Determining the Uncertainty in Lithium Ion Battery Micro-Structural Parameters Extracted from Tomographic Data

摘要

Over the past few years X-ray tomographic microscopy (XTM) has advanced to an indispensable tool to characterize and quantify lithium ion battery (LIB) electrodes in terms of mi-crostructural parameters. Its non-invasiveness, as well as its 3D character render the technique not only suitable for ex-situ sample analysis, but also capable to perform in-situ and operando studies. However, in particular tomographic data of graphite negative electrodes typically suffers from low contrast between the pore space and the active material, due to the weakly X-ray attenuating nature of carbon based materials. As a result data binarization is error-prone, which directly translates into large uncertainties of the calculated microstructural parameters. This study systematically investigates the mentioned error propagation from the binariza-tion to the set of microstructural characteristics porosity, tortuosity and specific surface area on XTM datasets of seven commercial LIB negative electrodes. To allow for a controlled variation of the binarization procedure, we parametrize it with two parameters that are varied in meaningful uncertainty ranges: (i) a threshold, which mostly controls the foreground to background ratio and (ii) a morphological filtering parameter that controls the smoothness of the binarization. We find that the computed microstructural characteristics are very prone to binarization er-rors in particular for low porosity electrodes, where nonlinear error propagation leads to uncertainties in the tortuosity exceeding 100 %. Finally, while the presented analysis is computationally expensive, we propose a criterion that allows to estimate the expected uncertainties for the different microstructural parameters based on a simple histogram analysis. This analysis can easily be conducted for example dur-ing test experiments at an X-ray tomographic imaging beamline to optimize the imaging pa-rameters for subsequent data analysis.