Measurements of the geometric configuration of objects and their material composition are needed for nuclear treaty verification purposes. We experimentally demonstrate a simple method based on monoenergetic fast neutron transmission to realize crude imaging of the geometric configuration of special nuclear material, confirm its fissionable content, and obtain information on its approximate fissile mass. In the experiment, we used monoenergetic neutrons from D(d, n)3He and T(d, n)4He reactions and a linear array of liquid scintillation detectors to perform spectroscopic neutron imaging of up to 13.7 kg of highly enriched uranium in a spherical geometry. We also show an example of detection of material diversion and confirm the presence of fissionable material based on the measurement of high-energy prompt fission neutrons, including estimating the quantity of material from the comparison of measured and predicted fission neutron emission rate. The combination of crude imaging and fissionable material detection and quantification in a simple approach may be attractive in certain treaty verification scenarios.

The effectiveness of international treaties aimed at preventing nuclear weapons proliferation and promoting disarmament relies on the technological capability to monitor and confirm the compliance of all participants. With the recently-signed Joint Comprehensive Plan of Action on the Iranian nuclear program, as well as the new START treaty with Russia, there is a clear need for a robust framework of verification technology. For example, the multi-stage dismantlement process includes warhead authentication, separation of the nuclear material from non-nuclear components (such as the delivery vehicle and high explosives), and storage of the special nuclear material (SNM) cores in an appropriate container, all of which require evidence of adherence to treaty protocols. Measurement of the characteristics of SNM cores such as geometry, composition, and fissile mass can provide the necessary information to ensure compliance with such treaty regulations and general safeguards standards.1–3 

However, the design of a verification system is complicated by the fact that such measurements necessarily involve the collection of highly classified data. As such, the protection of state secrets represents a significant hurdle that must be cleared for a verification technique to be considered viable. While this problem has traditionally been addressed through engineered information barriers, which prevent the inspector from directly observing the classified information being measured, this approach involves a high degree of complexity and susceptibility to tampering through the use of information trapdoors to falsify results or leak sensitive information.2,4 Current information barrier systems can also be expensive to implement and reduce confidence in verification measurement results.5 The political reality associated with such mechanisms thus presents a significant barrier to their adoption and implementation. As a result, there is considerable interest the application of zero-knowledge protocols, in which measurements do not record any sensitive information directly, but instead may form part of a differential comparison against a declared standard or template.4 Such templates usually consist of a complex radiation signatures that are used to “fingerprint” SNM components.2,6 For example, with the aid of a 252Cf active-interrogation source, analysis of an array of induced neutron signals in the time and frequency domains has been used to successfully distinguish and identify nuclear weapons components based on reference signatures.7,8 In addition to more complex signature-matching techniques, template-matching methods have been proposed that involve detector arrays which are pre-loaded with geometric or other characteristic data in such a way that only the differential information between the test object and the template is ever recorded by the measurement system.4 This paper focuses on characterizing the performance of a measurement technique that has a potential for future incorporation into such future zero-knowledge protocols.

There are a number of potential methods by which a host may attempt to deceive inspectors, including replacement, diversion, or dilution of nuclear material. By measuring the type of material, as well as its composition, fissile mass, and geometry, a verification system can detect potential anomalies such as spoofs (false geometries or materials designed to mimic the true SNM core) or diversion of material.2 One measurement technique which is particularly well-suited to the detection of geometric anomalies is transmission radiography. Radiographic measurements are typically performed with high-energy photons or fast neutrons, though other types of particles, such as muons, are sometimes used.1 Photon radiography leverages the Z-dependence of the mass attenuation coefficient to localize regions of high-Z or dense material in the transmission image. Sources that span a multitude of photon energies can be used to determine the effective Z (Zeff) of the material by exploiting differences between either the photoelectric and Compton-scatter cross-sections (at lower energies) or Compton-scatter and pair-production cross-sections (at higher energies).9,10 Elemental identification through photon radiography is limited by the fact that a thick mass of low-Z material may attenuate source photons similarly to a thinner mass of high-Z material. Although accuracies to within 5–10% when determining Zeff have been achieved,11 difficulties are still present in distinguishing SNM from similarly dense, high-Z materials such as lead or tungsten.

Neutron transmission radiography using fast neutrons has also been used to construct high-resolution geometric images of complex configurations of mixed materials.12 In a process analogous to the one employed for photon radiography, the composition of the material can be reconstructed based on the neutron interaction cross-section, which varies greatly with energy for lighter elements.10 Simple transmission radiography using fast neutrons has been used to image SNM components, and the performance of such a system in restricted-dose or high-background situations can be further improved through use of the associated-particle technique.13 In addition to transmission radiography, many established techniques exist for neutron imaging based on passive or induced emissions.14–16 Fast-neutron imaging using coded-aperture measurements can precisely locate neutron sources, and may be useful for verification applications such as warhead counting.14 High-fidelity imaging of SNM samples such as plutonium mixed oxide (MOX) fuel has also been achieved through the use of emitted-neutron computed tomography in a collimated slit imaging system.15 While the sensitivity of such high-resolution neutron imaging poses problems for information barrier development, algorithms which prevent the storage of image data may increase the viability of such systems in a verification setting.17 Alternatively, a low-resolution fast-neutron radiograph may be attractive because significant diversions could be detected without collecting detailed geometric information. Such a radiographic image could play a part in a geometric template-based verification method, which has shown promise for application to a zero-knowledge protocol.4 

It is also desirable for verification measurements to confirm whether SNM is present and in what quantity. While SNM may be difficult to distinguish from other dense, high-Z materials using simple transmission radiography alone, it can be readily identified from the presence of fission neutrons induced by a well-chosen interrogation source. By isolating the induced-fission neutron signal from source/background interference, the relative quantity of fissile material may also be deduced. More accurate determinations of fissile mass have been achieved by interrogating SNM with continuous or pulsed neutron beams and measuring neutron multiplicity moments.18,19 Imaging based on induced-fission neutrons has also been demonstrated using a pulsed D-T interrogation source and time-of-flight measurements to discriminate active background.16 While much of this prior work has demonstrated excellent radiographic and neutron-source imaging capabilities, the increase in performance often comes with added system complexity, specialization, and cost. In this paper, we present a simplified spectroscopic neutron radiography system that is cost-effective, easily implemented, and constructed from readily-available components. We demonstrate that such a basic design can be used to perform crude geometric imaging, confirm the presence of fissile material, and estimate its quantity. We also demonstrate the detection of material diversion using both transmission images and the abundance of the detected neutrons attributed to nuclear fission.

