A digital solvent screening method for lithium–sulfur batteries using a partial least squares model based on cycling data and Conductor-like Screening Model for Real Solvents calculations is presented. Screening≈3,000 compounds, promising electrolyte candidates are identified. Experimental validation confirms the model's effectiveness, with pimelonitrile showing exceptional cycling stability, demonstrating the potential of data-driven approaches in optimizing battery electrolyte performance.

Lithium–sulfur batteries have attracted great research interest due to the high theoretical capacity of sulfur of 1672 mAh g⁻¹. However, they have various problems due to the shuttle current caused by molecular sulfur dissolving in the electrolyte. Hence, electrolyte design is a key focus when optimizing the batteries. This study investigates the relationship between cycling data and electrochemical properties measured with cyclovoltammetric measurements, shuttle current measurements, and impedance spectroscopy. Using the acquired data, a partial least squares model to screen solvent candidates in reference to these findings is introduced. This model is based on cycling data as well as density functional theory-calculated Conductor-like Screening Model for Real Solvents data of the solvents and (solvated) lithium–polysulfides. The usefulness of the converged method is demonstrated by using it to identify new possible electrolyte systems. A subset of ten selected electrolyte systems is evaluated experimentally and their performance is reported. One of those electrolytes, 1.4 M LiTFSI, in pimelonitrile solution and without any further additives, displays exceptional cycling stability already on the first attempt, reaching a state of health of 50% after 115 cycles and maintaining a Coulombic efficiency of close to 100% during the entire cycling procedure.