Multimodal high-throughput approach assisted by deep learning for the analysis of ceramic saggars

The ceramic saggar is a container utilized in the production process of cathode materials for sodium-ion or lithium-ion batteries, and the characterization of saggar damage mechanisms often relies on the analysis of microstructure and composition of micro-sized areas. However, these results are often unreliable as they derive from a limited amount of data and cannot thus be utilized to perform a comprehensive and accurate analysis of the corrosion resistance mechanism of the saggar. In this study, deep learning combined with multimodal high-throughput in situ characterization methods, including array-SEM, μ-XRF, and μ-XRM, is employed to comprehensively analyze the multiscale characteristics of cordierite–mullite saggars. The 3D distribution of minerals in a 20×13×12 mm3 saggar sample is obtained. The results indicate that the direct dissolution of cordierite corrosion resistance reaction products is the primary damage mechanism of the saggar.