Assessing the suitability of three proxy sources for the development of detectors of special nuclear materials

Numerous techniques and equipment have been developed to provide a capability for the detection of special nuclear materials (SNM), but due to the necessary security measures surrounding these materials alternate, or proxy, neutron sources are often utilised in their stead. In this paper we report the neutron and gamma pulse shape discrimination response of plastic scintillator to mixed neutron/gamma beams produced from two radionuclide neutron sources, and also from an SNM source of weapons-grade plutonium. We discuss the suitability of using radionuclide sources, with appropriate shielding configurations as proxy sources for SNM. A 3σnth-γ discrimination level has been achieved for an SNM source at a low-level energy threshold of ∼220 keVee when a shielding configuration of 5 cm of lead was implemented. Varying amounts of lead and high-density polyethylene (HDPE) shielding were also investigated with the 3σ limit being reached by ∼240 keVee. This work shows that an AmBe neutron source serves as an appropriate SNM proxy achieving a comparable value for figure of merit above ∼1 MeVee. For energies below 1 MeVee down to ∼100 keVee a closer approximation of the expected FoM for SNM can be attained when using 252Cf as a proxy source or by utilising an ‘enhanced’ AmBe source with the addition of a further low energy γ ray source.

Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
The use of neutron discriminating plastic scintillators, such as the material produced by Eljen (most recently as EJ-276) [1,2], coupled with a silicon photomultiplier (SiPM) readout has become increasingly commonplace in detection of special nuclear materials (SNM).
In this work we study the neutron/gamma pulse shape discrimination (PSD) properties of a plastic scintillator-SiPM system [7,8] to assess the suitability of various radionuclide neutron sources as proxy sources for SNM.Various groups have previously reported the performance of PSD plastic scintillator using isotopic neutron sources of the (α,n) reaction type, such as AmBe [9][10][11], AmLi [12], 238 PuBe [13], and 238 PuC [14].However, the energy spectra of both neutrons and gammas emitted from such sources are quite distinct from those of special nuclear materials (SNM) [3][4][5], which can lead to an overestimation of neutron-gamma separation performance when used in isolation.In order to assess a system's ability to identify SNM, it is often useful to use 'proxy' sources where a sample of SNM may not otherwise be available, namely 252 Cf [15][16][17], or one of the previously mentioned (α,n) sources.However, in these cases one must carefully consider the respective differences between the sources, namely, the energies of emission, and the multiplicities of those emitted particles.Similarly, it should not be overlooked that detector volume can play a role beyond simple detection efficiency with attenuation and timing becoming important factors for large-scale systems.Here we report a comparative study of the performance of a PSD plastic scintillator using two different radionuclide neutron sources 241 AmBe and 252 Cf plus a sample of weapons-grade plutonium (<7% 240 Pu [18,19]) that will be referred to as the SNM source.
The neutron energy spectrum resulting from nuclear fission [20][21][22] follows a smooth Watt or Maxwellian distribution, whereas (α,n) sources, such as the AmBe used in this work, typically have greater structure in their neutron spectra [23,24].The spectra reproduced in figure 1 show the different energy spectra from of these common neutron sources.The AmBe source has a higher average neutron energy than the equivalent fission sources, with structure in the neutron energy spectrum due to the various transitions involved in the AmBe source reactions [3,[24][25][26].The fission neutron spectra are distinct from that of the AmBe source and share similar distributions to each other, with the peak intensity at ~1 MeVee [21].For general use as a neutron source, 252 Cf was highlighted as having an advantage over (α,n) sources due to its 'soft neutron spectra and low gamma activity' [27], although in the case of acting as a proxy for SNM, the reduced γ-ray emission may be considered a disadvantage.
A second key property of the fissile materials, that of the neutron multiplicity, is shown for key radionuclides in table 1. Clearly this is not a characteristic of an (α,n) source such as the AmBe, but for spontaneous fission of 240 Pu and 252 Cf (and to a lesser extent with its extremely low cross section, 239 Pu) there are more neutrons available per disintegration.Other techniques and algorithms focus on this property, separating based on the 'burstiness' of detected events [28] while PSD is not concerned with intra-event timing in its analysis beyond the avoidance of pulse pile-up for extremely high rates.Example neutron spectra (normalised to their peak emission) reproduced from [3], [4], and [5].Note the use of MeV electron equivalent here (and throughout) due to the light yield dependence on particle type and the respective quenching effects [6].

