Integrated Computational Materials, Process and Product Engineering – IC-MPPE 2022

Lithium (Li)-ion based batteries are one of the most used systems in decentralized storage systems, e-mobility or mobile electronic devices etc. However, an essential problem of state-of-the-art cells concerns the fast fading of the capacity during electrochemical cycling. The use of silicon (Si) as an active material in the anode provide promising prospects. The understanding of the microstructure of silicon-based anodes and its change with electrochemical cycling in connection with the electrochemical properties is highly crucial to develop more advanced Li-ion batteries. However, a big problem for the microstructure characterization concerns the complex hierarchical structures of the anode material going from m- down to nm-scales. Here, multi-method approaches are essential to cover the different scales with respect to resolution, contrast, and the representative volume of interest to gain sufficient statistical information while maintaining the ability to extract the needed information at relevant scales and with appropriate contrast modalities. Another challenge concerns the accurate and efficient analysis of the “big” microstructure data. In this work we discuss a multiscale correlated workflow suitable to investigate the microstructure of Si-based anodes in correlation with chemical element information and to gain a comprehensive, multiscale, representative picture of the intricate microstructure dictating the ultimate electrode performance. To handle the big image data accurately and to retrieve statistically relevant microstructure information we implement a machine learning based architecture for the segmentation of the gained image data of the Si-based composite anode.