Experimental data were collected in a series of measurements over a four-day period at the Device Assembly Facility (DAF), Nevada National Security Site. All measurements were made using a horizontally oriented linear detector array of eight 2×2 inch EJ309 organic liquid scintillators coupled to Hamamatsu R7724 photomultiplier tubes, located at a height of 89.5 cm from the floor, and evenly spaced over an array length of 106.5 cm. The detectors were powered by two CAEN DT-5533 high-voltage power supplies, and the signals digitized using a CAEN DT-5730 14-bit 500 MS/s desktop waveform digitizer.

The target used for transmission radiography measurements was a spherical configuration of the Rocky Flats highly enriched uranium (HEU) shells. The shells have a bulk density of 18.664 g/cm3 and an isotopic content listed in Table I.20 

TABLE I.

Isotopic Content of Rocky Flats HEU.

Uranium IsotopeWeight Percentage (1971)
233Not Recorded 
2341.02 
23593.16 
2360.47 
2385.35 
Uranium IsotopeWeight Percentage (1971)
233Not Recorded 
2341.02 
23593.16 
2360.47 
2385.35 

Each individually numbered shell is hemispherical in shape, and each consecutive pair of shells (1-2, 3-4, etc.) represents two nearly identical hemispheres that form one spherical shell. Two different configurations of the Rocky Flats shells were used in this experiment: a “full” configuration including shells 01–24, and a “half” configuration including only shells 13–24 (inner shells 01–12 were removed, leaving an empty core). Figure 1 shows a schematic representation of the shell geometry, and the exact dimensions for each shell can be found in Reference 20.

FIG. 1.

Geometric configurations of the Rocky Flats HEU shells: (a) full configuration; (b) half configuration.

FIG. 1.

Geometric configurations of the Rocky Flats HEU shells: (a) full configuration; (b) half configuration.

Close modal

Two monoenergetic neutron sources of 2.45 MeV and 14.1 MeV neutrons produced from D(d, n)3He (DD) and T(d, n)4He (DT) fusion reactions were used, respectively. A Thermo Scientific model MP 320 neutron generator with interchangeable DT and DD tubes was employed with an approximate isotropic neutron yield of 106 and 108 neutrons/s for the DD and DT reactions, respectively. The operating parameters for the neutron generator using the DD tube were 95 kV and 60 μA, while in the DT tube case the operating parameters were 70 kV and 45 μA.

The detectors were tested and calibrated using a 252Cf source and a 137Cs source. Neutron transmission measurements were subsequently made using the DD source with the target in the “full” configuration. The laboratory setup can be seen in Figures 2–3.

FIG. 2.

Laboratory schematic for DD transmission measurements. DT measurements used a geometrically identical setup.

FIG. 2.

Laboratory schematic for DD transmission measurements. DT measurements used a geometrically identical setup.

Close modal
FIG. 3.

View of the laboratory setup.

FIG. 3.

View of the laboratory setup.

Close modal

The DD source and HEU target were placed on a table at a height of 77 cm, such that the height of the source and target centers aligned with the central height of the detectors to within 1 cm. The source-target distance was 30 cm from center to center, and the source-detector distance was 155.5 cm from center to center. Four transmission measurements were taken for a cumulative measurement time of approximately 3 hours, for an average measurement time of about 45 minutes. The HEU target was then replaced with a hollow tungsten sphere with an inner radius of 6.4 cm and an outer radius of 8.9 cm, and a 35–minute measurement was made to obtain transmission data for a non-fissile material. Measurements of the raw DD source (beam on, no target) and background (beam off, no target) were also recorded for about 17 minutes each.

The following measurements were made on the HEU target in the “half” configuration using both the DD and DT sources. Five DD transmission measurements of the HEU target were taken over a total measurement time of 3 hours and 20 minutes, for an average measurement time of 40 minutes. Three DT transmission measurements were then taken over a total of 21 minutes, at an average measurement time of about 7 minutes. Much shorter measurement times were used for the DT generator because the greater source strength allowed for much quicker accumulation of statistics. The dimensions of the experimental setup remained unchanged from the prior measurement in the “full” configuration.

The final set of DT transmission measurements was conducted on the “full” HEU target, as well as two shielded configurations. The first shielded configuration used a hollow polyethylene sphere with a thickness of 3.81 cm placed around the target with a gap of 2 cm between the HEU and polyethylene shield. The second shielded configuration used a hollow tungsten shield with a thickness of 2.54 cm and a 0.7 cm gap between the shield and HEU sphere. Total measurement times were 85 minutes for the bare target, 60 minutes for the poly shield, and 15 minutes for the tungsten shield.

Simulations of the laboratory measurements were made in the Geant4 framework21 with an accurate description of the experimental geometry. The HEU target was simulated using the isotopic contents listed in Table I and the spatial dimensions for each shell.20 For the EJ309 organic liquid scintillators, a density of 0.959 g/cm3 and an atomic composition of 9.85% H and 90.15% C by weight were assumed.22 Conversion of neutron interactions into light output pulses for the 2-inch detectors utilized the exponential model described in Reference 23, which was then tailored to produce light output spectra that matched experimental results. The DD and DT generators were modeled as 2.45 MeV and 14.1 MeV monoenergetic neutron sources, respectively, and projected into a cone encompassing the full detector array. All Geant4 simulations utilized the QGSP_BERT_HP physics library, and each simulation generated a total of 100 million events.

