Isotopic gamma lines for identification of shielding materials

TL;DR

This paper presents a brute-force statistical framework for identifying shielding materials using high-resolution gamma spectroscopy and semi-analytic uncollided flux expressions for nested spherical and cylindrical geometries. By extracting discrete gamma-line intensities and exhaustively evaluating trial material configurations against a physics-informed loss, the method accurately identifies material combinations when geometry is known, performing best for one or two unknown materials and providing a ranked list of candidates for more complex cases. Validation against simulated and experimental data (DU shells and BeRP ball) demonstrates viability and highlights the impact of model mismatch and Poisson noise, pointing to improvements in continuum subtraction and extension to 3D geometries. The approach offers a practical pathway for material discrimination in security and industrial contexts where traditional dual-energy CT is insufficient, leveraging line-pattern information from HPGe spectroscopy.

Abstract

Identifying the constituting materials of concealed objects is crucial in a wide range of sectors, such as medical imaging, geophysics, nonproliferation, national security investigations, and so on. Existing methods face limitations, particularly when multiple materials are involved or when there are challenges posed by scattered radiation and large areal mass. Here we introduce a novel brute-force statistical approach for material identification using high spectral resolution detectors, such as HPGe. The method relies upon updated semianalytic formulae for computing uncollided flux from source of gamma radiation, shielded by a sequence of nested spherical or cylindrical materials. These semianalytical formulae make possible rapid flux estimation for material characterization via combinatorial search through all possible combinations of materials, using a high-resolution HPGe counting detector. An important prerequisite for the method is that the geometry of the objects is known (for example, from X-ray radiography). We demonstrate the viability of this material characterization technique in several use cases with both simulated and experimental data.

Figures (10)

• Figure 1: General problem formulation. Left: an illustration of an X-ray radiograph of a scene with image segmentation. Middle: spherical object of interest with multiple shielding. Right: high-resolution spectrum and lines
• Figure 2: Top: Distribution of the ratio of counts in two strongest lines of $^{238}$U, at the energies $E_1 = 1000.99$ keV and $E_2=766.37$ keV, for a range of various shielding materials. The red curve indicates normal PDF of the true value of the line counts ratio for Al, with standard deviation $\sigma=0.01$, or 1% (an error corresponding to about 10000 counts in both lines). Bottom: 2D distribution of relative line counts, for the $E_2/E_1$ ratio (horizontal axis) vs the $E_3/E_1$ ratio (vertical axis), where $E_3=94.67$ keV. The latter ratio is plotted on a log scale. A sequence of highly eccentric nested ovals centered around the ratio for Al represents PDF in 2D space of two line ratios, with the same relative standard deviation $\sigma=1\%$, at the levels of 0.003, 0.05, and 0.32. The levels correspond to the standard 1$\sigma$, 2$\sigma$, and 3$\sigma$, confidence intervals. Materials include polyethylene (PE), polycarbonate (PC), air, boron carbonate (B$_4$C), steel, lithium-deuterium (LiD) compound, and elemental materials labelled by their usual chemical notation.
• Figure 3: Sketch illustrating integration geometry.
• Figure 4: Experimental spectra for 1/4" DU shell with 1/2" Al shielding (top), and BeRP ball with 1" Ni shielding (bottom). The red lines indicate integrated peak area counts extracted with PeakEasy (see main text for details).
• Figure 5: Uncollided total (top) and simulated detected (bottom) count rate from a spherical 1/4" DU shell, shielded by 1/2" of Al. The red lines represent line fluxes as reported by GADRAS for this object, either uncollided (top), or from integrating the peak areas (bottom). Black circles and blue crosses represent our trial analytic estimates for Al and C, respectively. The black solid curve on the right shows the continuous spectrum, simulated for a HPGe detector using GADRAS. The lower part of each plot zooms on the ratio of trial to test line fluxes (in percentage of difference from 1), with the right panel also showing the uncertainties for each ratio due to the peak fitting. It should be noted that some of the lines do not have analytic estimates. This is because they are only computed for the dominant isotope, $^{238}$U, while other isotopes are also present in the DU composition. Additional lines can be identified in the spectra, such as 511 keV line visible in both panels, are not directly produced by radioactive decays, but rather other processes such as neutron captures or inelastic scattering on Al (in particular case of 511 keV line, by $e^+ e^-$-annihilation).