Identifying the impact of chemical functional groups on ionic liquid conductivity.

Ionic liquids are non-flammable, electrochemically stable electrolytes with promise as next-generation battery electrolytes. However, strong ion correlations and deviations from classical ion transport prevent the predictive design of application-specific ionic liquid electrolytes. While machine learning models can help address such limitations, standard cheminformatics tools are not well-suited to incorporate electrostatic interactions and thus do not account for the strong intermolecular interactions within ionic liquid materials. In this work, we present a molecular fragment representation that reflects charge-carrier resonance using SMARTS molecular substructure searching to explicitly capture the electrostatic contributions to conductivity of ionic liquid compounds. We find that this representation simplifies structure-conductivity relationships and improves predictive performance in low-data regimes. Additionally, we report how learned contributions of molecular structure impact ion transport and map the ionic liquid chemical design space using principal covariates regression to visualize the underlying structure-property relationships. We find that charge delocalization and anion flexibility are the most influential molecular features in improving ion transport in single-component ionic liquids. Further, we find that polar, alkyl, and fluorinated cation or anion functionalization decrease ion transport by introducing additional intermolecular interactions. Finally, many cation charge centers similarly enhance ion transport, suggesting that ionic liquid anions and ion functionalization are key for tuning intermolecular interactions and ion mobility. This work provides an interpretable approach for building structure-property relationships for currently available ionic liquid structures and opens new perspectives on how to design ionic liquids for desirable transport behavior. All data and models are shared as open-source code.