Uncovering ion transport mechanisms in ionic liquids using data science

Abstract

Batteries play a key role in the energy transition but suffer from safety concerns arising from the electrochemical instability of organic electrolytes. Ionic liquids are emerging as promising, non-flammable electrolytes for next-generation batteries. Yet, designing ionic liquids to facilitate redox ion transport has proven challenging, because ionic liquids are concentrated electrolytes where ion–ion interactions cause pronounced deviation from classical electrolyte scaling theories which assume viscosity governs mobility. Machine learning studies show that ionic liquid transport properties are challenging to predict from molecular descriptors, preventing rational design. Here, we pursue a broader data-centric approach to provide insight into ionic liquid design by merging databases of experimental properties and computational molecular features for 218 ionic liquids across 127 publications. We find that ionic liquids are well-described by a modified Arrhenius model that captures structure-driven ion transport in correlated electrolytes, yielding energy barriers of around 20–30 kJ mol⁻¹. This exhibits remarkable agreement with the approximately 25 kJ mol⁻¹ screened ion pair interaction energy derived from surface forces measurements, suggesting links between mechanisms of ion transport and interfacial screening. We also use machine learning models to find that molecular features can predict some properties, such as density, while failing to predict properties that rely on long-range correlations, such as viscous dissipation. Our study reveals that data science tools can be leveraged to reveal non-classical transport scaling relationships and alternative materials descriptors that promise to be transformative for designing ionic liquids and other correlated electrolytes for next-generation batteries. All data and models are shared as open-source code.