High Accuracy Battery Modeling : Fully 3D-Resolved Lithium-Ion Battery Mesostructure Including Carbon Binder Domains

Abstract

In the literature, reported 3D-resolved models rely on oversimplifications, such as an implicit representation of the carbon-binder domains (CBD) through the use of effective parameters for porosity and tortuosity or by merging CBD with the Active Material (AM) as a single solid phase. This work’s novelty relies on the explicit representation of CBD, leading to a new level of accuracy in terms of electrochemical modeling. This achievement is made possible thanks to an in-house algorithm, called INfinite Number Of phases meshing through Voxelization (INNOV) [1] . INNOV can generate a volumetric mesh from data of different types due to its flexible input format. INNOV takes as input binary stack of images to reconstruct the 3D structure. Such an input can result from tomography imaging, from slicing a 3D object or from Coarse Grained Molecular Dynamics (CGMD) simulations [2] . For the latter a function has been developed to convert its output (coordinates of the centers and radii of the particles) into a binary stack of images. The segmentation method of Nielson and Franke [3] has been translated and optimized for MATLAB language and modified to suit the COMSOL Multiphysics meshing importation process. This algorithm is designed in the scope of the ARTISTIC Project [4] to import a multi-phase volumetric mesh of an electrode (from a CGMD simulation) into COMSOL Multiphysics to simulate the performances of the cell. The core of this work is to increase the current level of precision of the modeling of batteries by separating active and inactive materials. In doing so, one must not sacrifice the integrity of the mesostructure geometry. To ensure this, INNOV provides a number of observables, which can be compared to experimental numbers (e.g. arising from tomography characterizations). Once the integrity of the mesh is ensured, electrochemical simulations can be done to characterize the structure (Figure 1). Such characterizations can capture the role of CBD to further understand the limiting phenomena and predict the optimal LIB structure to achieve the optimization of the fabrication process. Simulations on NMC/Li half-cell have been carried out to demonstrate the versatility of this algorithm. In conclusion, INNOV offers a time-efficient tool to perform meshing without requiring substantial computational resources. Simulations can later be performed to characterize these meshes with the CBD explicitly considered. It is also a powerful tool to couple with tomography imaging. FIGURE TABLE Figure 1. From CGMD simulation to a volumetric mesh (35.1*35.1*42 µm). In yellow is the CBD, in red the AM. At the far right is the discharge curve for this specific mesh with COMSOL. REFERENCES [1] M. Chouchane, A. Rucci, A.A. Franco (submitted 2019) [2] A.C. Ngandjong et al. Multiscale Simulation Platform Linking Lithium Ion Battery Electrode Fabrication Process with Performance at the Cell Level. J. Phys. Chem. Lett. (2017). 5966–5972 doi:10.1021/acs.jpclett.7b02647 [3] G.M. Nielson, R. Franke Computing the separating surface for segmented data. IEEE Vis. (1997). 229–233 doi:10.1097/MOP.0000000000000041 [4] https://www.u-picardie.fr/erc-artistic/ Figure 1