In a new study published this week in MRS Bulletin, a Stanford research team led by doctoral student Brandi Ransom and Associate Professor Evan Reed, and including Aionics CEO Dr. Austin Sendek, reported the discovery of two promising new cathode materials for lithium ion batteries using machine learning (ML) models.
The two materials, CrOF4 and NbFe3(PO4)6, appear to demonstrate high cathode performance, including high voltage, while also minimally expanding upon cycling. This near zero expansion can help mitigate cracking of the cathode, a major cause for Li-ion battery degradation.
These two materials were identified out of more than 38,000 candidates using ML models for predicting the electrochemical properties of voltage, capacity, energy density, and expansion. Further investigations using the quantum mechanical simulations of density functional theory confirmed the materials match or improve upon the theoretical performance of commonly used cathodes such as LCO and LFP.
According to the study’s lead author Brandi Ransom, “The new and interesting structures of these materials show how influential it can be to take a broad approach or bird’s-eye-view towards materials discovery. We hope that processes similar to these can help set jumping off points for investigating new materials in order to jump straight into the optimization process, and give confidence to new directions in battery materials.”
The study can be accessed online here: https://link.springer.com/article/10.1557/s43577-021-00154-9.
Dr. Sendek added: “This work is really exciting because, not only do we now have two promising new cathode materials for further study, but we also have a convincing proof point that machine learning models, even when trained on small data sets, can really accelerate the design and discovery of new cathode materials.”
The other co-authors on the study were doctoral student Nathan Zhao, Dr. Ekin D. Cubuk of Google Brain, and Associate Professor of Materials Science & Engineering William Chueh.