Source
Mean multiplicity Neutron yield (s −1 g −  Detection of nuclear materials, such as uranium and plutonium, is rarely in the context of a clean signature performed in ideal laboratory conditions.The anticipated search location may well contain some level of moderation or shielding, whether intentional or otherwise.Therefore not only is it important to assess the performance of any detection system on SNM itself, but also using a variety of high-Z and hydrogenous shielding configurations in order to vary the γ and neutron flux incident on the detector respectively. In this work we discuss PSD tests on plastic scintillator using three different radio-isotope sources; a fast neutron/gamma AmBe source, and two fission sources containing 252 Cf and special nuclear material (SNM).We also investigate the effectiveness of adding additional gamma sources in combination with the neutron sources to more closely replicate the spectra and detector neutron/gamma separation performance when in the presence of an SNM source.

Experimental setup
All measurements in this work used a detector consisting of PSD-sensitive scintillator directly coupled to SensL SiPMs.The detector was used to measure the energy response and neutrongamma discrimination performance for each of the neutron sources studied.As potential 'SNM proxy' sources we used AmBe and 252 Cf radionuclide sources.The AmBe source had an activity of ~18 GBq and was contained within a water tank of dimensions 80 cm × 76 cm × 94 cm.An air-tube was inserted to select fast neutrons, resulting in a source-detector offset of ~50 cm.Additionally, the standard configuration included 5 cm of lead shielding in order to reduce the γ-ray signal and achieve approximately a 50:50 n-γ ratio, full details can be found in earlier works [7,8].The 252 Cf source used in this study was mounted in air, offset from the detector due to the 5 cm lead shielding, and had an activity of 238 kBq.
Measurements were also taken using an enhanced AmBe source that consisted of an additional 118 kBq 133 Ba source placed in close proximity to the detector whilst setup as described previously.The addition of the 133 Ba signature, in particular the strong 356 keV γ ray gives a good approximation of the γ rays produced at low energies by 239 Pu, a (non-exhaustive) selection of prominent de-excitations are included as table 2.
The system was independently calibrated using laboratory γ ray sources at each measurement stage to ensure any potential drift in gain was accounted for.Due to the interaction of γ rays with plastic scintillators dominated by Compton scattering, it is not possible to resolve the primary photopeaks of commonly used sources ( 22 Na and 137 Cs) and so calibration is performed using the Compton edge of these γ rays, the full description is found in a previous publication [8].Additionally, the current drawn by the SiPM was monitored throughout all experimental stages.To enable direct comparisons to be made between datasets measurements were taken with same level of shielding for the different sources such that the nature of the emitted radiation from each source was the only significant variable.
The performance of the detector was compared to measurements obtained using an SNM source consisting of 75 g of Pu with isotopic vectors consistent with those of low burn-up or weapons grade [31] plutonium (WGPu) at ~6% 240 Pu and ~93% 239 Pu.The majority of the neutron yield for plutonium sources is a result of the 240 Pu content [32], as shown in table 1 alongside yields for the other sources, whereas the γ signature is due to the presence of 239 Pu, its decay chain, and the 241 Am and 237 U daughters of 241 Pu [33,34].As well as the direct comparison of 5 cm of lead shielding further configurations of different thicknesses of both γ ray and neutron shielding as well as stand-off distances were investigated.

