Improved Rate Capability for Dry Thick Electrodes Through Finite Elements Method and Machine-Learning Coupling

Abstract

A coupled Finite Elements Method (FEM) and Machine-Learning (ML) workflow is presented to optimize the rate capability of thick positive electrodes (ca. 150 µm and 8 mAh/cm²). An ML model is trained based on the geometrical observables of individual LiNi0.8Mn0.1Co0.1O2 particles and their average state of discharge (SOD) predicted from FEM modeling. This model not only bypasses lengthy FEM simulations, but also provides deeper insights on the importance of pore tortuosity and the active particles size, identified as the limiting phenomenon during the discharge. Based on these findings, a bi-layer configuration is proposed to tackle the identified limiting factors for the rate capability. The benefits of this structured electrode are validated through FEM by comparing its performance to a pristine mono-layer electrode. Finally, experimental validation using dry processing demonstrates a 40% higher volumetric capacity of the bi-layer electrode when compared to the previously reported thick NMC electrodes.