(733c) A New Charging Algorithm for Lithium-Ion Batteries Based on a Constant Lithium Gradient Protocol

This paper presents a new optimal charging methodology called Constant Gradient Constant Voltage (CGCV), which limits mechanical crack propagation at the surface of the particles in electrodes by regulating the current to maintain a constant lithium gradient at the surface. One important cause of battery degradation is the stress developed by the concentration gradient inside active particles. The proposed new charging protocol actively controls the concentration gradient by using the lithium diffusivity as strongly state-of-charge (SOC) dependent. Here, the diffusivity is not real time measurable, so a modeling-based approach is incorporated, in which the charging protocol is guided based on the concentration gradient predicted from the modeling results. The upper limit of the concentration gradient is defined based on the maximum stress during the conventional Constant Current Constant Voltage (CCCV) charging. Two gradient options of Mid and High (which is set by the maximum stress limit relative to 0.1C). The experimental results under three charging scenarios of high CGCV, mid CGCV, and standard CCCV. CGCV charging methodologies show slightly slower degradation rates with a 56% (High CG) or 49% (Mid CG) reduction in charging time when comparing CC to CG. This is a new protocol considering battery physics on particle level and has great potential to address the key challenge of regulating battery degradation during fast charging of Electric Vehicles (EVs).

Generally, charging methodologies are experimental approaches with the primary focus on improving cycle life and do not directly consider specific degradation mechanisms. There is no effort to improve mechanical degradation, which is a critical degradation mechanism. The extraction and insertion of ions distorts the structure of the electrodes. Analysis via TEM has shown that if there is insufficient time for the ions to diffuse through the electrode, inhomogeneity results in stress being relieved via fracturing. Research into LiCoO2 shows that a model can determine a critical fracture point based on particle size, diffusivity, and current rate by solving the elastic boundary conditions . Therefore, the main method to prevent crack propagation is to lower the critical facture energy by reducing the particle size or the C-rate. This model determines critical particle size is 250 nm at 2C and decreases to 190 nm at 6C during discharge. Another important factor to consider is if any phase shift occurs for the electrode particles. Another stress fracture model studying LiMn2O4 shows that if Jahn-teller distortion occurs during the transition from spinel to cubic then fracture is unavoidable. This occurs due to the mismatch in lattice parameters between the two crystal structures causing an immediate stress high enough to induce fracture. This region can be avoided for particles between 500 nm to 800 nm during charging if the cell is kept above 3 V. However, if the particle is kept in single phase, then fracture tends to occur in 4 V plateau region if current exceeds 5.36C for 5 nm particles . Though lesser subcritical fracturing can occur due to surface interactions between cracks and the electrolyte due to natural wear and tear.

This work proposes an approach designed to address that issue. An electrode is assumed to have preexisting cracks on its surface; these cracks will not grow when the maximum energy release is smaller than the fracture energy. This maximum energy release is caused by the gradient of Li ion distribution determined by the relationship between diffusivity and flux. Here, the diffusivity is strongly dependent on SOC, so the proposed idea is to regulate the gradient by varying the current based on the SOC dependent diffusivity. Specifically, the concentration gradient is controlled and is fixed to a certain value based on a CC profile. This is called Constant Gradient (CG), which is designed to allow higher current input during the high diffusivity SOC regions, while minimizing the stress development in the low diffusivity regions by using a smaller current input. For ideal implementation, the SOC dependent variable diffusivity would be accessible on-line for the control, but this is difficult to directly measure especially when the cell is running. Therefore, a modeling-assisted approach is incorporated instead to provide the current profile by considering the SOC-dependent diffusivity.