Researchers used a Lawrence Berkeley National Laboratory, or Berkeley Lab, X-ray device to gain insight into the process by which lithium-ion batteries, such as those found in electric vehicles, lose battery life over time.
These findings point toward a key to designing electric vehicles that charge much faster and maintain battery health. The study, a collaboration between Stanford University, the Massachusetts Institute of Technology and Toyota Research Institute, used imaging technology at Stanford University and UC Berkeley, as well as novel machine learning techniques to understand how fast-charging battery particles interact with lithium ions.
“The findings establish new design rules to make longer lasting batteries that could be recharged quickly,” said Will Chueh, study author and Stanford University materials science and engineering associate professor, in an email.
This could lead to more effective batteries for devices as small as laptops and as large as renewable energy grids. A primary aim of the research will be to help create electric vehicles that need much less time to recharge, bridging the refueling gap between electric and gasoline-powered cars.
The faster a lithium-ion battery charges, the shorter its life span will be, Chueh said in a YouTube video. Scientists previously thought this was because lithium ions moved in and out of battery particles at a uniform rate during charging, leading to damage over time.
This study, however, revealed a different picture, according to Chueh.
“Some particles immediately release a lot of their ions while others release very few or none at all,” Chueh said in the email. “The quick-to-release particles go on releasing ions at a faster rate than their neighbors – a positive feedback, or ‘rich get richer,’ effect that had not been identified before.”
The imbalance puts more strain on individual electrode particles, Chueh added, which the researchers now believe causes fast-charging batteries to degrade and lose life over many charges.
Berkeley Lab’s Advanced Light Source was used to view individual battery electrodes undergoing charging, while another device at Stanford University observed particles as a group. The data from these machines was then processed using artificial intelligence.
“The scientific machine learning used in this work- combined with advanced characterization and physics-based modeling- could significantly accelerate the (research and development) cycle,” Chueh said in the email. “The new study builds on two previous advances where we used more conventional forms of machine learning to dramatically accelerate both battery testing and the process of winnowing down many possible charging methods to find the ones that work best.”
The computer was taught to choose the right equations to describe the data, revealing the physics and chemistry of the battery models. This study marks the first time this technique, called “scientific machine learning,” has been used in battery research.