Daniel Hellfeld, a Research Scientist in the Applied Nuclear Physics Program, traveled to the Pacific Northwest in April 2022 to participate in a week-long measurement campaign to collect gamma-ray spectra from Type-30B UF6 cylinders used to store and transport uranium of various enrichments for use in nuclear fuel assemblies (see Figure 1). During the week, 12 different cylinders with U-235 enrichments ranging from 0.2% (depleted) to 4.95% (low-enriched) were measured using a room-temperature operated, semiconductor-based, 3D position-sensitive pixelated CdZnTe (CZT) detector system available commercially from H3D Inc.
The collected data (see Figure 2) are being used in the development of new algorithms to improve the spectrometric performance (e.g., energy resolution) of H3D’s pixelated CZT detectors. The International Atomic Energy Agency (IAEA) has expressed interest in these detectors as a replacement for medium-resolution scintillators used for non-destructive assay measurements in nuclear safeguards such as facility inspections and material declaration verification. LBNL is exploring the use of data-driven machine learning approaches (e.g., Non-negative Matrix Factorization) to learn latent spectral features in pixelated CZT detector data in order to optimally cluster and combine list-mode events for overall improved energy resolution.