Detector light output was calibrated using the Compton edge for 137Cs characteristic gamma rays at 478 keVee. The location of the edge was determined through comparison of simulated light output spectra with and without light output resolution broadening, with a resulting crossing point at 90% of the Compton maximum. The light output resolution function was defined using a previously described parameterization method.24 

CAEN’s DPP-PSD Control Software was used to collect and save two preset integral regions, Qlong and Qshort, for each event. Based on differences in the light output pulse profile for neutron and photon detection events, comparison of the areas of these two integral regions provides the basis for discrimination between both particle types. The waveform integration bounds were set to [ts, ts+28 ns] and [ts, ts+160 ns] for Qlong and Qshort, respectively, where ts is the start time of the waveform (trigger position–gate offset). The pulse shape parameter (PSP) was calculated as

PSP=(QlongQshort)/Qlong.
(1)

Figure 4 shows an example PSP distribution for a single detector during a DD neutron transmission measurement with the full HEU target in place. A threshold of 200 keVee was applied, and neutron pulses were selected by fitting a Gaussian model over the recoil region (PSP range of 0.27–0.45). The Gaussian fits for the neutron and photon regions are also shown in Fig. 4, with a figure of merit of 1.50 for the two curves. The neutron region was then defined by a one-sigma cut around the centroid of the Gaussian fit for each detector channel. At high light outputs, curvature in the photon region caused it to overlap with the neutron-pulse PSP range, so an upper light output cut of 4 MeVee was also applied.

FIG. 4.

Left: Example DD measurement PSP distribution for a single detector channel. Red lines indicate the neutron cut region. Right: Gaussian model fits for neutron and photon PSP regions.

FIG. 4.

Left: Example DD measurement PSP distribution for a single detector channel. Red lines indicate the neutron cut region. Right: Gaussian model fits for neutron and photon PSP regions.

Close modal

For the DT neutron measurements, significant pile-up effects in the lower light output range caused a curvature in the PSP distribution. To eliminate this skewing effect and provide a more robust definition of the neutron PSP range, a light output threshold of 4 MeVee was applied. The PSP range for each detector channel corresponding to neutron pulses was then defined using the same process used for the DD data, though the PSP range where the Gaussian fit was applied was much smaller (0.16–0.22). The resulting figure of merit was 1.60. Figure 5 shows the PSP distribution for the DT measurements for a single detector both before and after the application of the 4 MeVee light output threshold. Gaussian model fits for the neutron and photon regions are shown in Fig. 6.

FIG. 5.

Example DT measurement PSP distribution for a single detector channel. At right is the distribution with a 4 MeVee threshold applied. Red lines indicate the neutron cut region.

FIG. 5.

Example DT measurement PSP distribution for a single detector channel. At right is the distribution with a 4 MeVee threshold applied. Red lines indicate the neutron cut region.

Close modal
FIG. 6.

Gaussian model fits for neutron and photon PSP regions using the DT source and a 4 MeVee light output threshold.

FIG. 6.

Gaussian model fits for neutron and photon PSP regions using the DT source and a 4 MeVee light output threshold.

Close modal

Neutron transmission data were used to construct one-dimensional image profiles for each tested target configuration. This technique could be relatively easily incorporated into a zero-knowledge protocol, as a comparison of the measured transmission profile to a pre-loaded template for a particular target would allow a differential determination to be made without revealing the geometric characteristics of the target. Figure 7 shows the DD neutron transmission images for the full and half target configurations, representing a hypothetical differential measurement. Each channel corresponds to a single detector in the array, forming an eight-pixel one-dimensional transmission profile of the target object based on the total neutron count rate seen by each detector. Even for such crude resolution, the two scenarios are clearly distinguishable, and the diversion of material is apparent.

FIG. 7.

Experimental DD neutron transmission images for full and half HEU target configurations. Statistical error bars are included in the plot, but are too small to be easily visible.

FIG. 7.

Experimental DD neutron transmission images for full and half HEU target configurations. Statistical error bars are included in the plot, but are too small to be easily visible.

Close modal

Validation of the experimental results and further explorations of potential measurement system capabilities were conducted by Geant4 simulation. To begin, reconstruction of the laboratory conditions in simulation was beset by a number of uncertainties, including an unknown neutron flux from the DD source and the magnitude and spectrum of the room-return neutrons. Based on experimental measurements of the raw DD source and initial simulations of absolute detector efficiency at each channel position, the source neutron flux was estimated to approximately 106 neutrons/s. However, using this flux as the basis for simulation, the simulated and experimental data still exhibit a significant disagreement.

Since the room geometry is complex and the information is lacking to model it accurately, an alternative method is needed to account for the effect of room return neutrons on the measurements. We hypothesized that raising the light output threshold would have a disproportionate effect in discriminating lower-energy scattered neutrons, and thus greatly suppress the room-return contribution to the measured spectra. Figure 8 shows a comparison of experimental and simulated results for increasing light output threshold increments of 100 keVee from the initial threshold of 200 keVee up to 600 keVee, which was the upper bound for the light output cut before measurement and simulation statistics start to significantly deteriorate.

FIG. 8.

DD transmission images, experiment (black) vs simulation (red). Light output cuts: (a) 200 keVee; (b) 300 keVee; (c) 400 keVee; (d) 500 keVee; (e) 600 keVee. Progressively increasing light output cuts show an increased suppression of room-return contribution.

FIG. 8.

DD transmission images, experiment (black) vs simulation (red). Light output cuts: (a) 200 keVee; (b) 300 keVee; (c) 400 keVee; (d) 500 keVee; (e) 600 keVee. Progressively increasing light output cuts show an increased suppression of room-return contribution.

