Generative Artificial Intelligence Navigated Development of Solvents for Next Generation High-Performance Magnesium...

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Traditional trial-and-error solvent discovery for magnesium (Mg) metal-based batteries is inefficient and costly.
• Next-generation Mg batteries require novel electrolytes for enhanced performance and stability.

Purpose of the Study:

• To establish an efficient screening criterion for novel Mg battery solvents by integrating artificial intelligence (AI) and a virtual molecular database.
• To revolutionize the traditional solvent design pathway for Mg battery electrolytes.

Main Methods:

• Generated 823 potential solvents using a self-developed algorithm.
• Utilized machine learning (ML) models to analyze physicochemical properties (LUMO, ΔLUMO, ESPmin, ESPmax, Eb) for screening.
• Experimentally validated two novel solvents, DOX and DMP, for Mg battery applications.

Main Results:

• Identified 18 candidate solvents from the initial pool of 823.
• DOX solvent, paired with Mg[B(hfip)4]2 salt, demonstrated significantly enhanced electrochemical performance in Mg batteries.
• Achieved 99.54% average coulombic efficiency for Mg plating/stripping over 5200 cycles at 1.0 mAcm-2.
• Mg//Cu cells reached a cumulative capacity exceeding 2000 mAhcm-2, surpassing previous benchmarks.

Conclusions:

• Developed a streamlined screening methodology for Mg battery solvents based on physicochemical properties, reducing the candidate pool by ~45-fold.
• The AI-driven approach significantly improves efficiency and can be extended to other battery systems.
• Validated novel solvents DOX and DMP, demonstrating their potential for high-performance Mg batteries.