AI has revolutionized the battery design process, resulting in a new material that enables the construction of a functional battery using up to 70 percent less lithium compared to some existing designs.
The significance lies in reducing dependence on the expensive and environmentally damaging mineral, lithium, commonly used in lithium-ion batteries that power various everyday devices and electric vehicles.
Nathan Baker and his team at Microsoft leveraged AI to expedite the discovery and testing of materials for the battery, focusing on solid-state batteries. The researchers explored alternatives for the electrolyte, the component through which electric charges move.
Starting with 23.6 million candidate materials, the team used an AI algorithm to eliminate unstable options and those with weak chemical reactions. Within a few days, the list was narrowed down to a few hundred candidates, some of which had not been previously studied.
Despite not being material scientists, Baker consulted experts, including Vijay Murugesan at the Pacific Northwest National Laboratory. Murugesan’s team refined the AI’s suggestions and eventually selected a novel electrolyte recipe that replaced half of the expected lithium atoms with sodium.
The resulting working battery exhibited lower conductivity than comparable prototypes using more lithium, but Murugesan emphasized the need for further optimization.
The collaboration between the Microsoft team and Murugesan’s lab, from the initial conversation to the functional battery capable of powering a light bulb, took approximately nine months.
Rafael Gómez-Bombarelli at the Massachusetts Institute of Technology commended the project, highlighting the integration of bleeding-edge machine learning tools to accelerate and enhance traditional physics calculations.
However, he cautioned that challenges may arise in the future, particularly concerning sparse data for training AI and the potential complexity of combining elements in materials beyond battery components.