Close modal

The increased light output cuts demonstrate effective suppression of the room return, but not all of the room return is rejected by this method. A significant amount of high-Z material is present in the experimental environment, and even for abundant lower- and mid-Z concrete constituents such as silicon or oxygen, elastically scattered neutrons can still retain a large fraction of their initial energy. Furthermore, there is some spread in the energy of the neutrons emitted by the generator. For neutrons emitted at an angle of π/2 from the direction of the generator ion beam, which was the configuration used for this experiment, the energy spread is small and the neutron energy is near 2.45 MeV, but neutrons emitted in the same direction as the ion beam can vary in energy from about 2.7 MeV to as much as 3.1 MeV.25 As a result, the energy range of room-return neutrons extends above the full energy of the primary transmission DD neutrons, making it impossible to discriminate room return entirely using light output cuts alone.

In addition to their effects on the detector array, scattered neutrons from the DD source can also return to the HEU object to cause additional fission events. This would lead to a higher fission rate than predicted from fast fission alone, and may account for a large part of the initial discrepancy between simulated and experimental data. To validate this idea, the effect of a uniform, randomly-directed scattered neutron flux over the HEU sphere was simulated in Geant4, which resulted in a nearly uniform increase in the observed neutron count rate across the detector array.

It is also apparent from the comparison of simulated and experimental data that an additive correction for room return and an increased fission rate does not fully account for the discrepancy, as the two data sets are not separated by a near-constant offset. Therefore, the original estimation of DD source neutron flux, which has a scaling effect on the count rate, also needs to be adjusted. Notably, the discrepancy between simulation and experiment at higher light output thresholds was not fully accounted for by a constant scaling factor either, meaning that the differences were most likely due to some combination of the uncertainty in the knowledge of the DD generator flux and the magnitude of additive effects from room return and fission from scattered neutrons. To account for these effects, the simulated neutron transmission rate, Ntrans, was adjusted according to two correction parameters. The first is a source strength scaling factor, S, which is intended to provide a more accurate estimate of the overall DD generator flux, and an additive factor, R, which accounts for the effects of room return and fission from the scattered neutron flux. The model for the detected neutron count rate is thus given by:

Ntotal=SNtrans+R.
(2)

The correction in Eq. 2 shows a linear relationship between the total additive effects and the intensity of the source. This is expected since the contributions of neutron multiplication within the target and scattering in the room should scale linearly with the source intensity. The cosmic neutron background was measured to be negligible, so it is ignored by the model. Notably, in the presence of a significant cosmic neutron background, the model would need to be adjusted to account for an additional background that would not exhibit a linear relationship to source intensity.

Once determined, the model was then optimized for the combination of S and R that led to the best least-squares fit to the experimental data. From the optimized neutron flux scaling factor, the DD generator flux was estimated to be 1.5 × 106 neutrons/s. A separate validation measurement was later performed at the University of Michigan, in which the neutron count rate was recorded using the same model of DD generator and the same size and type of detector used in the DAF experiment. The EJ309 detector efficiencies were calculated based on Geant4 models, resulting in an intrinsic neutron detection efficiency of about 30% for neutrons in the energy range produced by the generator. Based on this detection efficiency, the total neutron flux produced by the DD generator during the validation measurement was estimated to be 1.6 × 106 n/s, which is in keeping with estimates from the DAF laboratory setting. For the estimated DD generator flux, the additive factor was determined to be about 2.4 neutron counts per second. Based on Geant4 simulation, the scattered neutron flux entering the HEU object would need to be about 2–3 orders of magnitude lower than the DD generator flux in order to achieve this count rate from fission events induced by scattered neutrons. Further simulations were carried out to determine whether this magnitude of scattered flux was consistent with the DAF laboratory setting. Using the approximate geometry for significant laboratory objects such as the aluminum table and concrete floor, the total scattered neutron flux through the HEU sphere was determined to be approximately 0.3% of the total DD source flux. Therefore, the predicted flux from scattered neutrons is sufficient to cause an appreciable effect on the overall count rate, providing further physical justification for the additive factor in the mathematical adjustment model.

Figures 9–11 show the comparison of simulated and experimental data at a 600 keVee light output threshold after application of the corrected neutron flux estimate and additive factors. After correction, the simulations are in good agreement with the experimental results, with Pearson correlation coefficients (PCC) of 0.963 and 0.956 for the “full” and “half” configurations, respectively. There is still a significant disagreement between the simulated and experimental results in the outer detector channels, which is likely due to uneven room-return contributions from large objects at either side of the array, which cannot be well approximated by a constant additive term. For example, there was a large coded aperture sheet of high-density polyethylene located to one side of the experimental setup, which may be partly responsible for the higher count rates in channels 1 and 2. However, as Figure 3 shows, there are other objects (e.g. metal tables) that may also contribute significantly to scattering. When considering a region of interest that excludes the outermost channels in the detector array to focus on the area where the target object’s shape is defined, the PCC improves to 0.997 for the “full” configuration, and 0.991 for “half”.

FIG. 9.

Experimental and simulated DD neutron transmission images for the HEU target in “full” configuration.

FIG. 9.

Experimental and simulated DD neutron transmission images for the HEU target in “full” configuration.

Close modal
FIG. 10.

Experimental and simulated DD neutron transmission images for half HEU target.

FIG. 10.

Experimental and simulated DD neutron transmission images for half HEU target.

Close modal
FIG. 11.

Experimental and simulated DD neutron transmission images for the DD source only (no object in place).

FIG. 11.

Experimental and simulated DD neutron transmission images for the DD source only (no object in place).

Close modal

Having established that the transmission imaging system can detect a relatively large diversion of material, it is of interest to determine what degree of spatial resolution may be achieved. To this end, neutron transmission was simulated for each incremental shell configuration between the full and half configurations. Figure 12 shows the simulated transmission results for each two-shell increment (corresponding to one spherical layer) for a measurement time of approximately 70 minutes.