Silicon photomultiplier
The SensL J-series SiPM was selected for its very low dark current and excellent photon detection efficiency (PDE) in the emission range of interest.The performance of this SiPM had been previously tested and shown to give good n-γ separation with plastic scintillator [8].The 6 mm J-series was used for all measurements, with two SiPM configurations used in this work.The MicroFJ-60 035-SMA is an evaluation board prepared by SensL housing a single 6 mm SiPM and was used with an appropriated sized scintillator described in the following section, these data are tagged as 'singles'.Additionally, a 2×2 array, the ArrayJ-60 035-4P-PCB, was mounted on bespoke read-out circuitry with all channels summed.Measurements with this detector are referred to as 'quads'.
Each scintillator material was tested for cubes of 6 mm and 15 mm for use with the single and quad SiPM respectively.All faces were highly polished (0.3 µm), with one face optically coupled to the SiPM with Silicone High-Vacuum grease and the remaining faces either painted with EJ510 titanium oxide paint in the case of the EJ-299-34 scintillator, or wrapped in several layers of PTFE tape for the EJ-276 [37].We have previously shown that there is negligible difference between the performance of the two reflective coatings when tested with identically prepared samples of EJ-299-34.

Data acquisition
Measurements were taken by acquiring waveform pulses in an event-by-event mode, using either a CAEN V1730C digitiser (AmBe data) or a CAEN DT5730B digitiser (fission source data).Both digitisers incorporated a 14-bit ADC operating over a dynamic range of 0.5 V pp at a sample rate of 500 MHz.The CAEN digitisers utilise a multi-buffer technique in order to minimise deadtime effects [38].However, for the unshielded SNM measurements the count-rate was sufficiently high that the detector was re-positioned further from the source to minimise deadtime.
The V1730C was connected via optical link cable to a CAEN A3818 optical controller in the acquisition PC, whereas the DT5730B was connected to its control computer with a standard USB cable.

Pulse shape discrimination
The PSD technique exploits the characteristic interaction mechanisms in order to identify the type of incident radiation.Full details of the phenomenon have been discussed in many previous works [39][40][41]; for scintillation light which contains emissions of different time constants 'the fraction of light that appears in the slow component often depends on the nature of the exciting particle' [6].Particles with higher linear energy transfer coefficients, such as neutrons in this case, produce scintillation with a greater proportion of the slow component, and thus can be distinguished from γ rays via a variety of algorithms.
In this study we used a charge comparison technique to determine a PSD parameter for each event [7,8], which uses the ratio of the integrated charge in the prompt portion of the pulse to The charge comparison method.Each pulse is processed in the same way, whereby the charge, Qs, in a defined 'short' time window from t 0 to ts, is compared to that of the whole pulse, where Q l is the charge integrated from t 0 to t l .Note how the regions of Qs and Q l overlap for this algorithm.
the total integrated charge.The value of this ratio is different for incident neutrons compared to γ rays due to the respective interaction mechanisms of each species [41,42].This is shown graphically in figure 2 and described by: where t 0 is the start time of the two gates, t s the end time of the short gate, and t l the end time of the long gate.The PSD parameter is then determined by a ratio of these components: Additionally, the quantity Q l , the long gate charge, is proportional to the energy of the event, which, when plotted against the PSD parameter forms a distinctive forked histogram (see later figure 3) from which the degree of neutron-γ separation can be determined.A projection of the PSD parameter shows a double Gaussian distribution, and the degree of neutron-γ separation is described by a figure of Merit (FoM) parameter [43] defined as: where S is the centroid separation, Γ n is the FWHM of the neutron peak, and Γ γ the γ FWHM.

Results and discussion
We describe the PSD performance of the scintillator in the presence of neutron-γ mixed radiation fields from the various sources, with the γ and neutron spectra shown independently.
The detectors' tolerance to various shielding configurations and the impact of shielding upon the neutron-γ ratio will be assessed.The effectiveness of the various radionuclide sources in acting as SNM proxy sources will be described.

