【Frontiers of HaiChuan Chemical Technology】Amazing! AI directly generates complete battery electrolyte formulas

Thread Content

The 25th National \"Safety Production Month\" in 2026: Everyone talks about safety, and everyone knows how to handle emergencies; identifying and addressing risks and hazards. -------------------------------------------------- According to the latest issue of the journal JACS Au, a research team from the Pritzker School of Molecular Engineering at the University of Chicago has developed a generative AI system that can create complete battery electrolyte formulations directly, rather than merely identifying individual candidate molecules. This AI model is named ElectrolyteGPT; it can not only determine which chemical components should be used in the electrolyte, but also design parameters such as the concentration of each component and their mixing ratios, in order to find the optimal balance among various performance indicators such as ionic conductivity, oxidation stability, Coulombic efficiency, and viscosity. The team used this system to generate a series of entirely new electrolyte formulations and conducted experimental verification. The results showed that the performance of several of these formulations in lithium metal batteries reached the level of current advanced electrolytes. Battery electrolyte design faces the enormous challenge of a \"combination explosion\": the number of potential electrolyte molecules far exceeds the total number of stars in the universe, and when considering all possible combinations of these molecules, the possibilities are virtually infinite. Traditional experimental methods struggle to find the optimal solutions in such a vast chemical space, whereas generative AI can explore areas that have never been studied before and propose entirely new candidate formulations. However, most existing generative AI models were originally designed for drug development, and what they generate are often molecules suitable as drugs. To this end, the team developed a database of electrolyte-related compounds and trained the model specifically so that it could learn the characteristics of electrolyte materials and generate new molecules that meet the required specifications. A key innovation of this research is the development of a new chemical description language called “fLine”. The widely used SMILES language is primarily employed to describe molecular structures, whereas fLine can not only record the chemical structure but also include formulation details such as solvent ratios, salt concentrations, and temperature. With this new language, AI can design the entire electrolyte system as a whole, rather than focusing only on individual molecules. This method is not only applicable to the development of electrolytes but also holds promise for being extended to the design of other complex chemical mixtures. This achievement represents an important step toward \"fully generative electrolyte design\". In the future, as training data and model sizes continue to grow, AI is expected to discover new electrolytes with performance surpassing that of current best products, thereby accelerating the development of next-generation high-performance batteries.