Scientists use machine learning to generate electrolyte formulations for next-gen batteries

University of Chicago researchers expanded ElectrolyteGPT to generate complete battery electrolyte formulations, not just screen individual ingredients.

The model was trained on a curated battery-focused dataset and proposed new candidates with concentrations and mixing ratios.

Several AI-designed electrolytes were synthesized and tested in the lab, matching the performance of top lithium metal battery electrolytes.

The work, from the Amanchukwu Lab and published in JACS Au, points to faster discovery across a vast chemical design space.