Insight into the accuracy and precision of SEM–EDS hyperspectral mapping for quantitative phase analysis in hydrated cement paste

Recent technological advances in the elemental quantification from scanning electron microscope (SEM) with energy-dispersive X-ray spectroscopy (EDS) and multi-elemental mapping allow us to identify the phase quantities and their spatial distribution in cementitious materials. While it can provide significantly useful characterisation data, its reliability and accuracy must be validated, and the limitation should be addressed to avoid any exaggerated data interpretation. This study evaluates the precision, repeatability, and quantitative reliability of edxia-based SEM-EDS hyperspectral mapping for phase analysis of hydrated cement pastes. Two systems, general purpose cement (GP) paste and a blended paste containing 70 wt% GP and 30 wt% fly ash (GPF paste), were investigated after 28 days of hydration. Three SEM-EDS instrument configurations, two classification strategies, and two specimen preparation methods were compared, with X-ray fluorescence (XRF) and X-ray diffraction (XRD) as independent validation methods. The hyperspectral mapping approach demonstrated high repeatability across multiple regions of interest under consistent acquisition conditions and successfully identified major hydration products, including inner and outer C-S-H, portlandite/calcite, AFm/AFt, residual clinker phases, and fly ash-related reaction products. In the GPF system, it captured the chemical heterogeneity caused by fly ash reaction, showing higher Si/Ca and Al/Ca ratios in C-(A)-S-H than those in the GP paste. However, cross-validation revealed systematic deviations in absolute quantification. SEM-EDS underestimated CaO and overestimated SO 3 compared with XRF and overestimated total C-S-H and aluminate phases while underestimating portlandite/calcite compared with XRD. These discrepancies are mainly attributed to mixed-pixel effects and nanoscale phase intergrowth.