Insights from Community Detection and Clustering through Modularity Function in 3D Model Glasses: an Unsupervised Machine Learning Approach
Please login to view abstract download link
Glass formation, particularly the structural signature of it, is a longstanding unresolved question in material science. There have been research investigations to study the atomic-scale structural rearrangement of glasses in response to an externally applied force, particularly for complex systems like silicate glasses. The current study reports the first application of a recently developed nature-inspired modularity (NIM) function to study the optimal clustering in the model silicate glasses. This NIM function, which in essence is an unsupervised machine learning protocol, is chosen since it was demonstrated to outperform other competing methods in other domains. The NIM function employs a modified Potts model-based community detection technique in which a modularity function is maximized. In this study, the modularity function is calculated in a higher dimensional space consisting of spatial coordinates as well as chemical identities of the molecules and their relative strength of interaction. Different glass systems were prepared using athermal quasistatic shear (AQS) deformation protocols. The best partition, which is nothing but the most modular structure, of a given network helps in revealing its naturally identifiable structures. This study seeks to explore the evolution of the best partition during the AQS deformation processes, which is likely to reveal insights into the mechanical properties of these glass systems.
