Large-Scale Dynamics Simulations of Complex Liquid Electrolytes with NequIP Equivariant Machine Learning

Electrolytes control efficiency, anode/cathode stability, battery recharge time as well as safety, thus their optimization is crucial for the design of next-generation energy storage devices. Here we focus on ionic liquid-based electrolytes thanks to their superior chemical stability compared to standard organic solvents, however, poor and anomalous transport properties are hindering their applicability [1]. Unfortunately, classical energy models struggle to provide the level of accuracy needed to reliably predict and investigate ionic conductivity.<br/>In this work, we study the applicability to such systems of state-of-the-art equivariant graph neural networks (NequIP [1]). Ionic liquid-based electrolytes provide a unique challenge due to their strong ionic interactions and viscous liquid nature. Additionally, substantially diverse inter-atomic environments are often present as a function of lithium-salt doping [2], raising the interesting question of model transferability. Our results show that we can obtain computational speed and near-DFT accuracy for large-scale ionic liquid molecular dynamics investigations.<br/>[1] Batzner, S., Musaelian, A., Sun, L., Geiger, M., Mailoa, J.P., Kornbluth, M., Molinari, N., Smidt, T.E. and Kozinsky, B., 2021. Se (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. arXiv preprint arXiv:2101.03164.<br/>[2] Molinari, N., Mailoa, J.P. and Kozinsky, B., 2019. General trend of a negative Li effective charge in ionic liquid electrolytes. The journal of physical chemistry letters, 10(10), pp.2313-2319.