FIG. 12.

Calculated DD neutron transmission images for incremental shell configurations. The legend indicates the shell numbers that have been removed from the full configuration of shells (1–24).

FIG. 12.

Calculated DD neutron transmission images for incremental shell configurations. The legend indicates the shell numbers that have been removed from the full configuration of shells (1–24).

Close modal

Each incremental configuration, corresponding to an average inner-radius difference of 0.3–0.4 cm, is individually distinguishable within statistical error bounds. Finer resolution is likely achievable with longer measurement times, with the limiting factor being the total time allotted for measurement in a particular verification scenario. Another limit on resolution comes from the geometry of the target. For smaller amounts of diverted material, the size of the empty inner cavity is small relative to the spacing of detectors in the array. Since the two central detectors are not placed exactly along the central axis of the target, the change in interceding material for small-cavity configurations (e.g. between the 1–2 and 1–4 configurations in Figure 12) is less pronounced than it would be for a more centrally located detector. While it is obvious that a detector array with more pixels would provide a more detailed transmission image, this suggests that a greater number of centrally located pixels would also provide better discrimination of shell configurations, especially when the amount of removed material is small.

Figure 13 shows the corresponding DT neutron transmission images for the full and half shell configurations with a light output threshold of 4.0 MeVee. The two scenarios are clearly distinguishable, with an even greater separation compared to the DD source measurements.

FIG. 13.

Experimental DT neutron transmission images for full and half HEU target configurations.

FIG. 13.

Experimental DT neutron transmission images for full and half HEU target configurations.

Close modal

Experimental results were once again validated with simulation in Geant4. Simulated results for the DT neutron measurements were scaled using the same modeling and optimization techniques applied to the DD data. For the Thermo Scientific MP-320 DT generator used in this experiment, the expected neutron yield for operational settings of 70 kV and 50 μA is roughly 5–7 × 107 neutrons/s.26 This agrees well with the source neutron flux of 5.0 × 107 neutrons/s estimated by the simulation scaling model. Figures 14–15 show the comparison of simulated and experimental data for the “full” and “half” shell configurations, respectively. Once again, the simulated transmission images agreed well with the experimental measurements, with PCC values of 0.999 and 0.997 for the “full” and “half” configurations.

FIG. 14.

Experimental vs simulated DT neutron transmission images for full HEU target.

FIG. 14.

Experimental vs simulated DT neutron transmission images for full HEU target.

Close modal
FIG. 15.

Experimental vs simulated DT neutron transmission images for half HEU target.

FIG. 15.

Experimental vs simulated DT neutron transmission images for half HEU target.

Close modal

As with the DD source, the spatial resolution capabilities of the DT-based system were explored in simulation. Figure 16 shows the resulting neutron transmission images for each incremental shell configuration.

FIG. 16.

Comparison of DT neutron transmission images for incremental shell configurations. The legend indicates the shell numbers that have been removed from the full configuration (shells 1–24).

FIG. 16.

Comparison of DT neutron transmission images for incremental shell configurations. The legend indicates the shell numbers that have been removed from the full configuration (shells 1–24).

Close modal

The transmission images appear to show even greater stratification compared to the DD data, and the 0.3–0.4 cm shell increments are once again well-resolved from each other. Due to the greater intensity of the DT source, this level of discrimination is achievable in about 80 seconds of measurement time.

The capabilities of the DT transmission imaging system were also tested under two shielding scenarios. Figure 17 shows the transmission image for the full HEU shell configuration with the 3.81 cm polyethylene shield in place. As a consequence of the high light output cut of 4.0 MeV, the transmission image is derived primarily from neutrons that do not lose energy by interacting in the shield. The neutron detection rate is dramatically reduced from the unshielded scenario, but any complications due to neutron moderation are eliminated by the use of a higher light output threshold. Though the difference in detected neutron rate between the detectors that are obscured by the target and those that are not is less pronounced in the overall image profile, the geometric shape of the target is still readily distinguishable, which suggests that detection of geometric anomalies would not be significantly inhibited by such a shield.

FIG. 17.

DT neutron transmission image for the full HEU sphere behind a 3.81 cm polyethylene shield.

FIG. 17.

DT neutron transmission image for the full HEU sphere behind a 3.81 cm polyethylene shield.

Close modal

Figure 18 shows the transmission image for the full HEU shell configuration surrounded by a 1-inch spherical layer of tungsten shielding. The profile is widened due to the greater radius of the target object, but there is no clear distinction between the central HEU core and the tungsten shielding. For 14.1 MeV neutrons, the total interaction cross-section of tungsten differs from that of 235U by only about 6%.27 Therefore, it is reasonable to assume that a sphere of pure HEU would generate a similar profile, which may not be distinguishable from the case where some of the HEU is replaced with a tungsten shield. Under such circumstances, additional measurements such as fission-neutron counting may be needed to differentiate the two scenarios.

FIG. 18.

DT neutron transmission image for the full HEU sphere with a 2.54 cm tungsten outer shielding layer.

FIG. 18.

DT neutron transmission image for the full HEU sphere with a 2.54 cm tungsten outer shielding layer.

Close modal

While geometric transmission imaging can detect material diversion that causes a change in the core geometry, differential determination becomes much more difficult if some or all of the SNM is replaced with a similarly dense, high-Z material. As shown in Figure 18, replacement of some or all of an HEU core with tungsten may still present a very similar neutron transmission profile. To guard against such cases, it is desirable that a verification system not only confirm the presence of SNM, but also provide an estimate of the quantity of fissile material. The energy spectrum for prompt neutrons from fast fission of 235U is well known, and prompt neutrons with energies as high as 8–10 MeV can be observed for incident neutron energies in the range of 0–7 MeV.28 Therefore, by imposing a lower detection threshold that is above the maximum light output for 2.45 MeV DD neutrons, only the signal from fission neutrons is measured, and the presence of SNM can be confirmed and quantified.

