Murakami et al. Supplemental Data for "Microstructural Analysis of Li-Ion Conductors with Deep Learning and SEM Images"

<p dir="ltr">For details, please refer to the paper: Murakami et al., "Deep Learning–Based SEM Image Analysis for Predicting Ionic Conductivity in LiZr₂(PO₄)₃-Based Solid Electrolytes" (DOI: *****).</p><h2><b>data_table.xlsx: </b></h2><h4><b>Description of columns in the dataset (Excel file):</b></h4><p dir="ltr"><b>sample.No</b>: Sample index number.<br><b>Ca, Si, Li, Zr, P</b>: Elemental composition of each sample (Li_1+2x+yCa_xZr_2-xSi_yP_3-yO₁₂)<br><b>1st heating temperature, 2nd heating temperature</b>: The first and second sintering temperatures of the sample.<br><b>Measured_Li_conductivity, Measured_Li_conductivity (Log)</b>: Experimentally measured lithium-ion conductivity (S cm⁻¹ at 30 °C) and its logarithmic value.<br><b>pred_1 (Log), pred_2 (Log), pred_3 (Log), pred_mean (Log)</b>: Predicted lithium-ion conductivities obtained by regression analysis (logarithmic values). The first three columns correspond to predictions from individual SEM images, and the last column is their average.<br><b>Reference</b>: Source of the data.</p><p dir="ltr"> Reference 1: H. Takeda et al., <i>Next Materials</i> 8 (2025) 100574, <a href="https://doi.org/10.1016/j.nxmate.2025.100574" rel="noopener" target="_new">https://doi.org/10.1016/j.nxmate.2025.100574</a> <br> Reference 2: H. Takeda et al., <i>Mater. Adv.</i>, 3 (2022) 8141–8148, <a href="https://doi.org/10.1039/D2MA00731B" rel="noopener" target="_new">https://doi.org/10.1039/D2MA00731B</a><br><b>Number of SEM pictures</b>: Number of SEM images obtained for each sample.</p><h2><b>SEM_images.zip</b>: </h2><p dir="ltr">These files consists of SEM images and numerical datasets (descriptors and objective variables) of composition, sintering temperature, and ionic conductivities for 52 samples (1-3 SEM images are included per 1 sample, total 130 images)</p><h2><b>python_codes.zip</b>: </h2><p dir="ltr">Python codes for four convolutional neural network (CNN) models used to investigate the relationship between these image data and ionic conductivity are provided.</p><h2><b>Segmentation_images.zip</b>: </h2><p dir="ltr">Positive and negative segmentation images for Li ionic conductivities in LCZSP materials. (See Figure 5 in the main text.) List of sample#, compositions and process conditions (heating temperatures are also included as csv formatted file.</p>