Vitaliy Yurkiv builds better batteries with $578K NSF grant

Lithium-ion batteries power everything from cellphones and electric vehicles to utility-scale storage facilities and hospital backup generators. But their unstable composition makes them susceptible to rare but dangerous fires and, in some cases, explosions.

Vitaliy Yurkiv, assistant professor of aerospace and mechanical engineering and an expert in the thermal-electrochemical characterization of batteries, won a three-year, $578,000 grant from the National Science Foundation to identify safer, more powerful battery materials. 

Since 2011, the U.S. Department of Commerce has documented an estimated 5,700 fires in electric and plug-in hybrid vehicles and nearly 200,000 structure fires tied to lithium-ion batteries, with consumer battery fires climbing 10% each year, according to a 2026 study.

“Since we are now using lithium batteries everywhere – rooftop energy, electric vehicles, even electric airplanes – this is a very severe problem, and it must be solved,” Yurkiv said. 

Creating more stable batteries for consumer and industrial use could save lives and improve the safety of the large-scale storage systems renewable energy depends on.

Searching for answers in alloys

Yurkiv used an artificial intelligence model to identify alloys, called high-entropy oxides, that could serve as the cathode for lithium-ion batteries.

“This is a very new class of materials. We did not even know about them 20 years ago,” he said. “They have very high electronic conductance, and no matter what you do to them, the material won’t release oxygen so they are extremely stable.” 

High oxygen levels contribute to thermal runaway, which occurs when a single cell in a battery pack overheats and triggers a chain reaction that causes nearby cells to heat, catch fire, or, in severe cases, explode.

Arranging complex alloys into a stable battery structure is extremely difficult, given their unpredictability. Yurkiv plans to arrange lithium, nickel, manganese, titanium, molybdenum and niobium into stable battery cathode formulations, so he trained his large language model to generate plausible arrangements of the elements.

"These elements have to be located precisely in this crystal structure, and in specific locations concerning the other elements,” he said. His model, he added, could be expanded for general research in materials science. 

“Once we generate our structures with the high-entropy oxide materials using this LLM-based principle, the same technique could be used to propose and create other novel materials.”