For the experimental data, fission-neutron thresholds were defined for each individual detector based on the measurements of the raw DD source with no target in place. The threshold was placed just above the endpoint light output for DD neutrons, providing a very conservative limit, and thus relatively high confidence that any measured neutrons originate only from fission. The measured fission-neutron rate for the “full” shell configuration (13.74 kg of HEU) was 0.733 ± 0.0082 neutrons/s, while the measured rate for the “half” configuration (10.04 kg of HEU) was 0.513 ± 0.0065 neutrons/s. A background rate of 0.18 ± 0.0135 neutrons/s above the fission-neutron threshold was also observed for the source-only data. Measurement of the tungsten sphere did not produce any appreciable fission-neutron signal above the background rate.

Figure 19 shows a comparison of experimental fission-neutron measurements with simulated results for each incremental shell configuration. The background rate from the source-only data has been added to all simulated results to better represent the laboratory setting. Since the count rate for high-energy fission neutrons is very low, and the probability is good that any of these neutrons would retain nearly all of their energy when scattering off the materials in the room, this measurement is very sensitive to room return. To mitigate this sensitivity, a concrete floor was included in the simulation, using the composition for ordinary concrete (NIST) as defined in the PNNL materials compendium.29 However, simulated transmission remained somewhat lower than the experimental data even after adjusting for background. As previously discussed, the increased rates in the experimental data are likely caused by the flux of scattered neutrons through the HEU object, which causes additional fission events beyond those caused by fast neutrons from the DD source.

FIG. 19.

Simulated and experimental fission-neutron count rates for various HEU shell configurations.

FIG. 19.

Simulated and experimental fission-neutron count rates for various HEU shell configurations.

Close modal

The full and half HEU configurations both exhibit an elevated fission-neutron count rate compared to the background and tungsten measurements, confirming the presence of fissile material. The two shell configurations are also well-resolved from each other, showing that the fission-neutron rate can be used to determine the approximate fissile material quantity. The experimental results represent cumulative data collected over a total live time greater than three hours, whereas the simulated results correspond to a measurement time of just over one hour. The trend of the simulated results in Figure 19 is encouraging, as the fission-neutron rate increases with HEU mass and suggests that differences of less than 1 kg are resolvable.

Though the system demonstrates the ability to discriminate two simple test cases based on fission neutron rate, it is important to note the limits of its applicability. In the presence of a significant amount of hydrogenous shielding material, high-energy fission neutrons would be moderated and thus could not be used to confirm the presence of SNM. Therefore, the current system requires some a priori knowledge that the test object is bare or minimally shielded. Furthermore, the single-view radiographic approach used in this experiment is limited to 2D measurements, and reconstruction of the full 3D geometry would require tomography. As a result, there could be some geometric ambiguities that may not be resolved by a 2D measurement, which could lead to inconsistencies between the inferred geometry based on radiographic measurements and the detected rate of high-energy fission neutrons.

The goal of this study has been to demonstrate the capabilities of a simple spectroscopic neutron transmission measurement system to realize crude geometric imaging of an SNM target, as well as confirm the presence of fissionable material and approximate its quantity. Using monoenergetic neutrons from DD and DT reactions, we successfully constructed low-resolution transmission profiles from which one could readily detect the removal of the central core (3.7 kg) of a 13.7 kg HEU sphere. Simulated results suggest that even greater spatial resolution could be achieved with this system, allowing for the removal of as little as 0.5–1 kg of material to be confidently detected. Spatial resolution could likely be improved further by increasing the number of image pixels (detectors), which would facilitate much greater distinction of relatively small diversions near the core of the target.

By applying high light-output discrimination to the transmitted neutron spectrum, we also successfully measured the emission rate of prompt fission neutrons, thereby confirming the presence of SNM and giving an estimate of its relative quantity. Removal of 3.7 kg of HEU from the original 13.7 kg target was readily detected on the basis of fission-neutron rate alone, and simulated results suggested that diversions of less than 1 kg could be resolved by the system. However, the measurements of fission neutrons were inhibited by the size of the array and its distance from the HEU target, which covered a very small fraction of the emission solid angle. A custom-designed fission-neutron sensing array placed much closer to the HEU target would likely be able to provide much greater resolution in the determination of fissile mass for much shorter measurement times.

Both geometric transmission imaging and estimation of fissile mass based on prompt fission neutrons may be useful in a treaty verification setting, such as a template-type zero-knowledge protocol. The transmission measurement approach we demonstrated could form the basis of differential measurements of the geometric profile and prompt fission-neutron emission rate of a purported SNM sample. Further refinements of the simple example demonstrated in this article could lead to a robust inspection tool for preventing the clandestine diversion of material during warhead dismantlement or for confirming compliance with certain nonproliferation treaty benchmarks.

The authors would like to thank Jesson Hutchinson of Los Alamos National Laboratory and John Mattingly of North Carolina State University for generous assistance with planning and executing the experiments at the Nevada National Security Site. This work has been supported by the Consortium for Verification Technology under Department of Energy National Nuclear Security Administration award number DE-NA0002534 and by the U.S. Department of Homeland Security award number 2014-DN-077-ARI078-02 and 2015-DN-077-ARI096.

