The verification of the absence of special nuclear material (SNM) may be a key requirement in potential future warhead treaty verification, but analysis of gamma-ray spectral features could reveal sensitive information that a host may not allow. This paper will report on extending the findings of the Simple Source Separator project, which identified a promising gamma-ray counting method for indicating a difference between nuclear objects (plutonium, highly enriched uranium) and nonnuclear objects (depleted uranium, low-enriched uranium) while protecting sensitive information. Here we adapt this method for a portal-sized system where an item could be moved through the portal and a measurement made in a short period of time. The algorithm under development uses the total counts in multiple, carefully selected energy windows to distinguish nuclear objects from non-nuclear objects and to determine the absence of nuclear objects. It leverages spectral shape without requiring detailed analysis of the spectrum, protecting sensitive information. The earlier effort focused on developing the described energy window algorithm using a hand-held sodium iodide detector as a compromise between detection efficiency, energy resolution, and cost. When moving to a portal-sized application, materials that are available in larger sizes at lower cost have been considered. Polyvinyl toluene (PVT) plastic is a common choice for portal monitor applications due to its availability in large sizes and good sensitivity to gamma radiation. However, the energy resolution and stopping power of PVT are both worse than sodium iodide. Since the energy window algorithm sums counts over wide spectral regions, the energy resolution may not necessarily be a large detriment. However, the lack of stopping power will lead to less ability to use any counts from higher energy gamma rays thus limiting sensitivity to differences in the spectral shape. Bismuthloaded PVT has been reported in the literature and has a higher stopping power due to the presence of the high-Z element bismuth but at the cost of reduced light output compared to PVT (about 35%). All three proposed detector materials have been simulated in portal monitor sizes in MCNP and GADRAS and simulated spectra for a range of materials, geometries, and shielding have been generated. These simulated spectra were used to evaluate the performance of the energy window algorithm. As expected, due to the different stopping powers as a function of energy for each detector material, the energy windows had to be customized for each material. The feature selection using neighborhood component analysis for classification algorithm (fscnca) was used to identify the best performing energy windows for each material. Examining the distribution of algorithm output for different object types, sodium iodide still appears to be favored but bismuth-loaded PVT is exhibiting some separation between different material types. The research is ongoing with the algorithm still being optimized for the detector materials. The latest status of the project progress will be reported.
Year
2024
Abstract