UIC professor using AI and supercomputing for next-generation energy storage
1
Professor Subramanian Sankaranarayanan received a grant from Argonne National Laboratory for a project titled “Development and Training of Reactive Force-fields for MD Simulations of Battery Materials.”
Sankaranarayanan’s research group at UIC is working with Ford Motors as part of a Department of Energy (DOE) project under the High-Performance Computing for Energy Innovation program supported by the Office of Energy Efficiency and Renewable Energy to design better energy storage systems using artificial intelligence. Through the initiative, researchers can deploy world-class supercomputers to help manufacturers achieve significant energy and cost savings, improve product performance, expand their markets, and grow the economy.
The researchers are utilizing the high-performance computing resources at Argonne National Laboratory and using the BLAST (Bridging Length-timescale via Atomistic Simulation Toolkit) multi-fidelity optimization framework developed by Sankaranarayanan’s group at UIC to understand how different elements — like carbon, lithium, hydrogen, Oxygen, Phosphorus, Manganese, Nickel, Cobalt, Silicon, Zirconium, Fluorine, and iron — interact inside energy storage systems.
“The solid-electrolyte interphase (SEI) interface is a very thin, protective layer that forms on the surface of a battery’s negative electrode – usually made of graphite or lithium – when the battery is first used. It’s created by a chemical reaction between the electrode and the liquid inside the battery, called the electrolyte. Modeling the SEI is critically important because the SEI plays a central role in the performance, safety, and longevity of rechargeable batteries, particularly lithium-ion, sodium-ion, and other next-generation chemistries. Our work will lead to predictive understanding that drives the rational design of stable interfaces, longer-lasting batteries, and safer energy storage systems,” Sankaranarayanan said.
Ford is leading the identification and collection of battery materials that are most important to them, and also collecting data about how those materials behave. Then, the UIC team will train a machine learning foundation model, which will describe the interactions between atoms in a way that allows for the breaking and formation of chemical bonds and how those materials evolve with time during a battery operation.
“Our calculations offer a predictive microscope and enable materials design atom by atom, molecule by molecule, which is essential for engineering next-generation battery technologies,” said Abibat Adekoya-Olowofela, a graduate student in Sankaranarayanan’s group.
Ford will check the model’s accuracy by comparing it to precise scientific calculations and experiments. As a deliverable, UIC will provide Ford a tested and reliable machine learning model for prediction and additional testing.
“From electrons to material interfaces, our quantum-level and atomistic modeling bridges the gap between fundamental chemistry and applied battery innovation,” said Sukriti Manna, a visiting research professor in MIE.