Small particle dynamics in glassy polymers: diffusion, relaxation, and machine-learned softness

In this work, we explore the simulated transport dynamics of small particles through a polymer melt at temperatures spanning from liquid-like behavior down to near and below the simulated glass transition temperature. Using softness, a machine-learned scalar quantity, we connect the structural neighborhood surrounding a particle with its dynamic behavior to relate the probability of a glassy rearrangement to its local environment. An energetic and entropic scale for the rearrangement process of these small particles emerges and is compared across systems of different particle sizes and interaction strengths. Diffusion coefficients and relaxation times both show strong dependence on our tuning parameters, and the barriers to rearrangement show increasing nonlinearity as the particles become smaller. The trends we observe provide some insight into how local structure plays a role in small molecule transport when the surrounding medium is undergoing a glass transition, leading to large changes in system mobility.