Generation of Virtual Microstructures for the Optimization of Lithium-Ion Battery Cathodes

Abstract

The microstructure of lithium-ion battery cathodes rules their performance. However, the production of a variety of cathodes with different microstructures through laboratory experiments is very expensive with respect to costs and time. Thus, model-based simulation approaches, i.e., virtual microstructures generated and analyzed on a computer, have become an important tool for microstructural optimization. In this work, tools for the generation of realistic yet virtual microstructures are presented. Basically, three dimensional reconstructions of cathode microstructures are established by FIB tomography, and the individually shaped particles of the active material phase as well as the carbon black are extracted. A representative set of extracted particles is used to create structures with specified characteristics, such as e.g. the desired mass content of active material and conductive additive or specific particle size distributions. In combination with performance models, a preselection of promising design concepts can be conducted. The complete workflow from the generation of realistic virtual microstructures based on 3D reconstruction via FIB tomography towards the simulation of discharge curves will be shown using the example of a commercial blend cathode consisting of the two active materials LiNiCoAlO2 (NCA) and LiCoO2 (LCO). Figure 1

Understanding Deviations between Spatially Resolved and Homogenized Cathode Models of Lithium‐Ion Batteries

“…Thereafter, study 4 introduces a “real‐life” cathode with particle shapes from an FIB/SEM reconstruction work, which violates several homogenization assumptions. [ 28 ]…”

“…The simulations presented in the following shall disclose, whether the discussed adjustment of the solid‐state diffusion length works for such a complex 3D microstructure or not. The microstructure shown in Figure 15 is created from active material particles and carbon/binder agglomerates by a microstructure generator, as described by Joos et al [ 28 ] The microstructure parameters are listed in Table 5 , and the simulation results for a 3C discharge are shown in Figure 16 . The EHM/P2D model was first set up with the mean active material particle size determined by the EDT (configuration 1).…”

“…Microstructure data sets generated by a microstructure model are used in the two 3D models (3D(SP) and 3D(PH)), whereas characteristic parameters (porosity and active material fraction, surface area, tortuosity of pore phase, and PSD of active material) are extracted for the two homogenized models (EHM and P2D). [ 28 ] The active material particles are assembled by a “dropping and rolling” method, and, in case of the 3D(SP) model only, the conductive phase (carbon black and binder) is stochastically placed between the active material particles. [ 35 ] The generation of these microstructures is completely implemented in MATLAB (The MathWorks, Version 2019b, Natik, MA, USA).…”

“…This microstructure generation method is newly developed and entirely described by Joos et al in this issue. [ 28 ]…”

“…For a systematic investigation of the microstructural influence, microstructures with desired, predefined properties are particularly helpful and are here generated by a newly developed virtual microstructure generator. [ 28 ] After defining a representative volume element for the 3D models, the simulation results of the three models are compared with each other with different degrees of simplification, which allows conclusions to be drawn about the relevance of homogenization of the conductive additive. Subsequently, the influences of 1) non‐spherical particle geometries and 2) overlapping particles on the discharge behavior are analyzed using specially generated microstructures.…”