High-accuracy uranium determination via functional decomposition to address matrix effects in XRF spectroscopy

Abstract

For quantitative analysis of uranium in core samples using X-ray fluorescence (XRF) spectroscopy, matrix effects, particularly the interference caused by overlapping peaks from rubidium, significantly compromise measurement accuracy. To address this issue, we propose a high-precision uranium quantification method based on a mathematical functional decomposition model of energy-dispersive XRF spectra. The procedure involves wavelet-based noise reduction, background subtraction using the Statistics-sensitive Nonlinear Iterative Peak-clipping (SNIP) algorithm, and adaptive Gaussian fitting to extract the net intensity of the uranium La₁ peak. To improve quantification accuracy, systematic parameter optimization within a functional model framework is also incorporated. Specifically, Coiflet wavelets are employed for spectral denoising, the optimal number of SNIP iterations is determined as k = 30, and peak fitting is performed using an adaptive dual-Gaussian model capable of adjusting to variations in peak position, full width at half maximum (FWHM), and intensity. Experiments with 10 standard samples demonstrate that the linear correlation coefficient (R²) between the La₁ peak counts and uranium concentration improves from 0.5918 to 0.9917 after applying the proposed method. Additional validation with samples of varying uranium concentrations shows relative deviations of less than 6.80% when compared with ICP-MS results, confirming the method's applicability. Moreover, the method exhibits high measurement stability, with a relative standard deviation (RSD) better than 1.97%, and achieves a detection limit (LOD) of 3 µg g⁻¹. These results highlight the excellent performance of the proposed method in resolving overlapping peak interference due to matrix effects, offering valuable guidance for the development of field-deployable XRF instruments for core analysis.