced sampling in future work.
Abstract
Crystals and other condensed phases are defined primarily by their inherent symmetries, which play a crucial role in dictating their structural properties. In crystallization studies, local order parameters (OPs) that describe bond orientational order are widely employed to investigate crystal formation. Despite their utility, these traditional metrics do not directly quantify symmetry, an important aspect for understanding the development of order during crystallization. To address this gap, we introduce a new set of OPs, called Point Group Order Parameters (PGOPs), designed to continuously quantify point group symmetry order. We demonstrate the strength and utility of PGOP in detecting order across different crystalline systems and compare its performance to commonly used bond-orientational order metrics. PGOP calculations for all non-infinite point groups are implemented in the open-source package SPATULA (Symmetry Pattern Analysis Toolkit for Understanding Local Arrangements), written in parallelized C++ with a Python interface. The code is publicly available on GitHub.
