BHATT, BOBBY YOGESH

Project Title
NUCLEAR FORENSIC ANALYSIS VIA MACHINE LEARNING ASSISTED LASER-BASED SPECTROSCOPY AND SPECTRAL IMAGING
Degree Name
DOCTOR OF PHILOSOPHY DEGREE IN PHYSICS
Project Summary

Nuclear forensics (NF) is a systematic and scientific methodology designed to identify, categorize, and characterize seized nuclear and radiological materials (NRM). The aim of NF is to reveal the geographical origin, process/production history, age and intended use of the NRM to prevent future diversions and thefts, thereby strengthening the national security of a country. The limitation in NF at the moment is the absence of suitable methodologies to directly, rapidly and non-invasively analyze limited size of NRM under concealed conditions. Laser based spectroscopy and spectral imaging techniques (laser induced breakdown spectroscopy (LIBS) and Laser Raman microspectrometry (LRM)) combined with machine learning techniques (ML) possess the power to conduct direct, rapid NF analysis of limited size NRM with accuracy and precision. The uranium lines at 386.592 nm, 385.957 nm and 385.464 nm were identified as nuclear forensic signatures of uranium in uranium trioxide bound in cellulose and uranium ore surrogates (uranium mineral ores and high background soil samples). The detection limit for uranium in cellulose was determined at 76 ppm. Uranium lines were
divided into resonant and weak lines, depending on the signal-to-background ratio. Resonant and weak uranium lines were utilized to develop multivariate calibration models in artificial neural network (back-propagation algorithm). The calibration models employing weak U-lines and resonant U-lines predicted uranium content in the certified reference material (CRM) (RGU1(400 ppm)) with relative error of prediction (REP) at 4.32% and 9.75% respectively. The model using weak U lines predicted the uranium concentration in the mineral ores (uranium) obtained from various parts of Kenya utilizing weak uranium between (103 - 837) ppm. The models were further validated with RGUMix (101 ppm) in addition to RGU-1 (400ppm). The calibration model using weak U-lines predicted the uranium concentration in RGUMix (101 ppm) and RGU-1 (400 ppm) at REP = 2.97% and 2.14% respectively, while using resonant U-lines predicted at REP = 69.07% and 4.22% respectively. The poor sensitivity of the resonant lines to changes in the low concentration uranium may account for the high REP of the model using resonant U-lines. On successful validation of the calibration model utilizing
weak U-lines, the uranium concentration in the uranium mineral ores was predicted ranging from (112 - 1000) ppm. LIBS spectra of HBRA soils of Kenya combined with principal component analysis revealed patterns that related to their origin. The uranium mineral ores collected from various parts of Kenya were successfully grouped into their mineral mines (origin) applying PCA to selective spectral regions. NF signatures associated with uranium molecules in uranyl nitrate, uranyl sulphate, uranyl chloride and uranium trioxide samples were identified at 865 cm-1, 868 cm-1, 861 cm-1, and 848 cm-1respectively using LRM (laser λ= 532 nm, 785 nm). Spectral imaging using these signatures on sample spiked with trace uranium (150ppm), HBRA soil samples and uranium mineral ore samples demonstrated the distribution of uranium molecules. Thus, ML techniques, in combination with laser-based techniques and
spectral imaging techniques, have the potential to not only perform rapid, direct, minimally intrusive qualitative and quantitative analysis of trace uranium (typical of NF), but also aid in the attribution of uranium ore surrogates to their origin and distribution of uranium molecule through the different layers of these samples.