1.
R. C.
Runkle
,
D. L.
Chichester
, and
S. J.
Thompson
,
Nuclear Instruments and Methods in Physics Research, A
663
,
75
(
2012
).
2.
C.
Comley
,
M.
Comley
,
P.
Eggins
,
G.
George
,
S.
Holloway
,
M.
Ley
,
P.
Thompson
, and
K.
Warburton
,
Confidence, Security & Verification: The Challenge of Global Nuclear Weapons Arms Control
(
Atomic Weapos Establishment
,
Reading, UK
,
2000
).
3.
R. T.
Kouzes
and
J. L.
Fuller
, in
Proceedings of Symposium on International Safeguards: Verification and Nuclear Material Security, October
(
International Atomic Energy Agency
,
Vienna, Austria
,
2001
).
4.
A.
Glaser
,
B.
Barak
, and
R. J.
Goldston
,
Nature
510
,
497
(
2014
).
5.
C. J.
Macgahan
,
M. A.
Kupinski
,
E. M.
Brubaker
,
N. R.
Hilton
, and
P. A.
Marleau
,
Nuclear Instruments and Methods in Physics Research, A
844
,
147
(
2017
).
6.
D.
Spears
, ed.,
Technology R & D for Arms Control
(
US Department of Energy, Office of Nonproliferation Research and Engineering
,
Washington DC
,
2001
).
7.
J. T.
Mihalczo
,
V. K.
Pare
,
E. D.
Blakeman
,
B.
Damiano
,
T. E.
Valentine
,
L. D.
Phillips
,
R. B.
Banner
,
D. B.
Bopp
,
T. R.
Chilcoat
,
J.
Declue
,
E. P.
Elliott
,
G. D.
Hackett
,
N. W.
Hill
,
D. J.
Nypaver
,
L. H.
Thacker
,
W. T.
Thomas
,
J. A.
Williams
, and
R. E.
Zumstein
,
Journal of Nuclear Materials Management
25
(
1997
).
8.
J. T.
Mihalczo
,
T. E.
Valentine
,
J. K.
Mattingly
,
J. A.
Mullens
, and
S. S.
Hughes
,
Active Neutron Interrogation for Verification of Storage of Weapons Components at the Oak Ridge Y-12 Plant
(
Oak Ridge National Laboratory
,
Oak Ridge, Tennessee
,
1998
).
9.
P. B.
Rose
,
A. S.
Erickson
,
M.
Mayer
,
J.
Nattress
, and
I.
Jovanovic
,
Scientific Reports
6
,
24388
(
2016
).
10.
J.
Rahon
,
A.
Danagoulian
,
T. D.
Macdonald
,
Z. S.
Hartwig
, and
R. C.
Lanza
,
Nuclear Instruments and Methods in Physics Research, A
820
,
141
(
2016
).
11.
S. V.
Naydenov
,
V. D.
Ryzhikov
, and
C. F.
Smith
,
Nuclear Instruments and Methods in Physics Research, B
215
,
552
(
2004
).
12.
T. A.
Wellington
,
B. A.
Palles
,
J. A.
Mullens
,
J. T.
Mihalczo
,
D. E.
Archer
,
T.
Thompson
,
C. L.
Britton
,
N. D.
Bull Ezell
,
M. N.
Ericson
,
E.
Farquhar
,
R.
Lind
, and
J.
Carter
,
Physics Procedia
66
,
432
(
2015
).
13.
P. A.
Hausladen
,
P. R.
Bingham
,
J. S.
Neal
,
J. A.
Mullens
, and
J. T.
Mihalczo
,
Nuclear Instruments and Methods in Physics Research B
261
,
387
(
2007
).
14.
P. A.
Hausladen
,
M. A.
Blackston
,
E.
Brubaker
,
D. L.
Chichester
,
P.
Marleau
, and
R. J.
Newby
,
Fast-Neutron Coded-Aperture Imaging of Special Nuclear Material Configurations
(
Idaho National Laboratory
,
Idaho Falls, ID
,
2012
).
15.
P. A.
Hausladen
,
M. A.
Blackston
, and
J.
Newby
,
Demonstration of Emitted-Neutron Computed Tomography to Quantify Nuclear Materials
, September (
Oak Ridge National Laboratory
,
Oak Ridge, Tennessee
,
2011
).
16.
P. A.
Hausladen
,
M. A.
Blackston
,
J.
Mullens
,
S.
Mcconchie
,
J. T.
Mihalczo
,
P. R.
Bingham
,
M. N.
Ericson
, and
L.
Fabris
,
Induced-Fission Imaging of Nuclear Material Induced
(
Oak Ridge National Laboratory
,
Oak Ridge, Tennessee
,
2010
).
17.
S. M.
Robinson
,
K. D.
Jarman
,
W. K.
Pitts
,
A.
Seifert
,
A. C.
Misner
,
M. L.
Woodring
, and
M. J.
Myjak
,
Nuclear Instruments and Methods in Physics Research, A
662
,
81
(
2012
).
18.
N.
Ensslin
,
M. S.
Krick
,
D. G.
Langner
, and
M. C.
Miller
,
Active Neutron Multiplicity Counting of Bulk Uranium
(
Los Alamos National Laboratory
,
Los Alamos, NM
,
1991
).
19.
I.
Israelashvili
,
C.
Dubi
,
H.
Ettedgui
,
A.
Ocherashvili
,
B.
Pedersen
,
A.
Beck
,
E.
Roesgen
,
J. M.
Crochmore
,
T.
Ridnik
, and
I.
Yaar
,
Nuclear Instruments and Methods in Physics Research, A
785
,
14
(
2015
).
20.
R. E.
Rothe
,
