(Battery Division Student Research Award sponsored by Mercedes-Benz Research & Development) Degradation of Carbonaceous Negative Electrodes in Lithium-Ion Batteries across Multiple Length Scales

Abstract

The graphitic negative electrode is widely used in today’s commercial lithium-ion batteries. However, its lifetime is limited by a number of degradation modes, particularly growth of the solid electrolyte interphase (SEI), lithium plating, and electrode inactivation. Two major challenges to better batteries are the range of length scales (nanometers to centimeters) over which degradation modes occur, as well as slow development times. In this presentation, we overcome these challenges by studying the degradation of carbon electrodes in both model systems and commercial devices, and by using machine learning methods for accelerated battery optimization. First, we study SEI growth on carbon black via microscopy, electrochemistry, and modeling. We first use cryogenic transmission electron microscopy (cryo-EM) to image the SEI on carbon black and track its evolution during cycling. We observe an evolution of inorganic components in thin (~2 nm), primary SEI directly interfaced to the carbon black, as well as deposits of SEI that span hundreds of nanometers.1 Second, we electrochemically measure the dependence of SEI growth on potential, current magnitude, and current direction during galvanostatic cycling. We find that SEI growth strongly depends on all three parameters; most notably, SEI growth rates increase with nominal C rate and are significantly higher on lithiation than on delithiation.2 Third, we model the SEI as a mixed ionic-electronic conductor, where the ionic concentration modulates the electronic conductivity. This model can account for the previously observed directional dependence.3 This work illustrates the MIEC-like nature of the SEI on carbonaceous anodes and illustrates the strong coupling between charge storage (i.e. intercalation) and SEI growth. Second, we characterize the cell-level degradation of commercial lithium iron phosphate (LFP)/graphite cylindrical cells during fast charging.4 We find that the graphite electrode exhibits significant and highly heterogeneous degradation during fast charging, with large inactive regions located near the tab. This ionic inactivation of the electrode occurs via large-scale SEI growth, preceding more conventional fast charging degradation modes such as lithium plating. Third, we optimize a six-step fast-charging protocol that achieves 80% state of charge in ten minutes on commercial LFP/graphite cylindrical cells. We first develop a machine learning algorithm that uses cycling data from the first 100 cycles to predict cycle lives that reach up to 2300 cycles with ~9% error.5 Then, we use an optimal experimental design methodology for fast-charging protocol optimization, with two key elements to reduce the optimization cost: early prediction of failure, which reduces the cost per experiment, and adaptive Bayesian multi-armed bandits, which reduces the number of experiments required.6 The fast charging protocols identified by this algorithm are unexpected given the battery literature.