Disentangling glassy polymer dynamics: combining simulations and machine learning

Glass-forming liquids are an example of non-equilibrium systems: upon cooling, a glass undergoes a transition from a liquid into an amorphous solid without apparent long-range order. The problem of the glass transition combines the concepts of self-organization, collective and heterogenous dynamics and poses an unsolved fundamental problem in condensed matter physics. We investigate the role of connectivity and topology in the glass transition of polymers, combining coarse grained molecular dynamics simulations with machine learning. ML allows us to quantify the relation between local structural properties of polymers and their topology using clustering of correlated motions of polymer chain segments as an example descriptor to create a training set, leading to the development of a data-based model.