Quantitative analysis of Th and U in graphite matrix using femtosecond laser-induced breakdown spectroscopy

As one of the six nuclear energy systems selected by the “Fourth Generation Nuclear Reactor International Forum”, the Thorium Molten Salt Reactor (TMSR) has garnered significant attention due to its fascinating features, including excellent neutron economy, online fuel reprocessing, and reduced production of actinides. However, real-time online analysis of the nuclear elements remains a crucial technology throughout the fuel reprocessing. In this study, the femtosecond laser-induced breakdown spectroscopy (LIBS) was employed with both univariate and multivariate regression models for the quantitative analysis of Th and U elements in a graphite matrix. The concentrations of ThO 2 in six homemade samples were ranged from 4.998 to 40.012 wt % , while the concentrations of U 3 O 8 were ranged from 0.970 to 2.498 wt % . The mean relative error (MRE), root mean square error of prediction (RMSEP), root mean square error of calibration and coefficient of determination ( R 2 ) were employed as indicators to evaluate the quantitative accuracy and stability of the LIBS analysis. For the univariate regress, the standard curve method was performed, utilizing a non-interferential emission line (Th II 286.99 nm) as analytical line for Th measurement. However, no suitable analysis line was found for U element. For the multivariate regression, the random forest (RF) method and partial least squares (PLS) regression method utilized three spectral regions (285.00–288.00, 330.00–370.00 and 390.00–425.00 nm) as analytical bands for ThO 2 , while two regions (380.00–389.00 and 514.00–517.00 nm) were employed for U analysis. The results suggest that the PLS regression exhibited the best performance, yielding RMSEP T h O 2 = 0.785 wt % , RMSEP U 3 O 8 = 0.135 wt % , MRE T h O 2 = 4.24 % , and MRE U 3 O 8 = 4.90 % . Therefore, the combination of LIBS with multivariate regression models exhibits significant potential as a robust in-situ analytical approach for the online fuel reprocessing in TMSR.