A research team led by Professor Park Jin-woo from Gachon University's School of Chemical, Biological, and Battery Engineering has jointly developed an electrolyte design technology utilizing artificial intelligence (AI) with a research team led by Professor Kim Woong from Korea University's School of Materials Science and Engineering.
According to Gachon University on the 26th, this research represents an advancement in lithium metal battery development technology by establishing a foundation for rapidly identifying optimal electrolytes.
The research findings were recently published online in Energy Storage Materials (IF=20.2), an international academic journal in the energy storage field.
Lithium metal batteries are attracting attention as a core technology for electric vehicles and next-generation energy storage devices due to their ability to achieve high energy density. However, irregular lithium growth during charge-discharge cycles can reduce lifespan and safety, necessitating precise electrolyte design for improvement. Conventional methods required significant time and cost due to repeated verification of numerous candidate combinations.
The research team introduced AI-based data analysis techniques to overcome these limitations. They developed and applied a new molecular representation technology (e-ECFP) that simultaneously reflects molecular connection structures, repetition characteristics, and concentration information, establishing a system that learns both the three-dimensional structure and concentration information of electrolyte molecules. Furthermore, they implemented an "interpretable AI model" that explains performance improvement factors beyond mere performance prediction, and enhanced model reliability by training it to distinguish between the roles of salts and solvents.
Analysis of electrolyte candidates using this model revealed that fluorine-containing structures and cyclic ether structures play crucial roles in improving the Coulombic efficiency of lithium metal batteries. A newly designed electrolyte based on these findings recorded 99.72% Coulombic efficiency in lithium-copper half-cell experiments and demonstrated stable capacity retention characteristics after more than 500 charge-discharge cycles in lithium-lithium iron phosphate full-cell experiments.
The research team stated, "This research demonstrates that AI can be utilized not just for simple predictions but for identifying the causes of battery performance improvements," adding, "We expect this to be widely applied to the design of various next-generation energy storage systems in the future."
This research was conducted with support from Korea University.