(Invited) Optimizing the Binder Distribution in Battery Electrodes with Manufacturing Process Simulations and Machine Learning

Abstract

The performance of lithium-ion battery cells strongly depends on the microstructure of the electrodes, which is determined by the spatial distribution of the active material, carbon additive, binder, and pores within the electrode volume. The manufacturing process determines this spatial distribution, which, in turn, determines the interfaces between these materials and the pores, thereby impacting the practical properties of the electrodes, such as their electrical conductivity, energy density, and wettability. The binder plays a crucial role in this process, as its spatial distribution controls the degree of percolation between the particles and the adhesion of the electrode with the current collector.Since 2017, we have been developing the ARTISTIC computational platform,1 which constitutes a digital twin of the battery electrode and cell manufacturing process. Supported by a combination of physics-based models and machine learning, this platform has been demonstrated by us for lithium-ion, sodium-ion, and solid-state batteries with electrodes produced by wet processing.2 The platform simulates each step of the manufacturing process, including mixing, coating/drying, calendering, electrolyte infiltration, and resulting cell performance. It allows for predicting 3D-resolved electrode and cell sandwich microstructures with their associated electrochemical performance as a function of the manufacturing parameters.In this talk, I will discuss how the ARTISTIC computational platform can optimize the spatial distribution of active material, carbon additive, and binder within the electrode volume. Using Bayesian Optimization, the platform predicts the manufacturing parameters required to obtain optimal electrodes in terms of several properties, such as electric conductivity, tortuosity factor, electroactive surface area, and energy density. I will particularly discuss the platform's capabilities to capture the adhesion of the electrode with the current collector through the binder. Additionally, I will present an extension of this digital twin for the 3D-resolved simulation of the dry processing of active material, carbon additive, and binder based on extrusion and compare the results (in the form of 3D-resolved electrode microstructures) with experimental data. Results from the wet and dry processing simulations will be discussed for electrodes made with different active material chemistries, such as NMC and LFP, and the crucial role of the binder in the heterogeneity of lithiation/delithiation of the electrodes upon their electrochemical cycling will be explored.Finally, building on our previously reported virtual reality digital tools for battery manufacturing,3 I will present a novel virtual reality interface that allows for the interactive and immersive use of the ARTISTIC computational platform to track the influence of manufacturing parameters on the active material, carbon additive, and binder's spatial location, as well as the resulting electrode properties. I will conclude by discussing why this virtual interface is also a powerful tool to assist in the electrode manufacturing optimization.