Unsupervised learning of ion dynamics in electrolytes using graph dynamical networks

Understanding the dynamical processes that govern the performance of functional materials is essential for the design of next generation materials to tackle global energy and environmental challenges. Many of these processes involve the dynamics of individual atoms, ions, or small molecules in condensed phases, which are difficult to understand due to the complexity of local environments. Here we present graph dynamical networks [1], an unsupervised learning approach for atomic scale dynamics in arbitrary phases and environments from molecular dynamics simulations. We apply the methodology to ion dynamics in polymer electrolytes, and show how important features related to ion-ion correlations can be automatically captured. With the large amounts of molecular dynamics data generated every day in nearly every aspect of materials design, this approach provides a broadly applicable, automated tool to understand atomic scale dynamics in material systems.

[1] T. Xie, A. France-Lanord, Y. Wang, Y. Shao-Horn, and J. C. Grossman. Graph dynamical networks for unsupervised learning of atomic scale dynamics in materials, Nature communications 10, 2667 (2019)