PSD performance
Using the charge comparison technique the PSD parameter for each event is plotted against the total energy of the pulse as shown in figure 3 for the mixed radiation fields of each source as well as for a γ-only source of 22 Na to provide a comparison.For this algorithm, a mixed radiation field exhibits two loci, where the upper locus contains the neutron events whereas the lower locus contains the γ ray events.It can be seen that γ only events from a 22 Na source aligns with the lower locus of the other plots.The data displayed in figure 3 were obtained using the 15 mm cube of EJ-276 with the 'standard' 5 cm lead shielding configuration.The degree of neutron-γ separation is calculated from the PSD spectrum (a Y-axis projection of the 2D histogram), for various energy thresholds.Full details of our use of the figure of merit (FoM) calculation as defined by [43] are given in [7,8].As an example the projection of the AmBe PSD parameter and its double peak form, is shown in figure 4 with 100 keV ee energy threshold bands applied.It is clear that while the centroid of the γ peak remains relatively constant the neutron peak has a noticeable shift in centroid position.
Table 3 shows the measured PSD performance obtained for Eljen EJ-299-34 and EJ-276 scintillators when irradiated with a mixed fast neutron/gamma field from the AmBe source, using 'single' 6 mm and 'quad' 15 mm scintillator cubes.Good PSD performance is generally defined as a 3σ separation of the neutron and γ peaks [1,44], which equates to a FoM of 1.27.The 15 mm cube generally provides poorer PSD performance than the smaller scintillator, which fits the commonly observed trend in these plastic scintillators of worsening PSD performance as the volume of the scintillator increases [45,46].It can be seen from table 3 that for the EJ-299 material the PSD performance at a threshold of 500 keV ee fails to meet the 3σ criterion.In contrast, the PSD performance for the EJ-276 scintillator significantly exceed the 3σ criterion at all measured energy values.When considering the 'improvement' ratio of the FoM values between EJ-276 and EJ-299, for the small detectors the improvement is a constant factor of 1.4 for all three energy values.For the large detector the improvement in performance of EJ-276 compared to EJ-299 is even more noticeable, with a ratio of 1.9 at 500 keV ee , reducing to 1.4 at 1.5 MeV ee .Figure 5 shows the FoM values obtained from the 15 mm EJ-276 detector for source, as a function of energy up to 1 MeV ee .In general the same trend of increasing FoM as a function of increasing energy is observed across all four sources.The measured FoM from 252 Cf closely approximates the separation achieved for SNM over the full energy.In contrast, for energies below ~800 keV ee the FoM values obtained from the AmBe mixed radiation field are significantly higher than for the other sources.However, once the ~1 MeV ee threshold is reached, the four sets of FoM values become comparable.This convergence of source FoM is also shown in table 4 highlighting some key FoM values for the 'quad' detector with each  source.At 100 keVee the required level of separation is only achieved in AmBe source's radiation field, indicating an over-estimate of separation when used as a proxy at these energies.However, the difference between the degree of n-γ discrimination for an (α,n) source and a fission source reduces at 500 keVee, and is further reduced by the 1 MeVee threshold.
To explain the variation in measured FoM between the different sources (shown in figure 5  and table 4) it is necessary to examine the neutron and γ ray spectra in each case.These have been extracted from each of the 2D histograms and are displayed in figures 6 (neutron) and 7 (γ ray).Due to the nature of the proton recoil mechanism that is used to detect fast neutrons, the raw deposited energy spectrum from plastic scintillators do not directly reflect the incident neutron energy spectrum.For this reason, the measured neutron spectrum from the AmBe source shown in figure 6 does not show the structure that is visible in figure 1.However, the measured neutron energy spectrum does show a distribution of higher energy neutrons in the AmBe spectrum which is not present in the fission sources which follow the Watt distribution [20].
From the γ spectra shown in figure 7 it is clear that the most significant difference in spectral shape between the sources is the different intensities of γ rays at energies below approximately 400 keV.The SNM source shows a large contribution of low energy γ rays which is primarily a result of the decay chain products of 239 Pu [33,34].This feature explains the reduction in FoM, as displayed in table 4 and figure 5, for the SNM sample compared to that of the 252 Cf source at the lower energy thresholds, and the subsequent improvement in FoM of the SNM  source as energy increases.Similarly, the mixed AmBe and 133 Ba source shows a significantly higher intensity of low energy γ rays, e.g.due to the presence of 356 keV γ rays from 133 Ba and a resulting reduction in the FoM at low energies.

