Nuclear Forensics Analysis Of Fission Products By Means Of Chemometric Laser Induced Breakdown Spectroscopy

Abstract

In the wake of the current global nuclear renaissance, the threat of terrorism involving nuclear materials is real. The lethality involving detonated improvised nuclear devices (IND) “dirty bombs” or even radiological dispersal devices (RDD) is high as compared to any other form of terrorism. Such detonation events need to be timely detected and contained. As such, there is a need for a direct, rapid, non-invasive, remote and in situ state-of-the-art analytical techniques that can detect and characterize intercepted nuclear and radioactive materials (NRM) to apprise attribution. Nuclear forensics involves analysis of intercepted nuclear materials to provide sufficient evidence for attribution. Current nuclear forensics techniques are destructive, time-consuming, invasive and laborious besides requiring a sizeable sample size of NRM. As such, they are of limited utility in direct rapid nuclear security analytics and attribution. Laser-Induced Breakdown Spectroscopy (LIBS), when combined with chemometrics, has the potential to be developed towards overcoming these limitations. The goal of this work was to develop a chemometric enabled LIBS methodology for direct, rapid and non-invasive nuclear forensics detection, quantitative analysis and attribution of fission products (FP) in the vitrified glass, nuclear powders and high-level liquid wastes (HLLW) in support of nuclear security. In this regard, selected FP (Rb, Sr, Y, Zr) were spiked in a uranium augmented (0-5 % natural grade) matrix (composed of; SiO2, Na2CO3 Al2O2, and CaCO3) in typical levels at which they occur in high-level nuclear wastes. The resulting sample was prepared as fused glass to simulate nuclear glass, mixed with cellulose and pressed into pellets to mimic high-level nuclear powders as may be encountered in detonation events and a third batch prepared to mimic HLLW resulting from nuclear fuel reprocessing. Drop coating deposition (DCD) of HLLW on perspex was found to produce the best signal to noise ratio (SNR) of 57.895 for Rb I 780.011 nm of 9 ppm. Multivariate calibration strategy for quantitative analysis of Rb, Sr, Y, and Zr was achieved by utilizing artificial neural network (ANN). Various ANN algorithms were tested with feed forward back propagation algorithm, providing (R2 > 95 %) calibration accuracy. The relative error of prediction (REP) of <10 % was realized for each model. Validation of HLLW models was achieved using synthetic inductively coupled plasma (ICP) and atomic absorption spectroscopy standard solutions respectively and river clay PTXRFIAEA09 SRM with <10 % deviation from certified values. v Detection limits for each of the spiked elements in different samples types were: fused glass LoD≤200 ppm, powders pellets LoD≤71 ppm and HLW liquids LoD≤8 ppm. Simulate high-level nuclear waste fluids typical of a nuclear forensic scene (detonation or accidental spillage of HLW liquids or unlawful release of FP to the environment) were treated together with base matrix standards to achieve PCA clustering that differentiates the nuclear wastes and non-nuclear wastes based on trace FP. Support Vector Machine (SVM) was utilized in developing qualitative clustering hyperplanes based on the presence of FP (Rb, Sr, Y, Zr) in samples. We report > 85 % accuracy in discrimination of FP in terms of trace concentrations and their corresponding spectral response signatures. These SVM findings are an important component of nuclear forensic attribution as samples containing FP in particular concentrations are separated by the developed hyperplanes and hence furnish useful nuclear forensic interpretations. The novelty of the developed analytical methodology lies in small samples involved (2 μl for liquid samples with computed SNR of 57.895) and ( 3 mm for solid fragments) to furnish nuclear forensic signatures. Hence, chemometric-LIBS provides a robust tool which can be integrated into a suitable software-user interface of a handheld LIBS for rapid analysis of NRM in the context of nuclear forensics.