Identification of the key parameters for the organic electrolytes selection for lithium-sulfur batteries

A key component for long-lasting next-generation lithium-sulfur batteries (LSBs) are electrolytes that facilitate a stable bilateral solid-electrolyte interface (SEI) simultaneously on both the cathode and anode. Other typical factors used for electrolyte selection include dielectric constant (ε), viscosity, dipole moment, donor number, and orbital energy levels. To simplify electrolyte selection, we developed a machine learning (ML) model (based on just 27 samples) whose predicted capacity retention generally agreed well with the experimental data. Additionally, it was found to correlate the most with the ε (R² = 0.71). High ε also implied a low |%drop| in capacity after rate testing. Therefore, ML was used to predict |%drop| for 51 batteries, which were also prepared and experimentally tested. Nine electrolytes with ε > 35, showed an average |%drop| of 0.8%, well below the average |%drop| of 5.29 % for the entire dataset. However, 3 other electrolytes with ε > 35 (based on diglyme) exhibited a |%drop| > 1.4%. The presence of large diglyme molecules in the solvation shell (obtained using classical molecular dynamics) leads to reduced diffusivity and a high |%drop|. Thus ε > 35 and the absence of large molecules in the solvation shell are good criteria for LSB electrolyte selection. Additionally, ML predicted |%drop| agreed well with experimental data (mean absolute error of < 2.59%).