Detection of SNM
Further studies were made of the detectors' performance with the AmBe and 252 Cf sources using a variety of shielding configurations in addition to the 'standard' arrangement described previously (5 cm lead).These configurations are listed in table 5 along with a pictogram demonstrating the setup.Due to time constraints of the measurements, the counting durations were not constant for each configuration, with times ranging from ~10 min to overnight  runs of ~20 h.The shortest runs were those dominated by the significant γ flux of unshielded measurements where data rates were highest.The longest duration measurements were the three configurations: (i) 5 cm lead, (ii) 10 cm lead, and (iii) 5 cm lead plus 5 cm HDPE, which enabled the full PSD analysis to be performed.These results are shown in figure 8.These data show a consistent trend in FoM for the three shielding configurations, although at energies above about 1 MeV ee the reduction in statistics increases the error bars that are derived from the double Gaussian fit.This is particularly pronounced for the 10 cm Pb shielding dataset where the number of neutron events at high energy drops to ~10 per bin.The datasets for the remaining configurations were dominated by the significant γ-ray flux incident on the detector.As a result, for many of these configurations, it was not possible to achieve statistically meaningful FoM, particularly at higher energy thresholds.However, in all of these cases events were identified within the neutron locus established with the standard shielding configuration that were cleanly separated from the PSD γ-ray peak.This is shown in figures 9 and 10, both configurations with 'short' durations of ~1 h.
Figure 10 was acquired using an SNM source with Pb shielding of only 1 cm thickness.Compared to 5 cm of Pb shielding, this configuration only weakly attenuated the γ rays and the γ spectrum of the source becomes apparent as shown in figure 10.There are features at ~350 and 730 keV ee which corresponds broadly to the Compton edges of known energies for WGPu γ lines [33,47].However, it is the nature of plastic scintillators that such an assessment will always be an approximation.

Conclusion
We have shown in this work the functionality of a compact neutron-γ discriminating detector for use in detecting SNM.The relative response to a number of different neutron sources has been investigated and the degree of neutron/γ discrimination with respect to energy has been determined.We conclude that depending on the intended purpose 252 Cf and AmBe are efficient proxies when an SNM source is unavailable.However, care must be taken to consider the energy range of interest, with AmBe sources providing a more intense distribution of higher energy neutrons compared to SNM.In terms of γ emission, the low-energy γ rays emitted from SNM are under represented by AmBe and fission sources and these proxies will typically over-estimate the FoM performance of a system intended to detect SNM.To mitigate this issue a more realistic signature can be made using the addition of another radionuclide to the neutron source.The chosen enhancement depends somewhat on the target material, with 57 Co often used where highly enriched uranium (HEU) is of concern [48], but for 239 Pu this role is performed by 133 Ba [49], and this work has demonstrated that similar FoM results can be obtained from such a combination proxy source.The introduction of the 133 Ba into the spectrum has the effect of reducing the n-γ ratio from that of pure AmBe, to a closer approximation of SNM (in our case the respective activities more closely aligned to the dual shielded 5 cm Pb/HDPE measurement).However, this adjustment is no longer apparent after ~400 keV ee , a combination of the 356 keV γ ray and the harder neutron spectra of AmBe in comparison to that of 239 Pu.Further enhancement could be provided by considering narrower energy regimes, and providing a proxy source at each point.This has been tested by other groups [48,49], although with the plastic scintillator tested here, the low resolution prevents spectroscopic information being discernable.The further addition of 241 Am, 109 Cd, and 137 Cs was shown to provide a good replication of plutonium spectra [49].Conversely, here we have looked at the variation in detector response where a real SNM source is shielded, approaching the problem from the opposite direction.
With the successful performance of this system in mind, the next stages of development must surely be in the direction of deployable equipment.Current radiation portal monitors use substantially larger volumes of scintillator than have been investigated in this work, and it has been shown previously that scintillator size has a direct effect on PSD performance [45,46].The degradation observed in PSD as detector volume increases is a combination of parameters: the optical attenuation as the photon transverses the scintillator; the increased variation in transit time due to multiple reflections within the material; and the relatively similar durations of neutron and γ ray time constants.The increased attenuation results in smaller pulses and therefore less light/charge available for collection, which, in analogy to a low energy interacting particle, can be seen from figure 3 makes separation more difficult.The variation in transit time, coupled with the respective time constants [8], effectively smears out the average pulse shapes, resulting in less defined, and therefore less separable neutron-and γ-ray indicative signals.Ongoing parallel research threads are investigating larger scale PSD systems [50].
Similarly, this work made use of traditional crate-based data acquisition (DAQ) systems, which clearly are unsuitable for any field role.To be able to deploy a detector such as that showcased here it will be necessary for a compact DAQ to be developed alongside the frontend scintillator-photodetector system.

