A Machine Learning Approach for Describing Shear-induced Dynamics in Soft Particle Glasses

Soft particle glasses (SPGs) are deformable particles jammed at volume fractions beyond the random close packing of equivalent hard spheres. These athermal suspensions exhibit weak solid-like behavior at rest and a liquid-like flow above yield stress. Due to many factors affecting their properties, such as solvent viscosity, particle elasticity, and volume fraction, SPGs display fascinating shear-induced heterogeneous dynamics due to their ability to store and release internal stresses. In this study, we seek to understand the correlation between the structure and dynamics of SPGs using predictive machine learning (ML)-based methods. We employ linear regression and neural network models along with per-particle descriptors to predict the propensity of the particles to undergo heterogeneous dynamics. The accuracy of our predictions will be validated by comparing them directly to results obtained from three-dimensional particle dynamic simulations. ML model in conjunction with 3D simulations will be used to build relationships between the microstructure and localized yielding events in SPGs.