Extrapolated Experimental Critical Parameters of Unreflected and Steel-Reflected Massive Enriched Uranium Metal Spherical and Hemispherical Assemblies
(
Idaho National Laboratory
,
Idaho Falls, ID
,
1997
).
21.
S.
Agostinelli
,
J.
Allison
,
K.
Amako
,
J.
Apostolakis
,
H.
Araujo
,
P.
Arce
,
M.
Asai
,
D.
Axen
,
S.
Banerjee
,
G.
Barrand
,
F.
Behner
,
L.
Bellagamba
,
J.
Boudreau
,
L.
Broglia
,
A.
Brunengo
,
H.
Burkhardt
,
S.
Chauvie
,
J.
Chuma
,
R.
Chytracek
,
G.
Cooperman
,
G.
Cosmo
,
P.
Degtyarenko
,
A.
Dell’Acqua
,
G.
Depaola
,
D.
Dietrich
,
R.
Enami
,
A.
Feliciello
,
C.
Ferguson
,
H.
Fesefeldt
,
G.
Folger
,
F.
Foppiano
,
A.
Forti
,
S.
Garelli
,
S.
Giani
,
R.
Giannitrapani
,
D.
Gibin
,
J.
Gómez Cadenas
,
I.
González
,
G.
Gracia Abril
,
G.
Greeniaus
,
W.
Greiner
,
V.
Grichine
,
A.
Grossheim
,
S.
Guatelli
,
P.
Gumplinger
,
R.
Hamatsu
,
K.
Hashimoto
,
H.
Hasui
,
A.
Heikkinen
,
A.
Howard
,
V.
Ivanchenko
,
A.
Johnson
,
F.
Jones
,
J.
Kallenbach
,
N.
Kanaya
,
M.
Kawabata
,
Y.
Kawabata
,
M.
Kawaguti
,
S.
Kelner
,
P.
Kent
,
A.
Kimura
,
T.
Kodama
,
R.
Kokoulin
,
M.
Kossov
,
H.
Kurashige
,
E.
Lamanna
,
T.
Lampén
,
V.
Lara
,
V.
Lefebure
,
F.
Lei
,
M.
Liendl
,
W.
Lockman
,
F.
Longo
,
S.
Magni
,
M.
Maire
,
E.
Medernach
,
K.
Minamimoto
,
P.
Mora de Freitas
,
Y.
Morita
,
K.
Murakami
,
M.
Nagamatu
,
R.
Nartallo
,
P.
Nieminen
,
T.
Nishimura
,
K.
Ohtsubo
,
M.
Okamura
,
S.
O’Neale
,
Y.
Oohata
,
K.
Paech
,
J.
Perl
,
A.
Pfeiffer
,
M.
Pia
,
F.
Ranjard
,
A.
Rybin
,
S.
Sadilov
,
E.
Di Salvo
,
G.
Santin
,
T.
Sasaki
,
N.
Savvas
,
Y.
Sawada
,
S.
Scherer
,
S.
Sei
,
V.
Sirotenko
,
D.
Smith
,
N.
Starkov
,
H.
Stoecker
,
J.
Sulkimo
,
M.
Takahata
,
S.
Tanaka
,
E.
Tcherniaev
,
E.
Safai Tehrani
,
M.
Tropeano
,
P.
Truscott
,
H.
Uno
,
L.
Urban
,
P.
Urban
,
M.
Verderi
,
A.
Walkden
,
W.
Wander
,
H.
Weber
,
J.
Wellisch
,
T.
Wenaus
,
D.
Williams
,
D.
Wright
,
T.
Yamada
,
H.
Yoshida
, and
D.
Zschiesche
,
Nuclear Instruments and Methods in Physics Research, A
506
,
250
(
2003
).
22.
NEUTRON/GAMMA PSD LIQUID SCINTILLATOR EJ301, EJ309 (Eljen Technology, 1300 W. Broadway, Sweetwater, TX 79556, 2016).
23.
A.
Enqvist
,
C. C.
Lawrence
,
B. M.
Wieger
,
S. A.
Pozzi
, and
T. N.
Massey
,
Nuclear Instruments and Methods in Physics Research, A
715
,
79
(
2013
).
24.
G.
Dietze
and
H.
Klein
,
Nuclear Instruments and Methods
193
,
549
(
1982
).
25.
J. R.
Verbus
,
C. A.
Rhyne
,
D. C.
Malling
,
M.
Genecov
,
S.
Ghosh
,
A. G.
Moskowitz
,
S.
Chan
,
J. J.
Chapman
,
L.
de Viveiros
,
C. H.
Faham
,
S.
Fiorucci
,
D. Q.
Huang
,
M.
Pangilinan
,
W. C.
Taylor
, and
R. J.
Gaitskell
,
Nuclear Instruments and Methods in Physics Research, A
851
,
68
(
2016
), arXiv:1608.05309.
26.
R.
Remetti
,
L.
Lepore
, and
N.
Cherubini
,
Nuclear Instruments and Methods in Physics Research, A
842
,
7
(
2017
).
27.
M. B.
Chadwick
,
M.
Herman
,
P.
Oblozinsky
,
B.
Pritychenko
,
G.
Arbanas
,
R.
Arcilla
,
R.
Brewer
,
D. A.
Brown
,
R.
Capote
,
A. D.
Carlson
,
Y. S.
Cho
,
H.
Derrien
,
K.
Guber
,
G. M.
Hale
,
S.
Hoblit
,
S.
Holloway
,
T. D.
Johnson
,
T.
Kawano
,
B. C.
Kiedrowski
,
H.
Kim
,
S.
Kunieda
,
N. M.
Larson
,
L.
Leal
,
J. P.
Lestone
,
R. C.
Little
,
E. A.
Mccutchan
,
R. E.
Macfarlane
,
M.
Macinnes
,
C. M.
Mattoon
,
R. D.
Mcknight
,
S. F.
Mughabghab
,
G. P. A.
Nobre
,
G.
Palmiotti
,
A.
Palumbo
,
M. T.
Pigni
,
V. G.
Pronyaev
,
R. L.
Vogt
,
S. C.
van der Marck
,
A.
Wallner
,
M. C.
White
,
D.
Wiarda
, and
P. G.
Young
,
Nuclear Data Sheets
112
,
2887
(
2011
).
28.
D. G.
Madland
,
New Fission-Neutron-Spectrum Representation for ENDF
, April (
Los Alamos National Laboratory
,
Los Alamos, NM
,
1982
).
29.
R.
McConn
, Jr.
,
C.
Gesh
,
R.
Pagh
,
R.
Rucker
, and
R.
Williams
, III
,
Compendium of Material Composition Data for Radiation Transport Modeling
, rev. 1 ed. (
Pacific Northwest National Laboratory
,
2011
).