Researchers at the UChicago Pritzker School of Molecular Engineering are already using artificial intelligence and machine learning to create better vaccines and cancer treatments, advance how plastics are studied and find better materials for proteins, qubits and other scientific endeavors.
They’re also advancing the use of AI itself, charting a path toward sustainable machine learning, developing new AI-inclusive teaching methods and applying machine learning insights to arenas previously seen as unrelated.
Bolstered by a new Google Research Scholar Award, PME Neubauer Family Assistant Professor of Molecular Engineering Chibueze Amanchukwu hopes to use this new tool to tackle an issue at the heart of the global energy crisis – the hunt for better batteries.
As the globe transitions off fossil fuels, more and better batteries will be needed to store the energy created by solar and wind power for times the sun isn’t shining and the wind isn’t blowing.
Amanchukwu, who recently received the 2024 Camille Dreyfus teacher-scholar award, is turning to AI and machine learning to sort through the millions upon millions of potentially useful chemical compounds to find better materials for battery electrolytes.
“We are taking advantage of the generative AI approaches that have been explored in other fields and are using them for battery applications,” he said.
A massive chemical space
“The difficult side is that no one knows what the entire chemical space is,” Amanchukwu said.
Amanchukwu compared it to drug discovery, where the number of compounds that meet the criteria to be potentially useful in the manufacture of drugs is commonly estimated at 10 to the 60th power, or a 1 with 60 zeroes after it. That doesn’t mean there are that many drugs out there. It means that is the theoretical potential number of compounds that researchers can look through to find new drugs.
“If I use the same analogy for batteries, there could be 10 to the 60 potential options of battery electrolytes or battery additives,” he said. “But if you ask how many have we studied today, it is about 10 to the two.”
That means going from about 100 potential battery electrolyte materials to 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000.
“How do you efficiently go from 10 to the 2 to 10 to the 60?” he said. “That means there's so much potential in the design space for electrolytes we have not yet accessed today.”
Turning to AI
To help wade through this ocean of potential electrolytes, looking for materials promising enough to turn over to researchers for study, Amanchukwu is developing what he calls ElectrolyteGPT.
The proposal, “ElectrolyteGPT to accelerate battery electrolyte discovery,” recently won an unrestricted $60,000 Google Research Scholar Award in the Applied Sciences category.
Amanchukwu hopes this will just be the beginning of his collaboration with the computing powerhouse. His research is already underway, and finding possible new solutions may also mean bringing cleaner energy in the form of batteries to market sooner.
“We need clean energy solutions as soon as possible,” Amanchukwu said. “This effort could ultimately help us deliver solutions more quickly.”