Online program Overview Session: NSS-22

Abstract

Spectral templates used for radiological anomaly detection and isotope identification are most commonly derived from simulation, yet real-world source encounters are subject to a diversity of shielding configurations, scattering geometries, and encounter dynamics that simulation does not fully capture. The PANDAWN sensor network — fourteen statically deployed, multi-modal radiological detection nodes distributed across Chicago, IL — provides a unique opportunity to learn spectral templates directly from real-world data. In this talk, we present the application of regularized Non-Negative Matrix Factorization clustering to anomalous gamma-ray spectra, which recovers multiple spectrally distinct templates for nuisance isotope encounters observed across the network, each corresponding to physically interpretable differences in source attenuation and encounter geometry. Examining how these cluster components trigger isotope identification during known source encounters reveals distinct temporal patterns, further demonstrating that the learned templates capture meaningful structure in the real-world encounter environment. We will discuss what these results reveal about source encounters across the network, present an assessment of template transferability across nodes, and examine how developing a taxonomy of data-driven source templates of this kind can improve detection and isotope identification performance.

Abstract

Geometrical event removal (GER) via a position-sensitive multianode photomultiplier tube (MAPMT) is used to enhance mixed-field pulse-shape discrimination (PSD) in an organic EJ-276 scintillator and gamma spectroscopy peak-to-continuum ratio in a CeBr3 crystal scintillator. Significant (~21.3%) improvements in PSD figure-of-merit (FoM) are shown without the need for an energy cut-off, allowing for real-time spectrum-agnostic enhancement of time-domeain PSD methods. This allows greater separation of the neutron and gamma plumes at low energies without the need for a low-energy cutoff, thereby preserving the ability to measure across the full range of neutron spectra. The position sensitivity of the detector has also been exploited to create 2D flux map images of gamma fields attenuated through various lightweight 3D-printed collimators, and to identify the isotope and relative direction of individual point gamma sources in a multi-isotope combined field.

Abstract

Robust tritium monitoring is essential for fusion energy systems, where rapid detection of low-level airborne releases is needed to support safe operation and accident response. However, intense neutron fields in fusion environments activate surrounding air and produce radioactive products that can contribute beta, gamma-ray, and X-ray backgrounds, complicating tritium detection. In this work, we investigate the impact of air-activation products on a proposed tritium detection concept that combines a thin plastic scintillator beta cell with a surrounding NaI(Tl) anti-coincidence shield. The plastic scintillator provides sensitivity to the low-energy beta particles from tritium decay, while the NaI(Tl) detector enables rejection of coincident photon backgrounds. Monte Carlo simulations were performed using MCNP6 to estimate the detector response to tritium and relevant air-activation radionuclides, including beta emitters such as 13N, 14C, 39Ar, 39Cl, and 41Ar, as well as associated gamma-ray, X-ray, and bremsstrahlung contributions. A 1.3-mm-thick plastic scintillator cell was modeled to match commercially available thin-walled beta cells used in radioxenon detection. Preliminary results show that several air-activation products produce signals in the same low-energy region as tritium and may represent important background contributions, particularly when gamma rays are not detected in coincidence. Subsequent simulations will be performed using the MCNPX-PoliMi code to accurately treat the correlated detection of electrons and photons in the plastic and NaI(Tl) cells.

Abstract

We present Attention-Biased Additive Point Source Localization (ABAPSL), a fast reconstruction method for localizing multiple radioactive point sources from mobile gamma-ray measurements. The method extends our prior gridded multi-source localization work, which found sources one at a time and added confirmed sources into a growing background model. ABAPSL instead re-fits all active sources at each reconstruction while using an attention metric to place fine test points only where the data support high sensitivity or high source likelihood. This provides local 10cm-scale refinement without requiring a uniformly fine grid over the full survey area. We also introduce multi-resolution spatial measurement binning, replacing fixed time bins with trajectory-aligned spatial bins that remain fine near candidate sources yet are coarser in background-dominated regions. The resulting system matrix is small enough for online reconstruction on mobile computers such as the onboard compute used in a Localization and Mapping Platform (LAMP) system. We demonstrate the approach with simulations around multiple radioactive point sources, with LAMP measurement validation to follow.

Abstract

Gamma Reality Inc has developed and demonstrated, to our knowledge, the first robotic deployment of Associated Particle Imaging (API) for the application of rapid-assessment, detection and 3D imaging of buried landmines. We utilize a deuterium-tritium (DT) reaction to generate momentum-tagged neutrons, at rates up to 5E8 n/s, that stimulate characteristic gamma-ray emissions. The result is 3D voxel-spectral information that we combine with contextual sensed scenes, via 3D LiDAR, providing a centimeter resolution map of isotopic compositions within a specific scene. There are many non-destructive assay applications for such a system, however we identify landmine detection as our first application space. Modern landmine detectors are sensitive to secondary characteristics of mines, such as low mass-metal content within the mines, and struggle with false-positive probabilities. Our method targets the primary characteristics of mines, which has the potential for a very large reduction in false positive rate compared to current methods, which can be 30 to 1 or often worse. In this paper we present results of our field demonstrations and domestic tests, where we identify surrogate mine materials constructed of ammonium nitrate and carbon against common background and nuisance objects (organic materials, metal objects, & etc). We present results for 20 min, 5min and 1min detections over mine surrogates in soil. We will also present the results of field-testing over TNT and other explosive fills in replica casings.

Abstract

Dosimetry under the high dose-rate conditions is required such as internal survey of the Fukushima Daiichi Nuclear Power Plant for the decommissioning.  We have developed a dose-rate monitor consisting of optical fiber and scintillators. Since the optical fiber has lower collection efficiency of the scintillation photons compared to that of direct contacting between a scintillator and photomultiplier tube, higher light output was required. Moreover, long emission wavelength of over 650 nm is required to obtain sufficient sensitivity. Cr-doped oxides are expected to have emission wavelength in the red or infra-red region, and garnet crystals can be optimized the emission wavelength by changing crystal fields. Thus, we focused on Cr-doped (Y, Gd)₃(Al, Ga)₅O₁₂ materials as red oxide scintillators, and Gd₃(Ga_0.995 Cr_0.005)₅O₁₂ (Cr:GGG) single crystals grown by the micro-pulling-down method had an emission wavelength of Cr:GGG was 731 nm, which achieved the emission goal. Using this Cr:GGG sample we have developed the real-time dose monitor, and  such system was found to operate the survey even under the high dose rate of around 1 kSv/h. Moreover, we evaluated the absorption dose with different gamma-ray energies with geant4 (gamma-ray energy response), and the result was consistent with gamma-ray irradiation test with ¹³⁷Cs and ⁶⁰Co sources.