Learning the Interatomic Potential of Ionic Liquids and Salt-In-Ionic Liquids with Machine Learning

Ionic liquids are a promising class of electrolytes for battery and supercapacitor applications, as well as universal solvents for chemical reactions. In their pure form, these two-component liquids have complex molecular environments and their disordered structures can have large length scales. Moreover, how these electrolytes behave at interfaces, how they transport ions and how additives, such as Li-based salts or/and water, affect their properties is challenging to study using empirical interatomic potentials due to their limited accuracy. Here we develop machine learning interatomic potentials, based on the equivariant graph neural networks with NequIP [1], for representative ionic liquid and salt-in-ionic liquid [2]. As capturing the complex intermolecular interactions is subtle, training a potential for this system is not as straightforward as a simple solid-electrolyte, for example, and we highlight some of the challenges encountered with this process. In addition, we study the question of model transferability across the composition space of salt-in-ionic liquids, the effect of long-range interactions and uncertainty of the model.<br/><br/>[1] Batzner, S., Musaelian, A., Sun, L., Geiger, M., Mailoa, J.P., Kornbluth, M., Molinari, N., Smidt, T.E. and Kozinsky, B. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nat. Commun., 13, 1-11 (2022)<br/>[2] Molinari, N., Mailoa, J.P. and Kozinsky, B. General trend of a negative Li effective charge in ionic liquid electrolytes. J Phys. Chem. Lett., 10, 2313-2319 (2019)