A unified quantitative inversion framework for neutron resonance transmission analysis of low-content uranium considering self-shielding and matrix effects

Abstract: Neutron resonance transmission analysis (NRTA) is a powerful non-destructive technique for nuclide identification and quantification. However, accurate quantitative analysis of low-content uranium remains challenging because resonance signals are weak, matrix attenuation introduces interference, and self-shielding becomes significant in centimeter-scale samples. In this work, we propose a unified quantitative inversion framework to address these challenges. A Geant4-based neutron transport model is established to simulate the transmission spectra of uranium--silicate matrix mixtures for different sample thicknesses and uranium contents. The framework combines a thick-target transmission model, which corrects the nonlinear attenuation caused by self-shielding, with a multi-element attenuation model, which explicitly accounts for matrix contributions. The results show that, compared with conventional thin-target and single-element models, the proposed framework reduces the relative error of uranium-content inversion from approximately 57.8% to less than 1% for high-content samples and from 35% to less than 3% for low-content samples. A sample thickness of 1.5--2.0 cm provides an appropriate balance between resonance contrast and statistical reliability, with a relative error below 5%. This work provides a physically constrained analysis method and practical guidance for future NRTA experiments using the accelerator-driven neutron source at East China University of Technology.