Particle physics experiments

X-ray experiments and machine learning innovation could reduce years of battery R&D


Engineer Bruis van Vlijmen works inside Battery Informatics Lab 1070 at Arrillaga Science Center, Bldg. 57. Credit: Jacqueline Orrell / SLAC National Accelerator Laboratory

An X-ray instrument from the Berkeley Lab contributed to a battery study that used an innovative approach to machine learning to accelerate the learning curve of a process that shortens the lifespan of lithium batteries when charged. fast.

Researchers used Berkeley Lab’s advanced light source, a synchrotron that produces light ranging from infrared to X-rays for dozens of simultaneous experiments, to perform a chemical imaging technique known as microscopy Scanning Transmission X-ray, or STXM, state-of-the-art ALS light line known as COSMIC.

The researchers also used “in situ” X-ray diffraction on another synchrotron – SLAC’s Stanford synchrotron light source – which attempted to recreate the conditions present in a battery, and further provided a model of multi-particle battery. The three forms of data have been combined into one format to help machine learning algorithms learn the physics at work in drums.

While typical machine learning algorithms look for images that may or may not match a set of training images, in this study, the researchers applied a more in-depth set of data from experiments and other sources to allow more refined results. This is the first time this brand of “scientific machine learning” has been applied to the battery cycle, the researchers noted. The study was recently published in Natural materials.

The study benefited from the ability of the COSMIC beamline to distinguish the chemical states of approximately 100 individual particles, which was made possible by the high-speed, high-resolution imaging capabilities of COSMIC. Young-Sang Yu, an ALS researcher who participated in the study, noted that each selected particle was imaged at around 50 different energy stages during the cycling process, for a total of 5,000 images.

Data from the ALS experiments and other experiments were combined with data from fast-charging mathematical models and with information on the chemistry and physics of fast-charging, and then incorporated into machine learning algorithms.

“Rather than letting the computer determine the model directly by simply supplying it with data, as we did in the previous two studies, we taught the computer how to choose or learn the right equations, and therefore the right one. physical, “said the Stanford postdoctoral researcher. Stephen Dongmin Kang, co-author of the study.

Patrick Herring, principal investigator at the Toyota Research Institute, which supported the work through its Accelerated Materials Design and Discovery program, said: “By understanding the fundamental reactions that occur within the battery, we can extend its life. life, allow faster charging and ultimately design better battery materials.

Reference: “Fictitious Phase Separation in Layered Oxides of Li Driven by Electro-Autocatalysis” by Jungjin Park, Hongbo Zhao, Stephen Dongmin Kang, Kipil Lim, Chia-Chin Chen, Young-Sang Yu, Richard D. Braatz, David A . Shapiro, Jihyun Hong, Michael F. Toney, Martin Z. Bazant and William C. Chueh, March 8, 2021, Natural materials.
DOI: 10.1038 / s41563-021-00936-1