Unveiling redox potential behavior in electrolytes: A machine learning approach to Li-ion coordination effects

Determining the redox potential of electrolytes is crucial for exploring the electrochemical stability window in rechargeable batteries. However, the commonly used thermodynamic cycle method for calculating redox potentials does not account for the complex lithium-ion (Li⁺) coordination environment around redox-active sites. We conducted a systematic investigation of a 140 Li⁺-coordinated electrolytes library and discovered that the elevation in redox potential caused by Li⁺ is attributable to a combined effect of electrostatic and ion-dipole interactions. These two types of factors, however, can perform converse effects on the change of oxidation and reduction potentials, respectively. Among the various machine learning (ML) models examined, the extreme gradient boosting (XGB) algorithm demonstrates the best predictive accuracy for the redox potential of electrolytes, both with and without Li⁺ coordination. The integration of high-throughput calculation and ML creates a powerful platform for rapidly predicting electrolyte redox potentials, essential for accurately assessing practical electrochemical performance. Moreover, the current work highlights the role of electronic features in determining oxidation potential and structural features in influencing reduction potential, providing valuable insights for the design and application of new electrolytes. Copyright © 2025. Published by Elsevier Ltd.