Figure 1 .
Figure 1.Example neutron spectra (normalised to their peak emission) reproduced from[3],[4], and[5].Note the use of MeV electron equivalent here (and throughout) due to the light yield dependence on particle type and the respective quenching effects[6].

Figure 2 .
Figure 2.The charge comparison method.Each pulse is processed in the same way, whereby the charge, Qs, in a defined 'short' time window from t 0 to ts, is compared to that of the whole pulse, where Q l is the charge integrated from t 0 to t l .Note how the regions of Qs and Q l overlap for this algorithm.

Figure 3 .
Figure 3.The PSD Vs Energy plots acquired from the 15 mm EJ-276 detector.From top to bottom: reference22 Na, AmBe, 252 Cf, SNM, and enhanced AmBe (the Z-axis range for the enhanced AmBe has been adjusted to allow the PSD structure to be visible.

Figure 4 .
Figure 4.The PSD projections for AmBe source data after the application of 100 keVee energy gates.

Figure 5 .
Figure 5.The lower energy (0-1 MeVee) FoM for the data displayed in figure 3. The dashed line corresponds to the FoM = 1.27 threshold.

Table 5 .
Shielding configurations investigated throughout this work.The pictogram shows an image approximating the experimental setup, where the green star represents the neutron source, grey and blue boxes show where lead or HDPE was used respectively, and the detector position is given by the red trapezium.For the enhanced AmBe/ 133 Ba measurement, the 133 Ba source was placed adjacent to the detector and not shielded.

Figure 8 .
Figure 8.The low-energy FoM for different shielding configurations.Below 200 keVee it was not possible to achieve a good FoM fit for the '10 cm Pb' and '5 cm Pb and 5 cm HDPE' datasets.

Figure 9 .
Figure 9.The PSD V Energy histogram for 10 cm HDPE neutron shielding of the SNM source.The red and green gates were determined using the 'standard' shielding configuration data (5 cm Pb), with red the likely neutron events, and green surrounding likely γ-ray pulses.

Figure 10 .
Figure10.The PSD V Energy histogram for 1 cm Pb shielding of the SNM source.The red and green gates were determined using the 'standard' shielding configuration data (5 cm Pb), with red the likely neutron events, and green surrounding likely γ-ray pulses.

Figure 11 .
Figure 11.The γ-ray spectra from the 1 cm Pb shielding SNM source dataset shown in figure 10.

Table 2 .
[29,30,nt γ ray decay energies for 239 Pu[29,30, 47].The energy resolution of the plastic scintillator is generally insufficient to distinguish between peaks of these and their less intense satellite emissions.

Table 3 .
[8]ected figures of Merit for the reported scintillator-photodetector combinations (6 mm data previously published in[8]) highlighting the performance increase of the new scintillator as well as the reduction in FoM for larger sizes.

Table 4 .
Selected 'low energy' figures of Merit for the EJ-276 Quad detector in the mixed radiation fields of different sources.