Lithium–sulfur (Li–S) batteries are among the most promising next-generation energy storage systems due to their high theoretical energy density, low cost, and the natural abundance of sulfur. However, several challenges hinder their practical realization, including the dissolution of lithium polysulfide (Li–PS) intermediates, their insulating nature, and significant volume changes at the cathode during cycling. In this work, we employ an integrated computational and data-driven approach to design optimal electrolytes that address these challenges. Focusing on fluorinated ether solvents (FLS) as co-solvents with 1,3-dioxolane, we explore a library of ~1,000 candidate molecules using high-throughput density functional theory (DFT), classical molecular dynamics (MD) simulations, machine learning (ML), and experimental validation
1. Remarkably, only 14 of these solvents have been previously reported for Li–S applications. Through systematic screening, we identify two novel FLS with reduced polysulfide solubility.
2, 3 One candidate matches the electrochemical performance of the benchmark TTE solvent, while offering improved viscosity and ionic conductivity. Interpretable ML models reveal that fluorination degree and the proximity of fluorine atoms to ether oxygen play critical roles in tuning both oxidative stability and polysulfide solubility. This study not only introduces new high-performance co-solvents but also establishes a robust computational-experimental framework for rational electrolyte design in Li–S batteries.
Reference:
- Atwi, R.; Bliss, M.; Makeev, M.; Rajput, N. N., MISPR: an open-source package for high-throughput multiscale molecular simulations. Scientific Reports 2022, 12 (1), 15760.
- Atwi, R.; Rajput, N. N., Guiding maps of solvents for lithium-sulfur batteries via a computational data-driven approach. Patterns 2023, 4 (9).
- Rasha Atwi, D. G., Dan Thien Nguyen, Minyung Song, Agrim Babbar, Vijayakumar murugesan, Vilas Pol, Nav Nidhi Rajput, Knowledge-driven design of fluorinated ether electrolytes via a multi-model approach. (under review) 2025.
