Unlocking the Chemical Space for Rechargeable Batteries with a Generative Solvent Design System

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Current electrolyte discovery for rechargeable batteries is limited by heuristic trial-and-error and high-throughput screening.
• Developing new electrolytes is crucial for advancing battery performance and safety.

Purpose of the Study:

• To introduce a novel Generative Solvent Design System (GSDS) for de novo design of rechargeable battery solvents.
• To leverage deep molecular generation and machine learning for efficient electrolyte solvent discovery.

Main Methods:

• Constructed a large, battery-specific molecular dataset (Batt-SLM).
• Fine-tuned a graph-based molecular generator using physics-informed machine learning surrogates for key properties (redox potential, viscosity, melting point, etc.).
• Integrated property predictors with a generative model to design novel solvents.

Main Results:

• GSDS successfully rediscovered known fluorinated and phosphorus-containing compounds.
• The system proposed novel, application-specific solvent candidates for alkali metal batteries.
• Candidates passed rigorous verification including property evaluation, synthetic accessibility, and literature checks.

Conclusions:

• GSDS provides a viable solvent design component for broader electrolyte design frameworks.
• The system can be extended for advanced optimization, including salt-aware, interface-informed, and mixture designs.
• This AI-driven approach accelerates the development of next-generation rechargeable batteries.