Accelerated Screening of Electrolyte Solvents for High Safety Batteries Using Machine Learning

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Current commercial batteries utilize flammable carbonate-based solvents.
• Safer solvent alternatives are crucial for battery safety and performance.

Purpose of the Study:

• To develop a high-throughput screening approach for identifying safer nonaqueous battery solvents.
• To predict key solvent properties directly from molecular structure.

Main Methods:

• Trained graph neural networks using the DIPPR 801 database.
• Predicted flash point, melting point, dielectric constant, and liquid viscosity.
• Screened over one billion molecules from Mcule and GDB-13 databases.

Main Results:

• Identified thousands of dynamically stable and easily synthesizable solvent candidates.
• Discovered novel candidates, including food additives and fragrances, for battery applications.
• Found one known battery additive among the promising candidates.

Conclusions:

• The high-throughput screening approach is effective for discovering safer battery solvent alternatives.
• Diverse and readily available candidates were identified through a dual database screening strategy.
• This method opens new avenues for developing next-generation battery electrolytes.