Huanyu Wu, Yuan Zou, Chi Zhang, Wei Yang, Bo Wu, Kai-Leung Yung, Qi Zhao 

Chang’e-5 (CE-5) lunar regolith samples were scanned using X-ray micro-computed tomography (micro-CT), and over 0.7 million particles were extracted from the images through machine learning-based segmentation. This is the largest three-dimensional (3D) image dataset on lunar regolith particles to date, offering a unique opportunity to study the key characteristics of lunar regolith. The image intensity was correlated with mineral density, allowing for the assessment of the bulk density (1.58 g/cm3), true density (3.17 g/cm3), and mineralogy of the lunar regolith. Glass and plagioclase contributed 45.6 wt.% of the samples, while pyroxene and olivine made up 49.7 wt.%, and ilmenite accounted for 4.7 wt.%. The median grain size of CE-5 was 57.5 μm, smaller than the Apollo 11, 16 and Luna 16, 20 and 24 samples. Spherical harmonic (SH) analysis and axial ratio (AR) measurement revealed that the CE-5 lunar regolith particles have more complex shapes than two common terrestrial soils and exhibit less spherical shapes than Apollo 11, 16 and Luna 16, 20 and 24 samples. We recommend using size and shape characteristics cautiously when inferring lunar regolith maturity because the intrinsic crystal size of the protolith and complex lunar surface weathering can cause significant size and shape variations. Additionally, characterizing particle shapes requires a large sample size (>1000) to avoid skewed results from outliers. Our non-destructive examination method offers a novel and appealing approach for analyzing critical physical, mineralogical, and morphological properties of million-scale extraterrestrial soil particles, paving the way for future deep space explorations.